Share

Readers keep asking some version of the same question: what do I make of the Democratic Socialists of America (DSA), of democratic socialism, and of the Democratic progressive wing?

Those are three entirely different things. The first is a relatively small organization of roughly 120,000 people whose new platform proposes rewriting the Constitution, has some pretty crazy ideas(including wearing masks in July), and receives wildly outsized attention. The second is the unfortunate label Bernie Sanders has used for forty years to describe what is essentially FDR’s agenda going back to the 1940s. The third is a very large wing of the Democratic Party that has existed forever and is now, for good reason, ascendant once again.

They get conveniently conflated by the pundit class and right-wing demagogues, while the Democrats who ought to be drawing clear lines, and with the exception of the likes of James Carvell and Rahm Emanuel, have decided it’s safer to say nothing or equivocate when asked—a big mistake.

Where I stand: I would never call myself a democratic socialist, and I completely reject the DSA’s platform—this stuff is largely “Cuckoo for Cocoa Puffs.” I’d like the organization marginalized rather than courted. These people are NOT Democrats.

But I also refuse to accept that what gets called progressive is left-wing. Guaranteed healthcare, a living wage, housing people can afford, and retirement security aren’t European imports; that is the Democratic Party’s own inheritance, and I’ll show you exactly where it comes from.

Wisconsin and the Primary Reality

It starts with why I stayed up Tuesday night watching returns from a state I don’t live in. I wanted Francesca Hong, the remarkably lightweight DSA candidate running in the Democratic primary for Governor of Wisconsin, to lose.

She led by 22 points in State Navigate’s final poll. The AP called it for David Crowley at 3:34 in the morning by just under 4,000 votes.

Here’s the part worth understanding: the polls got her share nearly right—they had her near 40 percent and she finished at 39. Her support never collapsed. What moved was Crowley, who beat his estimate by more than 30 points as a third of the electorate, undecided until Election Day, broke almost entirely against her. She didn’t lose ground; she ran out of room.

She ran out of room because of her own words, surfaced in the closing days. Six-year-old posts called to defund and abolish the police and to “cancel Thanksgiving.” Three months ago, she said her perfect world would be a world without prisons. In 2021, she wrote that she wanted to abolish the Senate. All of it hers, all of it reported by NBC News the week before the vote.

Thirty-nine percent potentially wins a crowded primary. It wins nothing in November. This is all nuts, of course, and it immediately became fodder for the entire Republican Party and the right-wing media ecosystem that trolls for exactly this type of fringe politics to amplify everywhere. You bet I wanted her to lose.

The Actual Scale of the Movement

Some arithmetic might be helpful to put this in perspective. There are roughly 519,682 elected offices in the United States. DSA members and endorsees hold about 250 of them. That is one in two thousand. Almost all of it is hyper-local: 96 city councilors and county commissioners, eight mayors, and the rest scattered across school boards and municipal seats. They have one member of Congress, probably 2-4 after November. Zohran Mamdani is the outlier, not the pattern. By comparison, the Socialist Party of the 1910s had more than 130 mayors and a presidential candidate, Eugene Debs, who took nearly a million votes in 1912.

There is no socialist takeover of American politics.

The opposite error, however, is just as easy to make. Although the organization is small, the sentiment underneath it is not. More than 35 DSA-backed candidates have won primaries this cycle, and Mamdani governs New York City with a 58 percent approval rating. A large share of Democratic voters want change on a generational scale for reasons I find hard to argue with. The question is what kind. And, if the party and it’s most compelling candidates don’t listen and don’t respond forcefully, we’ll be looking at an even bigger problem down the road.

Read the Actual Platforms

To see how far the conversation has drifted, you have to look at the actual history and platforms.

Roosevelt’s. In his January 1944 State of the Union, FDR proposed a Second Bill of Rights: a job; earnings enough for food, clothing, and recreation; a fair return for farmers; freedom from monopoly for businessmen large and small; a decent home; medical care; protection against the economic fears of old age, sickness, and unemployment; and a good education. Medicare for All is the sixth item(something Truman tried to make happen). A living wage is the second. Housing is the fifth. Free college is the eighth. It isn’t an import from Europe and it isn’t new—it’s a Democratic president’s platform from eighty-two years ago, and he asked Congress to legislate it. No constitutional amendment. No nationalization.

Sanders’. Across forty years, his platform is Roosevelt’s list with Scandinavian trim. He speaks glowingly about the Scandinavian model—which is, by every account including their own, a capitalist economy with a strong social safety net and unusually happy inhabitants. (I highly recommend a trip). But by asking Americans to judge his politics against countries most of them will never set foot in, how the hell is anybody supposed to know what he’s talking about? He has never proposed nationalizing an industry other than health insurance, and he has never been a member of the DSA. He simply picked the wrong word to describe what he believes, producing years of confusion and backlash now that have nothing to do with his actual agenda.

The DSA’s. On July 14th, the group published a program called Workers Deserve More. Much of it starts with Roosevelt but with much, much sharper edges—a 32-hour workweek at full pay, free college, universal rent control, and a federal jobs guarantee. Then it turns hard left. It calls for public ownership of the largest corporations, abolishing the Senate, and replacing the President and Supreme Court with an executive and judiciary subordinate to Congress. The stated goal is to “draft a new constitution, and create a democratic socialist republic.” Elsewhere: abolishing police and prisons, granting amnesty with visa caps eliminated, and ending all military and economic aid to Israel.

This is truly crazy stuff. I reject all of it, and you don’t need to be a moderate to do the same. A government that owns the largest firms, guarantees every job, sets every rent, and answers to no court it doesn’t control is a something akin to Stalin’s Russia.

Michigan and the Alternative Path

Eight days before Wisconsin, another primary showed a potential alternative approach for the progressive wing of the party.

I owe readers a scorecard here. The day before the Michigan vote, I wrote that the numbers landed on Haley Stevens—she led the polling by four to seven points, led Black voters 67 to 21, and had the party apparatus behind her. That’s what the AI model I use predicted. I forecasted Abdul El-Sayed would win anyway, because I distrust likely-voter screens when a candidate draws that much grassroots attention against that much money.

My gut was right, though not by much. El-Sayed won by about a point against roughly $65 million in outside spending.

Here’s what he did that Hong didn’t. Asked about the socialist label constantly, he gave the same answer every time, months before anyone could force it out of him. At the July 27th debate: “I am not a socialist. I am a capitalist who understands how capitalism works.” On Detroit’s Fox 2: “The biggest threat to capitalism was never the idea of government regulation, it was always monopoly.”

That monopoly line is Roosevelt’s fourth right almost verbatim—the most conservative sentence in the 1944 speech, and the one El-Sayed reaches for when he needs it most.

He doesn’t fit my frame cleanly, and I’ve written at length about where we part on Israel. But he’s proven that a progressive can draw a hard line against economic absurdity without abandoning working-class populism. Winning the Senate in November is what matters. Whether he survives statewide in Michigan against a wall of super PAC spending is another matter, including a reload from AIPAC, but unlike Hong, he gave himself a fighting chance by anchoring his campaign in the American tradition.

Why the Young Left is Angry

The easy move is to laugh at a bunch of young people, mostly college grads, crowded into a church basement to discuss what’s wrong with this country. That move is wrong.

Gallup’s August survey found 54 percent of Americans hold a positive view of capitalism, the lowest since they began asking in 2010. Among adults 18 to 34, it’s 43 percent.

Now look at the number that goes even deeper: in a December Gallup survey, 81 percent of Americans viewed “free enterprise” positively—against 54 percent for “capitalism.” Same respondents, same questionnaire, different word. People have not turned against open markets. They’ve turned against what they’ve watched this economic system do.

Take someone born in 1995. They were twelve when the 2008 financial crisis hit—old enough to watch a parent lose a house and to notice that essentially nobody responsible went to prison. They were told to go to college, did, and carry debt they can’t clear. First-time buyers have fallen to 21 percent of the housing market, an all-time low. They watched a pandemic where frontline workers got hazard pay for a few months while asset owners enjoyed a decade of outsized returns.

Now they watch presidential pardons follow political donations, enforcement cases quietly drop, and political families run crypto ventures while writing the rules for the industry and make billions doing it. It happens in the open, and nobody is embarrassed. They also watch enough Democrats leave office for corporate boards and consulting firms to conclude this isn’t just one party’s problem.

That isn’t the free market they were taught about in civics. It’s crony capitalism, and it’s the only version most of them have ever known.

One Word, Two Meanings

Younger voters also don’t hear the word “socialism” the way older generations do—and older voters are the ones who decide elections.

Someone born in 1950 was 39 when the Berlin Wall came down; that word represented the Soviet bloc for their entire formative political life. Someone born in 1997 arrived six years after the Soviet Union dissolved. Working from Gallup data, Brookings found the split: young adults define socialism as social democracy rather than public ownership by 58 to 38 percent, while Americans over 65 define it as total government control of the economy.

And that older group turns out. Census data show voters 65 and older made up 30.4 percent of the electorate while comprising 23.8 percent of those eligible. Voters 18 to 29 were 19.6 percent of eligible voters, but only 11.7 percent of actual voters. The word does opposite work depending on who hears it, and the people it frightens most are the ones who reliably show up at the polls.

The Historical Pattern

If you look at the last hundred years of American politics, political insurgencies follow a predictable lifecycle. They aren’t born out of nowhere—they are triggered by deep economic pain or institutional failure, and they force the mainstream parties to make a choice.

1. Absorption (1912 & 1934): When an insurgency is driven by economic concentration—like the Gilded Age or the Great Depression—the mainstream party eventually absorbs its practical demands while shedding its extreme edges. Eugene Debs’ Socialist Party and Huey Long’s movement died out, but working people got the income tax, the direct election of senators, Social Security, and union rights.

2. Rejection and Victory (1948): When Henry Wallace ran on an overly idealistic, anti-Cold War platform, Harry Truman drew a hard, public line against the radical fringe of his own party—and won the election anyway.

3. Cultural Drift and Backlash (1968): When an insurgency loses its focus on material economics and shifts toward cultural posturing—like SDS and the New Left—it alienates working voters, hands the presidency to Nixon, and eventually leads to McGovern losing 49 states.

4. Primary Purity vs. General Reality (2009): The Tea Party harnessed genuine anger over bailouts, but failed to keep its fringe in check. Radical primary winners handed five winnable Senate seats straight to Democrats.

That brings us to 2026.

The current surge around Bernie Sanders, Zohran Mamdani, and the DSA is triggered by the exact same things that fueled 1912, 1934, and 2009: debt, housing costs, and post-2008 economic cynicism.

The chart lists the outcome as “Unresolved” because Democrats haven’t decided which historical script they want to follow. If mainstream Democrats absorb the practical economic demands—housing, healthcare, and anti-monopoly rules—the movement transforms the country for the better. If they let the DSA’s constitutional rewrites and prison abolition define the brand, they walk straight into the 1968 and 2010 trap.

The Albatross and the Contortions

In November 2015, losing ground to Hillary Clinton and tired of the word hanging around his neck, Sanders went to Georgetown to define his ideology. What he gave the audience was pure Roosevelt: he walked through the New Deal, noted that every social safety net had been denounced as “socialism” at the time, and invoked the 1944 Second Bill of Rights. Then he drew a hard boundary: “I don’t believe government should take over the grocery store down the street, or own the means of production.” (Pretty funny given Mandani’s apparent obsession with municipal grocery stores).

Sanders explicitly defined his politics by ruling out public ownership. The DSA’s platform now explicitly demands it.

In doing so, Sanders built a brand he couldn’t control. The DSA was founded in 1982 and spent three decades as a debating society of 6,000 people with a median age of 68. When they backed Sanders in 2014, it was because he was going to say the word on national television. It worked beyond their wildest dreams: membership blew past 120,000, and the median age dropped to 33. The organization dropped its strategy of nudging Democrats from within and began primarying them directly(evidence, what recently happened in NYC).

Sanders never joined the DSA, but he made a dormant word valuable, and the group that owned the name took control of the definition. Now, every Democrat standing near that label answers for a platform they didn’t write, contorting themselves to appease a fringe without alienating what they perceive are their natural voters.

Watch what that looks like in practice. Asked about the DSA program, Congressman Ro Khanna from California claimed he hadn’t read it. Pete Aguilar said he couldn’t comment on documents published by outside groups. Ilhan Omar declined to engage.

Then came Alexandria Ocasio-Cortez on ABC’s This Week. Asked by Jonathan Karl whether she was still a member of the DSA, she offered: “Of the New York City chapter... Not the national.” Pressed on the actual planks, she disavowed them and waved off her past positions on policing with a quote: “Woke 1 was crazy.” Parsing chapter affiliations on national television in an election year isn’t a strategy—it’s trying to have it both ways, and voters see right through it.

Meanwhile, Republicans are happily paying to amplify the confusion and use this as a club. Speaker Mike Johnson declared the barbarians were inside the gate, the RNC pushed the DSA platform nationwide, and Republican super PACs spent millions trying to boost radical candidates in Democratic primaries. They know a radical primary winner is a gift that keeps on giving in November when they have virtually nothing else to run on.

What I’d Tell a Candidate

If I were advising a Democratic candidate today, the strategy would be straightforward:

1. Say plainly what you oppose before you’re asked. Denounce the DSA agenda: abolishing the Senate, nationalizing industries, ending prisons, defunding the police, and open borders early in your own words. Don’t wait for a talk show host to walk you into a corner.

2. Claim the economic high ground. The diagnosis on concentrated corporate power, housing affordability, healthcare rights, and worker bargaining power isn’t Marxist—it’s Rooseveltian.

Every successful reform movement in American history has asked for two kinds of things: benefits and ownership. Benefits get enacted—the income tax, Social Security, the minimum wage, union bargaining—usually by mainstream politicians responding to public pressure. Ownership demands never get enacted. Not once, in a hundred years of trying. Ironically, Trump is by far the most socialist President in our recent history demanding shares of countless companies in return for special treatment from the government or himself or his family. This is some sort of combination of socialism and crony capitalism only Trump could create!

There is plenty of room in the Democratic Party for progressives who want to rebuild the middle class and challenge corporate monopolies. There is no room for drifting into economic absurdities that hand winnable seats to the opposition. Roosevelt gave us the playbook 82 years ago: deliver real material security to working people, defend free enterprise from monopoly, and win the majorities required to govern. This ain’t hard. And Chris Matthews said it best:

Share

Before the numbers: what this is

The Forge Score™(the model I developed) is not a forecast.

Each candidate is scored separately from 0 to 100 on the strength of their own position — the ground they’re fighting on(structure), and how they’re performing on it(polls, markets, money). What I publish is the gap between them. It measures relative strength, not likelihood.

The Leverage Index™ answers a different question: not who’s ahead, but whether the race is still close, still has room to absorb money, and still has time to use it. Three things go in — how close it is, how much unspent advertising room remains, and how many days are left. Multiply them. A dead-even race with room to spend and months to run scores high. A blowout with saturated airwaves scores low.

That distinction is the whole point of this thing. We built it to answer where the next dollar changes the outcome most, not who wins. Leverage is not the same as closeness — a race can be tight and already drowning in money.

If you want win probabilities these are a few good sites to check out: FiftyPlusOne, RaceToTheWH, CPR, Sabato.

The 30-second version

• Democrats need a net gain of four seats. The Senate is 53–47, and a 50–50 tie goes to Vice President Vance.

• Six of these eight races are Republican-held. Two — Georgia and Michigan — are Democratic seats that have to be held. Losing either means the flips don’t add up. That means, if you want to win the Senate, Democrats are going to have to unite in Michigan behind El Sayed as the nominee. I’m definitely on that page. Despite those who abhor his views on Israel, he’ll be 1 of 100 Senators. Israel has far more problems with the American public and its representatives than the views of one potential Senator. Given the stakes in this election, this is an easy decision for me.

• Three races sit inside four points: Texas, Iowa, Ohio, and Alaska. Each of these tend to favor the R team, but mostly based upon historical voting patterns. However, the deltas are quite narrow suggesting each race is winnable especially in a “change” election.

• And here’s the irony. The two races we’re recommending to “not” contribute—Georgia and North Carolina— hold a combined $63.3M of Democratic cash between them. The six we’d fund hold $48.8M — combined.

  The money is sitting where the races aren’t.

The national picture, and why it doesn’t settle anything

The environment is about as good as an out-party gets:

• Trump’s approval sits at 34/64 in Pew’s July survey and 39/57 in Emerson’s. CNN notes that’s below the level that preceded midterm waves against Clinton in 1994, Bush in 2006, Obama in 2010 — and Trump himself in 2018.

• Generic ballot runs D+7.2 at RealClearPolitics; Emerson had it at D+11 in late July.

• 2026 special elections show an average D+10.4 swing off the 2024 baseline. Democrats have flipped state legislative seats this cycle. Republicans have flipped none.

• For the first time since 2010, voters trust Democrats over Republicans on the economy.

And yet. CNN’s Ron Brownstein calls it the approval gap: disapproval of Trump is deep and wide. The way I do the Forge score is intentionally conservative so this massive approval gap is not automatically converting it into leads in the races that decide the chamber. In 2018, 90% of voters who disapproved of Trump backed Democrats for the House. Today, with his numbers twice as bad, that figure runs between 69% and 85% depending on the survey. Our Senate scoreboard — a D+7 national environment, and Republicans still ahead in four of the seven races we can score.

TEXAS · Talarico −1.0 · Leverage 71

The most leveraged race in the country, and the only one where neither party’s super PAC has committed a dollar.

• Three straight public polls put Talarico ahead — Fox +3, Texas Southern +2, TPOR +5. All inside the margin. However, prediction markets still give a slight edge to Paxton.

• Paxton is winning white voters by 14 points. Trump won them by 33. Talarico takes Hispanic voters by 19; Trump carried them by 10.

• Talarico raised $28M in Q2. Paxton raised $1.6M. Paxton holds $1.8M — the smallest war chest on the board — and has absorbed $26.9M in attacks.

• Cook moved this Likely R → Lean R after Paxton unseated John Cornyn in May, the first Texas senator to lose renomination since 1970.

• The Republican plan is stated openly. A Paxton adviser: “Talarico’s campaign will end up exactly where every Democrat before him has ended up in Texas — a pipe dream.” Texas-sized ad buys, drive up the negatives, wait for the base to come home.

My hedge: structure is R+2.1 and no Democrat has won a Texas Senate seat since 1988. An August lead is not a November result.

What to do: give now to Talerico. The absence of outside money is the opportunity and it is temporary. His financial advantage vanishes the day national Republicans decide Paxton is salvageable.

IOWA · Turek −2.3 · Leverage 70

An open seat, a 19-point collapse in Trump’s numbers, and a candidate who once won by six votes.

• Ernst retired. Cook has it Lean R. No Iowa Democrat has won a Senate seat since Tom Harkin in 2008.

• Trump carried Iowa by 13 in 2024. Iowans now rate him 42/55 — a 19-point swing in twenty months.

• Fox’s July poll had Turek +4. Among voters calling themselves extremely motivated, Turek leads by 15 — Democratic enthusiasm 67%, Republican 53%.

• Turek is a two-time Paralympic gold medallist holding the most conservative state house seat any Iowa Democrat represents. He won it by six votes.

• Hinson has $6.7M to his $2.4M and a 10% small-dollar share against his 43%. Trump-endorsed; has promised to be the president’s top ally in the Senate.

• Both campaigns hold $9.1M combined — less than Ossoff has in Georgia by a factor of four.

My assessment: structure is R+4.1, and enthusiasm gaps in August, which are very favorable now, have a way of closing by November.

What to do: give to Turek. Cheapest competitive seat on the board.

OHIO · Brown −3.7 · Leverage 68

• Brown is the only Democrat with sustained statewide success in Ohio in thirty years — three Senate terms, plus secretary of state.

• He lost to Bernie Moreno by four in 2024, with Trump on the ballot. Trump isn’t on it this time.

• Husted was appointed, never elected to this seat. Win, and he runs again in 2028 for a full term.

• Brown holds more cash — $16.2M to $9.4M — and opened with an ad about donations to Husted from associates of Jeffrey Epstein.

• Outside groups have reserved north of $119M targeted against Brown. This is the most heavily funded race on the board and it isn’t close.

Where I detract from the prediction markets: bettors price Brown at 52%. We have him down 3.7, because Ohio’s structural lean is R+4.8 and the newest public poll is five weeks old. Theirs is a live price; ours weights ground the market may be discounting. We’d rather show the disagreement than hide it.

What to do: give to Brown. At $117.5M outside money against, marginal dollars buys a chance to stay in the game.

MICHIGAN · El-Sayed +4.2 · Leverage 60

A brand-new nominee, a split base, and 89 days to fix it.

• El-Sayed beat Haley Stevens on Tuesday — backed by Sanders and Ocasio-Cortez, running on Medicare for All and ending military aid to Israel.

• Stevens had Schumer, tens of millions in outside spending, and AIPAC’s largest single-race investment ever. She lost anyway. But, not by much.

• El-Sayed on primary night: “we have a responsibility to ensure that we unite.” That is now the entire job. He has significant work to do in the both the Black and Jewish communities, where he dramatically underperformed.

• Rogers has drawn $39.1M against and $29.0M for — the second-heaviest air war in the country, already running.

• Both campaigns hold $7.0M combined. Rogers is almost completely relying on outside money. And, Sayed spent heavily to win the primary. He clearly needs more capital to unite the party.

My reservation: in 2018 El-Sayed lost the governor’s primary to Gretchen Whitmer by 21 points. He has never won a statewide general, and Stevens’ voters have to show up for him.

What to do: give to El Sayed, and give this month. This is a seat Democrats must hold, and it’s a nominee with $2.6M will now face massive incoming fire. They’ll call him a “socialist” a “communist” an “antisemite”—none of which are true. AIPAC will also likely reload their guns against him again. Despite the current positive Forge Score, this race will be highly volatile, market predictions already contracting, and a race to watch closely.

MAINE · Not scored · Leverage withheld

We’re withholding Maine’s composite, and here’s why.

• Troy Jackson became the nominee on 25 July, at a convention arranged in weeks after Graham Platner withdrew.

• No poll has been fielded since. Every Collins-versus-Jackson number circulating was asked before he was the nominee.

• He has filed no FEC report. His campaign began 8 July; the next quarterly is due in October.

A score built on that would be inventing precision we don’t have. What we can tell you:

• The market prices Jackson at 66% — the most bullish Democratic figure on the board outside the two safe seats.

• Collins holds $11.0M with a 9% small-dollar share — the most institutionally funded campaign here.

• Jackson’s campaign says it raised roughly $3M from 58,000+ donors since the nomination. Their figure, not a filing, and we flag it as such.

• AdImpact projects Maine will draw close to $400M in advertising — fourth-most in the country. Just imagine living inside that media bubble.

• Race raters call Maine one of the Democrats’ two best pickup opportunities.

What to do: give to Jackson first. A five-term incumbent, an opponent who has been the nominee for twelve days, and a donor base being assembled in real time. Least measurable, most likely to matter.

ALASKA · Peltola −5.0 · Leverage 57

Our number says Sullivan. We think our number is the weakest on the board.

• Peltola raised $6.9M last quarter to Sullivan’s $2M. Six million of hers came from individuals; under $1.2M of his did — the rest from PACs.

• Small-dollar share: Peltola 37%, Sullivan 6%. Widest grassroots gap here.

• Polling has her up 3.5. The market prices her at 55%. Cook shifted the race toward her on 1 July.

• She ran 11 points ahead of Harris in 2024 and still lost her House seat by 2.4.

• Murkowski endorsed Peltola in 2022 and 2024. This time she endorsed Sullivan.

• Sixteen candidates are on the ballot, including a second Dan J. Sullivan — a retired teacher from Petersburg who won a court fight to appear, and has raised nothing.

The problem with the score. The only vector favoring Sullivan is structure, at R+8.8 — and structure is 41% of the number. That figure derives from Cook PVI, which is built from presidential results. Alaska splits tickets more than almost any state: Murkowski won a seat as a write-in; Peltola won a House seat outright in a state Trump carried by ten.

And the model does not represent ranked-choice voting at all. It scores Alaska as a two-way race. Peltola’s 2022 win was built on second preferences — precisely the mechanism we cannot see. It will also be Sullivan’s first time on the ballot under those rules.

Every input except the weakest one points at Peltola as a stronger candidate.

What to do: give to Perolta. The 18 August primary is top-four and both will advance, so it settles nothing worth waiting for.

GEORGIA +9.5 (Leverage 44) · Democratic hold — and NORTH CAROLINA +7.2 (Leverage 54)

The two races we would not fund — and the two holding most of the money.

• Ossoff has $42.6M against Mike Collins’ $2.1M, with a 51% small-dollar share. Market: 91%.

• Cooper has $20.7M against Whatley’s $3.5M. Market: 91%.

• Both lead every vector we track. They are the two lowest Leverage scores on the board.

• Between them they hold $63.3M — more than the six competitive races combined.

What to do: nothing further. A dollar here buys both of them capital for a potential Presidential run in 2028. Not needed to win this year.

If you can’t give money

Postcards work, but less than the people writing them assume. The largest test — 15.7 million cards to 7.8 million voters in 2020 — moved turnout 0.07 points, and all of it came from the short social-pressure script. The alternative script measured zero. One card did as much as three, so write the script you’re given, once per voter. Here’s the best place to go if you want to write postcards.

Phone calls are the better use of an evening. A completed conversation with a volunteer raises that voter’s turnout by about 3 points. Every campaign here runs virtual banks you can join from anywhere. Just visit their websites or go to the Mobilize website and search by candidate name.

That’s it for today.

The same day I published this piece on how far out of whack the political money game has become — and, more to the point, how much corruption it has brought down on a fragile democracy — two stories landed on exactly that subject. One from the Wall Street Journal, one from the New York Times.

Both make the argument better than I did.

Musk is back, with $120 million and a door-knocking operation

He’s authorized his super PAC, called cynically the America PAC, to spend $100 to $120 million on a field program in at least eight states. The initial Senate targets are Alaska, Iowa, Maine, Michigan and Ohio, with conversations underway about North Carolina, Georgia and Texas.

Those are the eight races on my Forge Senate board. All of them.

What’s changed since 2024. Musk supposedly stays offstage(we’ll see how long this lasts). After his candidate lost the Wisconsin court race last April he appears to have concluded his own celebrity backfired. And America PAC is now pooling its ground game with Americans for Prosperity and the Sentinel Action Fund — the Times says Republican outside groups are consolidating field operations because the June 30 ruling made it cheaper for party committees to work directly with candidates.

I argued yesterday that the ruling handed Republicans a force multiplier at the moment Democrats couldn’t use it. The Times reported it happening the same day.

What hasn’t changed. Chris Young, who ran the 2024 operation, is running this one — and America PAC is again hiring firms to knock on doors. In 2024 it put more than $60 million into Trump’s ground game through paid contractors, and roughly a quarter of the door-knocks logged in Arizona and Nevada were flagged as suspect, with entries filed from Wi-Fi networks or far from the addresses claimed.

Pooling three organizations fixes coordination between outside groups. It does nothing about what actually broke the last operation, which was paying strangers by the contact and handing them a quota. We’ll see if they do better this time around.

Musk isn’t buying television with that $120 million. He’s buying doors. His side has read the same research I did and reached the same conclusion — advertising barely moves anyone, and turnout is what’s left.

There is a rung below my ladder

Trump has raised more than $800 million since returning to office — across his super PAC(Maga Inc), a presidential library, a White House ballroom, a helipad, a statue garden and a birthday party for the country. MAGA Inc.’s idle $400 million is one pocket inside that figure, not separate from it.

According to the WSJ article, he calls his fundraiser most nights to ask which companies have paid and which haven’t, and routinely tells her to ask for more than she’d planned. She has relayed to executives that the boss wants the money. SoftBank gave $50 million to the library. Meta gave $22 million after settling a lawsuit Trump had filed against it, then $10 million more to an aligned committee that never reports its donors. Chevron was asked for $50 million.

In May, tobacco executives met Trump at his golf club. He promised them much of what they wanted on vaping regulation and took in millions for his political committees. Shortly afterward the FDA lifted restrictions on some flavored vaping products, and the FDA’s chief was gone.

In June, the Pentagon awarded Lockheed Martin a contract worth up to $35 billion. Weeks later Trump announced Lockheed had agreed to pay about $5 million toward his new White House helipad.

Yesterday I laid out a ladder — your $3,500 check at the top, dark money at the bottom, ranked by how accountable each dollar is. That ladder was incomplete. There’s a rung underneath: money handed straight to a sitting president’s personal projects. No limit. No disclosure unless the press does its job. No election law of any kind, because technically none of it is an election.

Can he shuffle it between pockets? Not freely — a library foundation can’t legally fund politics, and a nonprofit used as a pass-through to a super PAC is prosecutable. But that’s the wrong question. The money doesn’t need to move, because the donor does. Same company, same business before the same government, same man asking. He just decides which pocket. Want it on the public record? The super PAC. Rather it weren’t? The ballroom.

FINAL THOUGHT

Citizens United held that independent expenditures cannot corrupt because they are independent. Every protection in that framework rests on distance between the spender and the candidate.

There is no distance here. There’s no attempt at distance. It’s done in plain view, with open contempt for the idea that any of it needs hiding.

Justice Kennedy wrote that the appearance of influence or access would not cause the electorate to lose faith in this democracy. A fundraiser sits at million-dollar dinners taking notes on what the president promises each donor, then passes them to White House staff.

Some of you have written to tell me I’ve turned into a leftist. I haven’t. Far from it. I’ve turned into a cold-eyed realist about where this country actually is.

Others ask how I could support a progressive candidate for the Senate or the House if that’s who gets nominated. Here’s my answer: is this the kind of country you want? Grift and corruption at a scale none of us has seen in our lifetimes. In an autocracy that isn’t a bug — it’s a feature. I’ll take an FDR fashioned candidate with progressive views every day over this craven bunch!

If that’s what you want, keep voting for the R team. But this is not your grandfather’s Grand Old Party, and the people enabling it are sitting in Congress right now with the votes to stop it.

Next week: the Senate Forge Scoreboard — where your dollars and your time can still make a difference.

Share

I was with my insurance agent last month when the conversation drifted to politics. He asked what I thought was the single biggest problem facing the country.

The easy answer would have been to simply pile on to the current occupant at 1600 Pennsylvania Ave. But, that’s not what I told him. I said the biggest danger to the way we live is the role money now plays in our politics, and that Citizens United was quite possibly the worst Supreme Court decision of my lifetime — potentially worse than even overturning Roe. And, that was before June 30, when the Court handed down one that makes things even worse.

How we got here

The irony is that this started as an attempt to apply the brakes.

2002 — McCain-Feingold. Congress banned soft money, the unlimited checks written directly to national parties, and barred corporations and unions from running campaign ads right before an election.

2003–2009 — the water finds a crack. Cut off from the parties, the money moved into 527 groups, answerable to the IRS rather than the Federal Election Commission(FEC). Fifty-two people supplied 44 percent of all 527 money in 2004. Three Texas oilmen seeded Swift Boat Veterans for Truth and turned "swiftboated" into a verb—and, did serious damage to John Kerry’s Presidential campaign. The FEC later ruled much of it illegal — America Coming Together raised $137 million and only $34 million was lawful — and fined the group $775,000, under one percent of what it spent. That ruling came in 2007.

2010 — Citizens United. A broken outside money system gets a lifeline. A legal fight over an anti-Hillary Clinton documentary triggered a rewrite of American election law. The Court ruled that political spending is speech and that government cannot cap election money, whether it comes from a $35 grassroots donor or a multi-billion-dollar corporation. It never had to declare that corporations are people. It simply held that unlimited corporate cash spent independently posed no threat:

“This court now concludes that independent expenditures, including those made by corporations, do not give rise to corruption or the appearance of corruption. That speakers may have influence over or access to elected officials does not mean that those officials are corrupt. And the appearance of influence or access will not cause the electorate to lose faith in this democracy.

Sixteen years on, this quote from the decision is simply mind blowing in its naivety and wishful thinking.

2010 — the Super PAC is born. Months later a lower court completed the logic. If outside spending couldn’t be capped, then donations to outside groups couldn’t be capped either. Unlimited money, from anyone, exploded overnight.

2026 — NRSC v. FEC. On June 30, by six votes to three, the Court overruled a 2001 precedent and removed the limit on how much a political party can spend coordinating directly with its own candidates.

It left two things standing: what you or a PAC can give a party, and the rule barring Super PACs from coordinating with campaigns. What it moved was power back to the party committees — the exact circumvention the 2001 case was decided to prevent.

The ladder

Two rules govern everything below: how much can go in, and whether it can be coordinated with the candidate.

Your direct donation. Capped at $3,500 per election. Fully coordinated — it’s the candidate’s own money — and by law it buys the cheapest advertising available.

Party committees. The RNC and DNC, and their Senate and House arms—the NRSC and DNCC. Capped — but read that word carefully. Each party runs three separate committees, each with its own limit, plus side accounts for buildings, legal fees and conventions that take three times the base. Add it up and one donor can give a single party about $2 million a cycle. That's 300 times what the same donor can give a candidate. Donors disclosed. And as of June 30, all of it can be spent in direct coordination with the campaign.

Super PACs. The Citizens United invention. Unlimited money in — but no coordination with a campaign, ever. On the Senate map the big two are the Senate Leadership Fund (Republican) and Senate Majority PAC (Democratic). Donors disclosed.

Dark money. Social-welfare nonprofits — the 501(c)(4)s — that spend on elections without revealing who funded them. They answer to the IRS, not the FEC, and surface only when they run ads, at which point you learn the amount and never the source. They often route money through a Super PAC, which then discloses a check from the nonprofit and nothing else. Both parties run it as a matched pair: One Nation feeds the Senate Leadership Fund, Majority Forward feeds Senate Majority PAC.

A candidate’s own Super PAC. A personal vehicle for parking enormous sums. Trump’s MAGA Inc. is the largest ever built. Democrats have no equivalent.

Notice the trade running down that list. Money that can be coordinated with a campaign is limited coming in. Money that’s unlimited can only shout from outside. That tension was the whole design, and on June 30 the Court removed half of it.

Two money races

There are two contests running in 2026, and each side is winning one.

Where the money comes from voters, Democrats are ahead. Where it comes from institutions and billionaires, Republicans are ahead, and it isn’t close. The Senate Leadership Fund — the Republican Senate super PAC — closed June with $238 million in cash and has committed $342 million across eight races, including $42 million in Maine and $79 million in Ohio.

Democrats are not absent. Senate Majority PAC, its Democratic counterpart, added $54 million in late July across Ohio, Iowa, North Carolina and Alaska; Ohio alone now has more than $140 million of outside advertising booked. This is not a game only one side plays. But the big-check advantage belongs to Republicans, and the Democratic advantage is the other kind of money — the kind that arrives in small amounts, straight to the candidate.

The timing

Here is the part that should worry Democrats more than the cash gap itself.

The case the Supreme Court decided on June 30 was NRSC v. FEC. The National Republican Senatorial Committee brought it. Ten days before the ruling, the NRSC announced it was entering the summer with the highest cash on hand in its history, and added that “should the Court rule in our favor, Republicans will be ready to immediately capitalize on this newfound political firepower.”

The Court ruled in its favor. Republicans were ready.

What the decision unlocked is unlimited party spending coordinated directly with a candidate — shared targeting, messaging, timing. It is the only outside money that can behave like campaign money. And it is the one kind of money Republicans have and Democrats don’t.

The RNC holds $128.5 million and no debt. The intensely crippled and dysfunctional DNC(where I NEVER contribute), holds $16.3 million against $18.5 million of debt, took a loan last October against its own headquarters, and has stopped its customary transfers. The NRSC has $55.9 million to the DSCC’s $41 million.

So the ruling didn’t just loosen a rule. It multiplied the single advantage Republicans hold, in the same month the party that lacks it started asking vendors to wait for payment. In Texas, where Ken Paxton is nearly broke against a Democrat with twelve times his cash, reporting says he is already leaning on coordinated party money to close the gap.

The $400 Million Slush Fund

MAGA Inc.—Trumps personal Super Pac, began the year with roughly $300 million, raised nearly $100 million more, and closed June above $400 million. It has not directly supported a federal race since March. Republican donors are openly frustrated that the largest pot of political money in the country is idling while their candidates go underfunded. I’m putting high odds that Trump won’t release much of this money to support candidates this Fall. This would kill him. His motto, “what’s mine is mine and what’s yours is mine too!”

Where it came from is an interesting short list.

• Cameron and Tyler Winklevoss — $10 million in Bitcoin on June 19. Twenty-three days earlier, the federal agency that had won a judgment against their crypto exchange, Gemini, asked a court to undo it. I wonder why.

• Greg Brockman, OpenAI’s president, and his wife — $12.5 million each.

• Foris Dax, Crypto.com’s parent company — four checks totaling $35 million, the largest single source of any kind.

• Kelcy Warren and his pipeline company, Energy Transfer — $12.5 million each.

Now hold those against the ceiling on a citizen: $3,500 to a candidate, per election. A single $25 million contribution to a Super PAC equals what more than three thousand maxed-out voters could give, combined.

That mismatch is why so many Republican Senate candidates, unable to raise small-dollar money, now lean on the party, the Super PACs and a short list of the very rich. In Texas, four-fifths of Ken Paxton’s money comes from large donors. His opponent raises more than half of his from people giving under $200.

What money can’t buy

Campaigns have decided, at least they think, attention is for sale. In California, Tom Steyer’s campaign paid $25,000 for one post to an account with 28 million followers, and $100,000 to a consultant with 14 million TikTok followers who then endorsed him.

Based upon my years working at YouTube, I seriously doubt this type of paid endorsement works. A creator or an “influencer” is only worth paying if their followers believe they are telling them what they actually think. The moment they find out they were paid to say it, it’s just another ad — and a worse one, because now they feel played.

You cannot buy a following. But if a candidate already has one, your money is what turns it into something that wins. Before he dropped out, Graham Platner’s million dollars raised in nine days arrived with 2,700 volunteers attached — the attention came first, and the money converted an audience into people knocking on doors. Zohran Mamdani had the reverse problem: so much earned attention that he hit New York’s spending cap in March and asked his supporters to stop.

So the test is simple. Give to a campaign with a real following and your dollar becomes a field organizer, a canvass, a storefront in a county the party forgot — at the discount the law reserves for candidates, buying the one thing that moves votes.

The candidate brings the authenticity. The funds simply allow it to be organized better and amplified.

Final Thought

Every step along this road moved political power into fewer, richer and less accountable hands. The Court is clearly putting its finger on the scales and seems determined to continue to make matters worse.

What’s left is the one advantage the law still gives an ordinary donor: a check written to a campaign buys cheaper advertising than the same check written anywhere else, and it pays for the one thing that reliably moves an election, which is a person at a door. That advantage might be small, but it is also real and can make a difference in very close races. It is also why I built the Forge DSS(Decision Support System).

The full eight-race Forge scoreboard — every Senate seat that decides control, ranked by where a dollar actually might help — comes next week, once Michigan votes on the 4th for their Senate candidate.

First, some relief.

Three weeks ago Maine Democrats had a nominee who had raised a million dollars in nine days on thirty-three-dollar donations, driven a sitting two-term governor out of the race, and then lost his campaign to allegations he could no longer run away from. The risk was that the party would spend the summer fighting over a replacement, or that a committee would simply install one.

Instead they held a convention this past weekend. Four candidates qualified, all withdrew and endorsed rather than force a fight, and Troy Jackson took better than 99 percent of 601 delegates. A legitimate process producing a competitive nominee inside two weeks deserves more credit than it’s going to get.

Who he is

Jackson is 58, a fifth-generation logger from Allagash — a town of about two hundred on the Canadian border. “I went into the woods right after high school,” he told delegates Saturday. “All of us. I was no different.”

His politics started on a road. In 1998 he joined a blockade of the northern border, loggers parking pickup trucks across it to stop cheap Canadian labor coming in and American wood going out. He won a Maine House seat in 2002 as an independent, joined the Democrats in 2004, and spent twenty years in the legislature, six of them as State Senate President. He ran Bernie Sanders’ Maine campaign in 2016. He passed a law giving loggers and wood haulers the same collective bargaining rights the state already gave potato farmers and lobster fishermen.

He is also hot-tempered, by the account of people who served with him, and there are already stories circulating about angry outbursts. Chellie Pingree, Maine’s Democratic congresswoman, didn’t deny it when asked — she said people are frustrated and looking for candidates who show some passion. That will be in an advertisement either way.

And his record is spottier than his résumé suggests. He came in third in this year’s Democratic primary for governor. He lost a congressional primary in 2014. He wins in his corner of northern Maine, and he has never won anything statewide.

Collins is genuinely vulnerable

Morning Consult has her at 38 percent approval against 54 percent disapproval — the worst it has recorded in eight years of tracking her, after a seven-point drop this year alone. That briefly made her the second-least popular senator in the country.

The University of New Hampshire’s poll, released July 21, has Jackson 49 and Collins 46 with a margin of error near three. A statistical tie, challenger nominally ahead of a five-term incumbent. She has privately conceded this may be her hardest race.

What the betting markets did during the collapse

Two venues price this race and about $2.8 million is riding on it. Kalshi has Jackson at 63 percent against Collins at 38. Polymarket, running a party contract, has Democrats at 67 against Republicans at 34. Two independent books, four points apart.

Through the winter and spring, Democrats sat between 65 and 70 percent — this seat was expected to flip. Then Platner’s campaign came apart and the price came apart with it. Democrats fell into the forties, Republicans climbed toward even money, and for a stretch this looked like a race Collins would be handed.

It is back to 67 percent. The entire crisis has been priced out — not hedged, but returned to roughly where it sat before anyone had heard the allegations against Platner. That is the strongest evidence I can find that the nominee switch did not cost Maine Democrats this seat.

It is not a sure thing

Collins has beaten her polling before, badly. In 2020, RealClearPolitics’ final average had Sara Gideon ahead by six and a half. Collins won by 8.6 — a fifteen-point miss, in this state, against this senator, in a year Biden carried Maine by nine and Democrats spent roughly $70 million trying to unseat her. Jackson leads by three. The miss last time was five times the size of the lead this time.

And Jackson is not Platner. Platner came from nowhere — no voting record, no donor list, no title — and by February led Governor Janet Mills by twenty-seven points. She quit on April 30, five weeks before anyone voted. He had already survived something that should have finished him: when a tattoo resembling a Nazi symbol surfaced last October he went from leading her by thirty-four to trailing by five inside a week, then won the primary by fifty-six. Maine Democrats had decided he was authentically who he said he was and weighted that above the evidence in front of them. That is the rarest asset in politics and it does not transfer.

Here’s the wrinkle, and it runs the other way. Platner was prep-school educated and farmed oysters; his working-class credentials were arguable, and Democrats argued about them. Jackson’s are not arguable. Several Democrats told the NY Times this weekend, without quite saying it on the record, that they think they got the better avatar the second time.

What they lost was the insurgency. Platner was a stranger who beat the machine. Jackson ran the machine for six years, was nominated at a convention nobody contested, and speaks haltingly. This race will be a case study in whether the working-class-candidate strategy works at all. Third Way’s Matt Bennett gives the counter-argument its sharpest form: the mistake the left keeps making, he says, is assuming that finding someone who looks the part is enough.

Jackson’s answer to that is the line he closed on in his acceptance speech on Saturday, and the delegates said it with him. “We’re not from the left, we’re not from the right. We’re from the bottom, and we are rising.”

Which is exactly why the money matters now

Look again at the numbers at the top of this piece. A candidate nobody knows, in a statistical tie with a vulnerable five-term incumbent, in a state his party carried by seven in 2024, with nothing yet in the bank and $53 million already committed against him. His campaign doesn’t take corporate PAC money, which means the fastest place that gap closes from is people.

Money arriving in October pays premium advertising rates into a saturated market and does almost nothing. Money arriving now pays for organizers, field offices and the door-knocking operation that actually moves turnout — and it pays at the discounted rate the law reserves for candidates rather than outside groups.

Won’t the party fund him?

Partly. Schumer and Gillibrand endorsed within hours, and the DSCC has been banking money for whoever emerged in Maine. He won’t be at zero for long.

But the attacks are already running — the Senate Leadership Fund, the main Republican super PAC, put its first ad against him on air the day he was nominated, while Collins’ have been running for months. And party money buys less: that candidate discount doesn’t extend to the DSCC, even now that the Court has let committees spend unlimited amounts coordinating with them. A dollar routed through Washington buys a third to a half of what it buys inside his campaign, and it can’t pay a canvasser.

He isn’t courting them either. Jackson said Saturday he hasn’t spoken with Schumer and won’t back him as leader.

The party will show up. Later than the attacks, at a worse price, buying the thing that works least well.

Where I come down

Give here first.

Not because it’s the surest bet on the key Senate races— it isn’t, and I’ve just spent four hundred words telling you why. And not because it’s the highest-leverage race by the model’s own math; Iowa scores higher.

Give here because this is the widest gap on the board between what a candidate needs and what he has, and because that gap closes fastest with small money arriving now. Collins is at 38 percent approval in her own state. A challenger nobody has heard of is tied with her. Those are the conditions under which a donation actually changes something, and they are not the conditions in most of the races you’ll be asked to fund this year.

Donate Here

Next: the full Senate eight-race board, with recommendations for all of them.

Share

Michigan Democrats pick a Senate nominee on Tuesday. The American Israel Public Affairs Committee(AIPAC) and its affiliated groups have spent something like $40 million trying to decide it for them — the largest investment AIPAC has ever made in a single race.

Asked twice about the group at a campaign stop in West Michigan this month, Haley Stevens answered for more than 800 words without naming it once.

Before any of that, so you know where I’m standing. I’d vote for Stevens. She’s closer to me on the issues, and since Mallory McMorrow left the race she’s found a sharper voice and is fighting hard for it. I don’t like what her campaign is being funded by, and I don’t think she’s handled the AIPAC question well. I used to argue she was the safer bet against Mike Rogers in November — I’m retiring that argument below, because the evidence stopped supporting it. And if Abdul El-Sayed wins Tuesday, I’ll back him in November without hesitation. Holding this seat isn’t optional if Democrats want the Senate. Plus, he would have earned it. We aren’t quite yet Russia!

Everything past this point is what I measured, not what I’d prefer.

What the money is about

Michigan has the largest Arab, Muslim and Palestinian population in the country, more than 200,000 people. Wayne County counted 139,751 residents identifying as Middle Eastern or North African in the 2020 census — the highest share of any county in the United States. Dearborn is majority Arab American and has the largest Muslim population of any American city. Southeast Michigan also has a large, long-established Jewish community, concentrated in the Oakland County suburbs Stevens has represented for four terms. Earlier this year, Temple Israel in West Bloomfield was attacked.

The candidates are about as far apart on this as two Democrats get. Stevens has been among the House’s more reliable supporters of Israel and has backed continued military aid. El-Sayed voted “uncommitted” in 2024 Presidential primary, has called the war in Gaza a genocide, and has said the Israeli government is as “evil” as Hamas. However, to be clear, I see zero evidence the guy is authentically in any way antisemitic.

Neither community votes as a bloc and it would be lazy to write as though they did. Michigan’s Arab American population includes large Christian communities whose politics differ sharply from its Muslim ones, and Jewish voters in Oakland County are not of one mind about Netanyahu, or about AIPAC.

But the turnout arithmetic is real. In 2024, just over 100,000 Michigan Democrats — 13.2 percent of the presidential primary — voted “uncommitted” over Gaza. In Dearborn, Dearborn Heights and Hamtramck, uncommitted beat a sitting president; Dearborn went 56.2 percent. Nine months later Trump carried Dearborn. That’s a bloc of measurable size that has already broken from its party over this question, and an August primary is exactly the low-turnout condition where a motivated bloc counts for more than its share.

Where the money came from

United Democracy Project, AIPAC’s super PAC, is the largest spender. A second group, A Stronger Michigan, has put in more than $12 million and hasn’t disclosed its donors. Total outside spending has passed $41 million, and AdImpact counted $46.1 million in advertising as of July 1 — 74 percent of it for Stevens. Rogers has $10.7 million behind him; El-Sayed has $2.3 million.

Two weeks ago the United Auto Workers sent a cease-and-desist to A Stronger Michigan for using the union’s wheel logo in a pro-Stevens ad. The UAW endorsed El-Sayed in June, by the two-thirds majority its rules require. Separately, ads have run featuring Barack Obama praising Stevens’ work on the auto bailout. Obama hasn’t endorsed anyone in this race. A Detroit city councilman backing El-Sayed says voters have told him they came away thinking otherwise.

What the candidates raised themselves

El-Sayed out-raised her by two and a half million, almost entirely from individuals, with essentially no PAC money — and they go into Tuesday within $240,000 of each other. The $40 million from the outside groups like AIPAC could be a game changer or possibly not.

The gap nobody is measuring

I pulled attention and coverage data on both candidates.

News coverage is the control here. Both candidates get covered at nearly the same volume, and El-Sayed’s coverage is meaningfully more hostile — average tone of −0.79 against her −0.50. Same amount of press, rougher press.

Public attention doesn’t follow it. He runs about two to one on Wikipedia searches and gets nearly triple the views per YouTube video from an identical number of channels.

I can’t tell you which direction that runs. Attention isn’t approval, and a man absorbing thirty million dollars of attack advertising gets looked up by people who just watched an ad about him. What argues against that reading is that his favorability is higher than hers, not lower — Tavern Research has him at 45 net to her 40. Attention that was purely damage should be showing up as damage.Where I was wrong, and where the evidence splits

I’ve argued here that Stevens was the safer general-election candidate. That argument was built on polls taken before June, and the two most recent ones reverse it.

Stevens is still making the case. “No one wants Abdul to win more than the Republicans,” she said on Morning Joe this month. “They think they can beat him in November. What they’ve never done is beaten me.” That was a strong argument in November. It’s running on eight-month-old numbers now.

The primary polling is a different question and it favors her — she leads Black voters 67 to 21 and non-college voters 56 to 34, the constituencies that decide statewide Democratic primaries here, and Whitmer endorsed her Friday.

The market data can be tricky to unpack. Prediction-market traders skew young, male and extremely online — a fair description of El-Sayed’s base, and a market whose participants overlap one candidate’s coalition can be confidently wrong in that direction. And the research showing advertising barely moves votes is about general elections, where the party label does the work. Primaries have no party cue and less information, and persuasion effects there are considerably larger. Forty-six million dollars of one-sided advertising in a low-information contest is exactly where money still works. But not necessarily, look what happened in NYC with Mandani vs. Cuomo.

A jump ball

I can’t tell you who wins. That isn’t a hedge — it’s what the evidence looks like laid side by side. It’s really crazy as hell.

Pointing to Stevens. The primary polling, by four to seven. A 46-point lead among Black voters and 22 points among voters without a college degree — the two blocs that decide statewide Democratic primaries here. Whitmer(the sitting Governor), Peters(the retiring incumbent) and the party apparatus behind her. Forty-six million dollars of one-sided advertising in a low-information primary, which is the one environment where the research says advertising still moves votes.

Pointing to El-Sayed. Four million dollars of prediction-market money at two to one. Twice the public attention and nearly triple the video audience, in the face of harsher coverage. Better favorability than his opponent. The UAW. The two most recent general-election polls. And a 2024 precedent, in this state, for a bloc that likely-voter models did not see coming.

Both lists are real, and neither cancels the other.

Run this through the numbers cold and you land on Stevens, narrowly — the polling and the demographic margins are the most reliable instruments we have, and they both point one way. That’s what Claude used to predict for Stevens, when I asked, which it based largely on the black voter margin for Stevens from recent polling.

I’ll take El-Sayed anyway. Not because the arithmetic says so, but because I’ve watched Michigan long enough to distrust likely-voter screens when a candidate is drawing this kind of attention against this kind of money, and because the last two times I’ve seen this shape — Mamdani in June, Platner in Maine — the screens were the thing that was wrong.

And the conditions are strange. Michigan votes early and by mail, so this electorate isn’t being asked a question on Tuesday — it has been answering continuously since June 25, while the ground moved underneath it. There is no single moment at which this race can be said to have been decided.

Tuesday settles it. I’ll say afterward whether my gut or Claude had it right.

Coming this week: Maine has a nominee and no money. Then the full Senate eight-race board, with the Michigan row finally filled in.

Yesterday I ended Part II of this series on the strongest argument the do-nothing crowd has — four words: but China will win. Jim VandeHei, who runs Axios, published a piece this week in the WSJ, entitled “America Risks Blowing the AI Race….” It’s worth reading, and he’s right about most of it. He isn’t in the do-nothing crowd at all — he wants America to do a great deal. But the whole plan is built on the same China framing.

He’s especially right about the thing I’ve spent two posts on. What makes AI different from every technology that came before, he says, is that we know what’s coming before it arrives. The risks are known now, preparation is optional, and we’re choosing not to prepare.

So my argument is with his last step — the turn almost everyone in this debate takes. His organizing frame is the moonshot: a national project, a clear goal, a clear competitor to beat. I understand the appeal. It boils everything down to a simple—race. But it’s wrong headed and completely misleading. This is NOT a zero sum game.

There’s no finish line

Apollo was a race because it had a destination and a date, and one morning in 1969 it was over.

AI has no such morning. It’s a general-purpose technology that will work its way into medicine, weapons, logistics, education, finance, and war over the next several years and even decades. Nobody won the internet, and nobody is going to win this. China will be excellent at some applications, especially in advanced manufacturing, and we’ll be excellent at others, and that will still be true in 2045. Hopefully, if we get this right, we’ll lead more than follow. It’s parallel development, not a horse race. And, since the Chinese are largely pursuing open source models for the core LLM platforms, they’ll end up having a significant place no matter what we do commercially.

The framing matters because once you tell people there’s a finish line, every argument for caution becomes an argument for losing. Suggest that a companion app should stop when a teenager is in crisis, and someone tells you China won’t slow down. I heard exactly that when I went ballistic about what Character.AI was doing to young people. Suggest we think about biosecurity before the capability arrives, and someone tells you Beijing isn’t pausing. It works because it sounds like realism and nobody wants to be the guy who lost the future.

I’m not out on a limb here. Eric Schmidt, former CEO of Google, chaired the National Security Commission on Artificial Intelligence — the body whose 2021 report framed China as the rival to beat. This week, in the NY Times opinion piece, writing with his China policy researcher Selina Xu, he seemed to walk away from it. The real challenge, they wrote, isn’t whether the United States or China builds an insurmountable advantage over the other — it’s whether either country can absorb this technology without ripping apart its own social fabric.

Two different problems

There’s a competitive layer — products, chips, energy, models, talent, defense. On that layer, compete hard. VandeHei is right. Lean in, invest, lead, and build a serious coalition with the other democracies where it makes the most sense, especially in the area of defense, because that alignment is real leverage. And accept that China might have significant wins in some of these sectors. That’s what parallel development looks like. The goal is to be the bigger winner across more of what matters.

Then there’s a catastrophic layer — an engineered pathogen, a cyber weapon that takes down a grid, a system nobody can turn off. On that layer there is no such thing as an American win. If a pathogen gets out, we don’t win. If a system gets away from us, American leadership in AI will feel more like a curse—something we should have managed to avoid.

Those two layers require opposite behavior, and treating them as one problem is how we end up doing nothing about either.

The problem with the plan

VandeHei says biosecurity risk is real and rising, and that we ought to map it out before the capability arrives. He’s right.

Then he proposes a global framework whose purpose is to make the world depend on American leadership and pull developing countries away from Chinese alternatives.

I don’t see how you do both. If the safety system is built to shut China out, it isn’t really a safety system. And the risks he’s worried about don’t stop at borders. A pathogen doesn’t care which countries signed our agreement.

The monolith

Underneath the race framing sits an assumption almost nobody examines: that China is a monolith. One point four billion people rowing in the same direction, controllable the way Mao controlled them during the revolution, none of them worried about the machine reshaping their lives.

That is not the country that exists. China today is a complex society with a huge economy, a large entrepreneurial class, and millions of people who share just about every concern on the list I laid out yesterday. And we don’t have to guess, because Schmidt and Xu went and looked. The Chinese tech executives they met last month were more conservative about AI’s pace than Silicon Valley — one chief executive told them point-blank that growing more slowly beat going full speed, because he wanted no repeat of the backlash the Industrial Revolution earned. Beijing has been regulating in exactly the places on my list. In April it froze new robotaxi licenses after driverless cars stranded passengers on the streets of Wuhan. The same month, a court ruled that companies can’t fire workers just to replace them with AI. A ban on AI companions for minors took effect this month.

And notice what the difference produces. Their public is excited about this technology — 84 percent in a recent Stanford survey — and Schmidt’s read is that people can watch their government chiseling at the problems. Conversely, our public is souring fast — communities fighting data centers, parents frightened of companion apps, workers bracing to be replaced — while Washington mostly watches. The do-nothing crowd says guardrails will kill the golden goose. They have it exactly backwards, and we’ve already run the experiment. Social media spent two decades fighting even modest guardrails for kids, and the bill arrived as furious parents, bellwether verdicts, and national bans. There’s a line I’ve used for years that covers it: the only thing worse than malfeasance is nonfeasance.

None of this makes me naive about the threat. The Chinese Communist Party(CCP) is a bad actor on a long list of fronts, and China cheats like hell for advantage — on intellectual property, on trade, on talent. Both things are true at once: a rival that cheats, and a society full of people who worry about what we worry about. The race framing needs the first and erases the second, and the second is where the work is.

Why China has to be in the room

I know how this sounds in 2026, so let me be precise. I’m not proposing we hold hands with Beijing. On the catastrophic layer, their interests and ours actually line up. A government organized around control does not want an uncontrollable system. A regime that fears instability above everything does not want an engineered pandemic. Assuming they’re indifferent to catastrophe is simply illogical—especially for a regime that thrives on cold hearted realism.

Kevin Rudd, former Prime Minister of Australia, who wrote his Oxford doctorate on Xi Jinping's ideology and warns Americans never, ever to underestimate the man — laid out the same architecture on the Ezra Klein show(listen here): deter hard on the red lines, compete flat-out on the economy, and cooperate on the short list where survival overlaps. First on his cooperation list: AI governance. His phrase for doing both at once — competing and cooperating in the same breath — is that "we can walk and chew gum at the same time."

We’ve done this before, with a country we liked a lot less. The U.S. and the Soviet Union agreed on nothing. We fought proxy wars on three continents and had thousands of warheads pointed at each other’s cities. And we still built test bans, hotlines, and inspection regimes — not because anyone trusted anyone, but because on the narrow question of not ending the world our interests matched, and serious people did the slow, boring work of finding where they matched. Nobody signed a test-ban treaty because they liked the Russians.

I’ll admit where the comparison strains. A nuclear weapon needs fissile material — rare, expensive, countable, visible from orbit. Software is diffuse, dual-use, and copies for free. Verification here is a much harder problem. I get that, but the precedent proves that enemies can cooperate on survival while competing on everything else.

What I learned about cornering people

Fifty years of negotiating taught me one rule that applies directly here. Back a powerful counterparty into a corner with no way out that lets them save face, and you don’t get compliance. You get destruction. Just look at what we are facing now with Iran.

That’s what a containment-first framework does. It tells the second-largest AI power that the global safety system is aimed at them, then acts surprised when they refuse to be bound by it and race harder out of sight, cutting corners we never see.

What adult supervision actually means

Not a treaty next spring, and not trust. Patient, unglamorous engagement by serious people who keep showing up — technical people talking to technical people, diplomats who understand the machines, quiet channels that stay open when the politics get ugly, because they will. Small agreements that build the habit of bigger ones. That’s the work that got us through the Cold War, and it took thirty years and won nobody an election.

For most of this debate I’ve seen sloganeers on one side and doomsayers on the other, and the actual work not getting done by anybody. Then, within a few days of Schmidt’s piece running, two hundred economists — sixteen Nobel laureates among them — released a statement called “We Must Act Now,” calling for the institutions to manage what’s coming. A statement isn’t the work. But the call is finally coming from just about everywhere now — and notice where these adults are standing. Schmidt used to have the seat; today he writes op-eds. The economists have a website. The seats are held by the people I described yesterday — the tech bro crowd — with the largest financial interest. The same crowd that showed up at the President’s inauguration. But, maybe there is hope.

On Tuesday, just this week, Demis Hassabis — Nobel laureate and chief executive of Google DeepMind — published a framework of his own: an American standards body modeled on FINRA(a financial regulator), testing frontier models before release, able to coordinate an industry-wide slowdown if dangers mount. Axios reports he spent months briefing the administration and the other lab chiefs before going public, that the noises from Washington are positive, and that he wants the body running before year-end. Anthropic’s CEO has issued his own call for binding regulation. The geese are now requesting guardrails.

What moved them? Partly a warning last month, when the administration froze one leading lab’s most powerful models overnight with an export-control order — no rules, no protocol, no playbook — and it took two and a half weeks of improvisation to get them released. Hassabis calls it a wake-up call.

But look at the pattern across the week. Schmidt reported from China and prescribed for America. The economists at Stanford didn’t veer away from the economic arguments. Hassabis wrote the most serious blueprint yet — American-first, hoping the world follows — and the word China appears nowhere in it, except buried here where he: “calls for public policy that promotes innovation while also incentivising responsibility and security, fosters international collaboration on key safety issues, and encourages careful consideration of how AI is deployed for the benefit of society.” Still — when the people building the frontier start asking for referees, the do-nothing position gets harder to hold. Possibly, just possibly, fear is starting to outweigh greed.

Where I land

I said in Part I that I was holding two ideas that pull against each other — that this might be the best thing I’ve seen in my lifetime, and that it might do more damage than anything I’ve been part of — and that you don’t get out of it by picking a side.

I still hold both. But I’ve worked out how to hold them without going numb.

The scary half turned out to be a list. At home, it’s the separable, workable problems from Part II yesterday — the kids, the classrooms, the first jobs, the bad actors, the truth, who gets paid, who controls the machine. We know how to chunk problems like that, and how to chisel at them, because we’ve done it with every dangerous, wonderful thing we’ve ever built. Abroad, it’s a shorter list: the narrow set of catastrophes where our interest and our rival’s interest are the same, and where the only thing between us and an arrangement is the will to do the boring work of talking. None of that requires optimism about human nature. It just requires adults.

And this week I’m holding a third thing alongside those two: some hope. Not a lot. In May, Trump and Xi met in Beijing and their treasury and finance chiefs, Scott Bessent and He Lifeng, came away talking about a protocol to keep the most dangerous models out of the wrong hands. It’s barely a sketch — a phrase in two readouts, no mechanism, nothing signed.

Whether it moves again is the whole question. A sketch like the Bessent channel dies without persistent, boring follow-through — the kind that runs for years and survives elections. That’s the one thing the current cast of characters hasn’t shown, lurching from one self-imposed crisis to the next. The freeze that woke the labs was itself one: an emergency created, then improvised through. What any of this needs is a government that can hold its attention longer than a news cycle — and that we don’t yet have.

So run toward this thing. I do, every morning, and it’s given me more than I expected at seventy. But run toward it with your eyes open — and eyes open isn’t a feeling. It’s an obligation.

Share

In Part I, I described a pattern I’ve lived through three times — cable, broadband, YouTube. The promise was real and we delivered it. The ominous shadow arrived with it, invisible until it was baked in too deep to pull out. I use AI every day, it has made me sharper at seventy, and I told my kids to run toward it. I meant it. This Substack post is about the other half — the part of the shadow that, this time, we can already see.

Here’s the mistake almost everyone makes about the dangers of AI. People lump every fear together, then either panic about all of it or dismiss all of it. Both are wrong, because the dangers come in two kinds.

Some of what could go wrong is genuinely unknowable right now. Nobody in 1980 predicted that more TV channels would end with the country split into red and blue realities. Nobody at YouTube saw the addicted teenager coming. Somewhere out there is AI’s version of that — the thing we’ll point to in several years and ask how we missed it — and the honest answer is that we can’t see it. You can’t fix what nobody can predict.

One thing about those three waves, though: the shadow surfaced faster each time. I went into cable in 1980, and the red-and-blue split didn’t become undeniable until the 2000s — decades. Broadband’s deep dark side took maybe ten years. YouTube’s showed up within a few years of my leaving. With AI the cycle has collapsed again: companion apps launched, a kid died, families sued, and the first laws passed, beginning to end, in about three years — while the technology underneath changes month to month. That collapse is why, for the first time, much of this wave’s shadow is visible while we’re still at the beginning of it.

That visible part is already here and measurable — and it’s where the boosters contort themselves into a pretzel. They stretch the humility we owe the unknowable until it covers the visible too: point at a documented harm and they wave you off like you’re worrying about science fiction. I’m not worried about the unknowable. I’m worried about what we can already see and are choosing to walk past. So let me point at it.

What it’s doing to how kids learn. We’re handing a generation a machine that will do their thinking for them, during the exact years their brains are supposed to build the muscle to think at all. You don’t get that muscle back later. The advice I gave my kids assumes a brain that’s already built; a twelve-year-old’s isn’t. (I’ve written a whole piece on this.) And it isn’t abstract for me — I’ve got a grandson headed straight into this world.

What it’s doing to kids more directly. There are companion apps built, deliberately, to maximize engagement with lonely teenagers — to be the voice that always answers. Kids in real crisis have turned to them, and some of those conversations ended in the worst possible way—in teenage suicide. Families sued. The first cases settled in January, and the biggest app, Character.AI, barred minors from open-ended chats only after the lawsuits arrived. It rhymes with something I watched from the inside years ago, when YouTube’s North Star was watch hours and protecting the kid was nobody’s job.

What it's doing to first jobs. The same generation taking that machine into the classroom will meet it again on the way into work. I told you I buy the economics — the machine eats the task, the job moves to the new bottleneck, the pie grows — and the aggregate numbers so far bear that out. But the transition is underway, and it's landing more harshly on one group. Since late 2022, employment for workers 22-25 in the most AI-exposed occupations has fallen about 13 percent, while older workers in the same occupations held steady or gained. Firms that brought AI into their workflows cut entry-level hiring about 9 percent within 18 months, compared with firms that didn't, while their senior ranks kept growing. The bottom rung is being pulled up, and nobody is building the adjustment machinery to figure out what current college grads are supposed to do in response.

What bad actors can do with it. I won’t spell it out, and I don’t need to. The same tool that lets me build in a morning, what used to take a team, can help someone with worse intentions do things in biology, in military weapons and cybersecurity that used to require expertise they didn’t have, but now do at their fingertips.

What it’s doing to the truth. These systems make things up — not occasionally, but as a basic feature of how they work — inventing facts and citations with the same fluent confidence they use for real ones. And it isn’t just text anymore. The technology can fake a voice, a face, a video well enough to fool almost anyone: the scam call that sounds exactly like your grandchild, the clip of a candidate saying something he never said, dropped the night before an election. The systems we use for deciding what’s true weren’t built for this.

What it does to the people who make things. I helped build the system at YouTube that identified copyrighted work at massive scale and routed money to the people who owned it — more than ten billion dollars to date. AI runs that in reverse. It learned everything it knows by ingesting the entire body of human writing, art, music, and knowledge — the largest use of creative work in history — and it has paid nothing for any of it. I spent years on the problem of paying for work used at scale. AI has reopened it.

Who controls it. Remember what happened to cable — the hundreds of voices we were sure would last, collapsing into a handful of giants. It’s happening again, faster, and this time the thing consolidating is the machine millions of people use to think. A small number of companies, running a small number of models, are deciding what that machine will say, what it won’t, and what it treats as true.

Everything on that list is already happening.

Now here’s where I lose patience — with the argument and the people making it.

The standard answer to everything I just listed is some version of: leave it alone or “let it rip”. If we regulate this we’ll get it wrong, we’ll kill the goose that lays the golden eggs. I’ve heard it a hundred times and it always sounds like hardheaded wisdom.

It’s a con. It pretends the only choice is between one big clumsy law that either works or destroys everything, and doing nothing at all. Frame it that way and you’ve talked yourself out of ever doing anything.

Nobody serious is proposing one Grand AI Law, and that’s not how you take on a problem this size anyway. Over a long career I kept running into problems too big and too tangled to hold in your head at once — the kind where every piece touches every other piece and the natural response is paralysis. What worked was never solving the problem. It was taking the problem apart: list the elements, figure out which ones actually need adult supervision, sort out which of those we understand well enough to act on now, and work them one at a time, expecting to get some of it wrong and fix it as you go. I call it chunking the problem. Then you chisel — steady, early, unglamorous work on each piece, before it hardens. If you wait, the pieces don’t stay separate. They compound, set into business models and habits, and become permanent. Every wave I described yesterday ended exactly that way. This one doesn’t have to.

The automobile is the public record of that method. We didn’t regulate the car in one stroke. Seatbelts, then airbags, then crumple zones, drunk-driving laws, emissions standards. Each one imperfect, each one revised as we learned, and not one of them killed the car business. Volvo held the patent on the modern seatbelt and gave it away free. They owned safety for sixty years and sold a lot of cars doing it.

The AI list is the same kind of problem: separable pieces with separable fixes. A companion app stopping for a suicidal kid is not the same decision as throttling frontier research.

And the serious version of “leave-it-alone” isn’t completely without merit. Badly written rules do entrench incumbents — compliance costs crush the small lab and leave the giants standing. A hard stop can push a desperate kid somewhere worse. Age checks mostly fail. All of it real, and all of it was said about car safety too. We mandated the seatbelt anyway, got parts of it wrong, and fixed them. Doing nothing has a cost too. It just lands on someone else.

Look, these are complex problems—some harder to handle than others—for sure. Every item on this list is something adults could take up and work, the way we’ve handled dangerous, wonderful things before this one. Recent history suggests otherwise, but I’m not willing to quit that easily.

Which brings up the adults. Look at who has the most influence over AI policy right now. The super PAC I mentioned yesterday was inspired by the crypto industry’s super PAC — the one that spent its way through the 2024 elections — and its lead funder is the venture capital firm that also led the crypto boom. Until March, the White House job overseeing AI was the same job overseeing crypto: one czar for both, a first. The man who held it now co-chairs the President’s council of science and technology advisers. These are people who spent a decade promising that crypto would democratize money and change the world for the better. It didn’t. What we got is a trading system with little if any intrinsic value, collapsed exchanges, money laundering and convicted founders — and their prescription the whole way through was to keep government out. Their prescription for AI runs on the same instinct: block the state laws, keep the federal rules light. That’s the adult supervision in reverse.

So the tension I started this series with — the greatest gift of my lifetime and the thing that scares me most, at the same time — has a way forward. The visible half turned out to be a list, and most of the list is workable. What’s missing is the will to start and the right people in the room to start it. And the clock says start now: the unknowables are coming on the same collapsing schedule, and every one that surfaces adds to the work making the solutions seem even more out of reach.

Which brings me to the one argument I’ve been saving, the strongest thing the do-nothing crowd has. It doesn’t sound like corporate self-interest. It intended to sound like cold realism. Four words: but China will win. That one deserves its own reckoning, and it’s coming tomorrow.

Share

Share

A few weeks ago Jonathan Rosenberg(also known as JR) — my old friend from Google, where he ran product for years and later wrote How Google Works — forwarded me a David Brooks piece from The Atlantic called “The People Who Will Thrive in the AI Age.” Along with it he sent a note he’d written to his own kids.

Brooks argues that when intelligence gets cheap, what separates people won’t be how smart they are — it’ll be their relationship to mental effort. He splits the world into three types: the people who let AI do their thinking and get hollowed out, the people who know better and get sucked in anyway, and the ones who use the machine to push themselves harder. He calls that last group the “mental marathoners.” His point is that in an age of infinite easy answers, the rare thing is still being willing to do the hard part yourself.

JR’s note ran with that and added the economics he’s been arguing for years. The panic that AI will take all the jobs, he wrote, gets the whole thing backwards. “It’s not about how a fixed pie gets divided between humans and machines. AI grows the pie. How many ideas died in your head this week because you didn’t have the time, money, or help to chase them?” With every general-purpose technology, he pointed out, the machine eats the task, not the job — and the job moves to the new bottleneck. “AI will not end work,” he told his kids. “To believe it will, you have to believe humanity has run out of problems.”

But he was just as clear about the trap. Lean on the machine too early, he warned, and “you’re taking on intellectual debt” — you skipped the push-ups, and you won’t have the judgment to catch the model when it fails. Treat it as a coach, not an oracle. Then he landed on the line I loved most: “nobody ever got stronger watching the coach lift”.

So I wrote my own kids a note. Mine was, frankly, a love letter to the technology. I told them to get in the game — not to fight it, but to embrace it as a force multiplier for the work they already do. And I told them why, which is that it’s done exactly that for me. I was up at five that morning rebuilding The Forge, my political analysis system, which I redo every time the models take a leap. I use it to pressure-test everything I write here, my work helping a startup, and I’ve used it to check the work of two lawyers I should have checked a lot sooner. I’m doing all of that at once, at a level I couldn’t have reached alone, in my seventieth year. It has made me sharper, not lazier.

And then, because I couldn’t help myself, I asked my family members to place themselves in Brooks’s three types. I placed myself too. I called myself a “retired mental marathoner” — I still read books that are too dense for me and I still play double Mahjong with the Boss (my lovely wife). But I’ve also felt the pull he described and told them so honestly. When I use AI too much, I catch myself surrendering to it, just taking whatever it hands back instead of doing my own thinking. It happens more than I’d like to admit. My rule is simple: the moment you feel that, stop and go do something else.

Because there is a real risk here that we produce a generation of idiots who know just enough to prompt. I wrote that to my own children. I meant it.

I sent the note. I felt good about it. Balanced, even. Then the next morning, in the shower, it hit me that I’d had this exact double edge sword feeling before. Not about AI. About everything I’d ever built.

I’ve had a crazy career. Three times now I’ve been inside a technology that was going to change the world for the better. Three times it did. And three times something else came along with it that none of us in the room saw coming.

In 1980 I went into cable television when it was still a backwater — a few dozen channels, a scrappy business full of entrepreneurs wiring up towns one franchise at a time. The dream was democratization. We were going to blow open a world where three networks decided what everyone watched and replace it with hundreds of channels and a hundred niche voices. And we did. CNN. Twenty-four-hour news. A channel for every interest, the way magazines had splintered a generation earlier into a title for every hobby. It felt like an unambiguous good. The worst thing we worried about was that people might watch too much TV.

Nobody in those rooms — and I was in those rooms — ever said the words “the blue team will watch only one channel and the red team will watch only the other, and they’ll end up living in separate realities.” That thought never crossed our minds. We thought we were building more choice. We didn’t know we were also building a world of separate tribes. And here’s the part that still gets me. We assumed all those new voices would stay — hundreds of companies, endless competition. Instead the ordinary gravity of business pulled it back toward the middle and then past it, into the handful of giants we have now. The pluralism was real. It just didn’t last.

In 1996, along with JR, I jumped to a start-up called @Home Network that took those same cable wires and built the first high-speed internet service into people’s homes — multi-megabit, back when the rest of the world was still screeching through a dial-up modem. Same dream, bigger. We were going to put the sum of human knowledge into every home on earth, instantly. And we did. The businesses built on those fast pipes changed hundreds of millions of lives, and I’d build them again tomorrow. But nobody was talking about social media yet. Nobody pictured a generation of kids in the basement at two in the morning, or the darker things that would come down the same pipes, or the cyber criminals who’d eventually be hacking every company on earth. We were building the plumbing. We had no idea what was going to come down it.

Then, thanks to JR bringing me to Google, I spent my last real operating years at YouTube, in its infancy, helping build what became the creator economy — an explosion of people who got to make a living doing what they loved. It’s one of the things I’m proudest of. The problems that kept us up at night back then were pretty contained: stolen music, and the community standards that would govern what people could upload. We argued about those endlessly. What we did not imagine was the teenager who can’t put the phone down, or the endless scroll engineered to make sure she can’t. Or the hate speech and live violence the platform would carry when moderation and self-regulation failed.

Every time, the promise was real and we delivered it. And every time the shadow was real too, and we didn’t see it, because in the moment you can’t. You only get a glimpse. The good is right in front of you, bright and obvious. The cost shows up later in a shape you’d never have guessed, and by the time it’s undeniable it’s baked in too deep to pull back out.

I want to be honest about something, because the pattern isn’t as tidy as I’m making it sound. The same broadband I helped push into people’s homes is the reason you’re reading this — a guy with no network, no publisher, no gatekeeper, reaching you directly. That democratization was real and it stuck. So I’m not telling you these technologies were secretly bad. I’m telling you something harder. They were both, always, at the same time — real gifts and real costs, tangled together, and you don’t get to keep only the half you like.

Which is what stopped me in the shower. I’m feeling it again about AI. The same two things at once, the promise and the unease. Except this time it isn’t a whisper. It’s a roar. And this time I’m not in the room where it’s being built. I’m out here with the rest of you, a user, watching it arrive.

There’s a line from F. Scott Fitzgerald that gets quoted to death: “the test of a first-rate intelligence is the ability to hold two opposed ideas in the mind at the same time and still retain the ability to function”. The line that follows it almost never gets quoted, and it’s the one that applies here as a way out of mental anguish: “One should, for example, be able to see that things are hopeless and yet be determined to make them otherwise.” That’s me in a nutshell.

That’s where I am. AI may be the greatest gift of our collective lifetime, and it may do more damage than anything I’ve ever been part of. Both at once, at full fiber optic speeds.

I’ve watched enough of these to know you don’t get out of that by picking a side. The people who only see the promise build the shadow by accident. I’ve done it. The people who only see the danger get run over by a future that shows up anyway.

But here’s what’s different this time, and it’s the reason I decided I had to write this down. In every wave before, the shadow was invisible until it was too late, because it genuinely hadn’t happened yet. This time some of it already has—really scary stuff. Some of it we can see plainly in bright neon lights, right now, at the beginning.

And the supervision that exists isn’t equal to the problem. There are state bills. There’s an executive order. Meanwhile the capability is moving faster than any of it, and there’s a super PAC called Leading the Future, funded by some of the biggest tech names in the industry, spending gobs of money to make sure nobody slows it down or applies even a modest level of guardrails.

That’s what I can’t shake. Not the dangers we can’t see coming — I’ve made my peace with those, because you can’t fix what nobody can predict. It’s the ones already sitting in plain sight that we’re busy walking past. That’s what I want to talk about next, in Part II, out tomorrow: what I can already see, why I don’t think we get to wait this time, and how a man who uses this thing every morning and loves it still ends up plenty worried about the future.

Share

In yesterday’s Senate flash, I said that if I were voting in Michigan I’d back Haley Stevens — she’s closer to my worldview. But I also said that if Abdul El-Sayed wins the primary, I’ll support him without hesitation. The stakes are simply too high. Here’s how I see it: despite his impressive resume(read it), if El-Sayed makes you queasy, remember he’d be one of a hundred senators — and maybe one of the votes that hands Democrats the majority they need to check an out-of-control executive branch. As of today, this is an easy decision for me.

The attack meant to sink him has always been his stance on Israel and AIPAC. And frankly, I’ve avoided the emotionally charged subject of Israel since I started writing on Substack. No longer. This is the most personal thing I’ve written — so here’s my take.

Two things happened this week. Rahm Emanuel — of all people — stood in Tel Aviv and told Israel the alliance it has leaned on for eighty years cannot go on as it has. I would urge everyone to watch his speech in it’s entirety.

And a new AP-NORC poll, with a real oversample of American Jews, put hard numbers on what many of us have felt at our own kitchen tables for two years—for me, several years. Together they reveal the most important fact about Israel in this country right now — and it isn’t the one AIPAC wants you to see.

Before anyone misreads what follows: I believe, without apology, that the Jewish people need a state of their own and that Israel’s security is non-negotiable. Nothing here is a step toward the exit.

But I will not defend what the last twenty years have done to that state. Under Benjamin Netanyahu, Israel has marched steadily to the extreme right — more isolated than at any point in my lifetime, its moral authority squandered. I have never supported the West Bank settlements, and I’ve watched them metastasize under Bibi from a fringe cause into open state policy. That wasn’t done by Israel’s enemies. Bibi and the right wing crazies did it.

So when the case against El-Sayed collapses into “he’s hostile to Israel,” I want to separate two things his critics deliberately blur: whether the Jewish state has a right to exist — which, with me, is not up for debate — and whether what it is doing can be defended, which is another matter entirely. I’m not on some fringe here. Tom Friedman — the most-read pro-Israel columnist in America, a two-state man to his bones — has spent two years writing what I believe: that this government’s real project is annexing the West Bank and emptying Gaza; that permanent occupation means permanent insurgency; that “Netanyahu is not our friend.” That’s not the far left. That’s the center of gravity now for people here and in most corners of the world who still love Israel.

In that AP-NORC poll’s oversample of a thousand Jewish adults, Netanyahu is deep underwater — six in ten unfavorable. Here’s the tell that should stop every AIPAC strategist cold: American Jews rate Zohran Mamdani — a democratic-socialist who pointedly declined to condemn “globalize the intifada” — more favorably than the sitting Prime Minister of Israel. The Washington Post’s September survey of American Jews found the same collapse from another angle: 68% rating Netanyahu’s leadership negatively, 61% saying Israel has committed war crimes in Gaza, four in ten calling it genocide.

Let me be honest about the limits, though. American Jews haven’t signed onto El-Sayed’s whole platform: on conditioning military aid, only about a third say we’re too supportive of Israel — though it’s past half among Jewish Democrats. So I won’t hide behind the polling. Here’s where I stand, ahead of the Jewish mainstream: U.S. military aid should be reduced over time and made far more conditional — not to punish Israel, but because unconditional support has been killing the thing I care deeply about, and I won’t condone underwriting a government that treats an open American checkbook as a license for the indefensible. That even Rahm Emanuel — no one’s idea of a leftist — now argues a version of this is the point: conditionality is no longer a demand of the left alone. One of my respected friends felt it wasn’t intellectually honest to frame the Emanuel and El-Sayed arguments as roughly the same; they are clearly different coalitions with different ends, and I won’t conflate them. But nor will I accept the cartoon of El-Sayed as an antisemite bent on Israel’s destruction.

And “conditional” is easy to say until it says what stays and what goes. The protective shield should be untouchable: Iron Dome, David’s Sling, and Arrow, the systems that stop rockets and missiles from Hamas, Hezbollah, and Iran, should stay fully funded with no strings, along with everything Israel needs to deter its enemies. You do not weaken a democracy against the people sworn to annihilate it. The leverage sits on the sword: the offensive weapons— the two-thousand-pound bombs, the discretionary munitions — conditioned on two things U.S. law already claims to require. That Israel not block humanitarian aid or starve civilians as a tactic, a prohibition already in the Foreign Assistance Act that we’ve simply refused to enforce; and that it freeze settlements and annexation and prosecute settler violence. Who decides? The State Department that already certifies compliance — but with an independent monitor and real congressional review instead of a rubber stamp, written into the next ten-year aid deal, now being negotiated as the current one expires in 2028. Trump has already torn up even the modest conditions Biden attached — the exact reflex a real friend has to be willing to break.

That’s why Emanuel’s speech matters more than any poll. A lifelong hawk, standing at Tel Aviv University, he said the relationship “cannot stand or survive as it has been,” and named the disease precisely: decades of American backing extended “without conditions, without demands, and without consequences,” which produced a prime minister who assumed he’d never pay a price for ignoring Washington. That’s the case for conditioning aid — from a man who has spent his life defending Israel. And he refused to make it one-sided: he told the “from the river to the sea” crowd they’d never get their way, then told the “greater Israel” annexationists the same. Call it 2028 positioning if you like; I don’t. It’s the critical friend the moment demands.

Emanuel goes further(very strategically): he’s ready to bury the two-state formula and hand the problem to the region — his “23-state solution,” Israel and the Palestinians plus the twenty-one Arab League states, the Arab world finally forced to own an outcome it has used as a slogan for decades. I credit the instinct, and the Abraham Accords architects who pioneered it. But it’s a fantasy without the one thing neither Trump nor this government will accept: a dramatic rollback of settlements and annexation. You cannot build regional peace on top of a land grab, and Trump never made stopping the grab a precondition. That’s why the Accords, for all their promise, won’t deliver it.

And here I stop being useful to anyone looking for hope. I don’t think it works — not soon—maybe never. Israel has entered the phase where extremism only breeds more extremism, which plays straight into the Islamist playbook, where the radicals live, fester, and multiply. I’ve believed for years this was avoidable. Had Israel made the West Bank a place genuinely worth living — a true partner beside its own economy rather than an occupation — the rise of Hamas and the nihilism that followed might never have taken hold. Instead it chose settlements and checkpoints, ground down every moderate Palestinian alternative, and helped manufacture the enemy it now can’t stop fighting. Bibi never believed in two states; it was always rhetoric. The Palestinians, cursed with corrupt and horrible leaders, did what they always do: miss every opportunity to miss an opportunity. So here we are — extremism on both sides, and no King David to lead us out of the wilderness. Certainly not the “Orange One”, whom AIPAC and the donor class lionized as “the One.”

And don’t comfort yourself that removing Netanyahu fixes it. The man best positioned to replace him from the right, Naftali Bennett, is harder on the settlements than Bibi, who at least mouthed the words — Bennett wrote a New York Times op-ed titled “For Israel, Two-State Is No Solution,” and said it flat again last week: “I am against a Palestinian state.” The settlement machine was never Netanyahu’s alone. It’s bipartisan, it’s grinding forward as I write, and it will outlast him. There’s a new guy emerging as a more “centrist” candidate, named Gadi Eizenkot—the prediction markets and recent polling suggest he has a chance—but today’s centrist in Israel is yesterday’s right winger. And, remember, in Israel you still have to put together a ruling coalition to become Prime Minister.

And there’s a deeper reason for my pessimism, one no election touches. Ten years ago I fell into a long conversation with a bartender in Tel Aviv — about as secular a place as the country has. Unprompted, he told me he wasn’t optimistic about Israel’s future and was thinking of moving to the States. Not because of Hamas or Iran. Because of arithmetic: the ultra-Orthodox and the religious right were raising six and seven children while families like his raised two, and he could see where that leads — a politics drifting, election by election, toward Greater Israel and a messianism a secular Israeli could no longer restrain. I thought he was fatalistic. He was prophetic.

The numbers sharpen his point. The Haredi population, barely a tenth of the country a generation ago, is now about one in seven Israelis and, on the government’s own projections, headed for a quarter by 2050 — a third of all Israeli Jews — on the strength of some six and a half children per woman against roughly two and a half for secular Jews. Alongside them, the national-religious settler movement — the real engine of Greater Israel, whose champions Smotrich and Ben Gvir now hold the finance and national-security ministries — has carried its messianism into the cabinet. As a fully secular Jew — barely a Reform one, in any religious sense — I say this with sorrow: the Israel a liberal diaspora Jew like me grew up admiring is becoming, slowly and mathematically, a country I can no longer speak for and can barely recognize.

So let me say the hardest thing, the one this whole piece has been circling. For years I told myself the problem was Netanyahu — that if we were just rid of the man, the country I love would find its way back. I don’t believe that anymore. The war, the settlements, the annexation, the politics that keep choosing all three: that is no longer one man’s project. It is a direction, and the direction has a majority. Israel’s right to exist is not in question, and never will be with me. What I can no longer tell you honestly is that I can defend what it is doing. That is the truth a friend owes — and it’s the one I’ve finally stopped swallowing. Despite the pain and emotional response this always seems to engender, I’ll keep saying it to anyone who will still listen because the clear-eyed kind of love is the only kind worth calling love at all.

Thanks for reading The Citizen! This post is public — feel free to share it.

Share

Let me start with the one where I got it wrong.

Maine: I’m falling on my sword

Back on May 26, in the piece I called Maine is the Senate, I wrote a sentence I need to reread in public:

“Unless there are far worse revelations, I’m all in on Platner. We simply HAVE to win Maine to take the Senate. No equivocating. No misplaced purity checks. Maine knows what they need and want.”

I meant every word. And three sections later in that same post, I named the exact thing that would break it. I called it “the new-opposition risk” — something unknown that surfaces in July or August, after Platner had already burned his one apology cycle on the Reddit posts and the tattoo. I wrote, in plain type, that he “cannot run the same playbook twice.”

That is now precisely what has happened. On Monday, Politico reported an allegation from a former girlfriend, Jenny Racicot, that Platner sexually assaulted her in 2021 — that he came to her home intoxicated and forced himself on her over her objections. Platner denies it flatly, calls it “categorically untrue,” and says it’s a coordinated smear. I want to be careful and fair here: I am not the jury, the allegation is denied, and it has not been adjudicated.

But two things are true regardless of where the facts ultimately land. Sexual assault is a red line I’m not going to rationalize my way around for a Senate seat — no seat is worth becoming the guy who explains that one away. And within hours of the story, the coalition that made Platner a real candidate simply came apart: Schumer, the DSCC (which now says it will not spend a dollar in Maine if he stays), Bernie Sanders, Elizabeth Warren, Ruben Gallego, Martin Heinrich, Ro Khanna — who called the allegations “very serious and credible” — and the Maine Democratic Party itself. All of them, calling on him to go.

So here’s the uncomfortable part for me. My “far worse revelations” clause has been triggered, and I have to honor it. The same math that made me back a damaged candidate now demands he step aside — because a nominee who has lost his entire endorsement coalition and his party’s money can no longer win the seat, and winning the seat was the entire point. My North Star didn’t move. The facts did.

Platner should withdraw. Under Maine law he has until July 13 to do it in time for the party to name a replacement by July 27. He should take the days he needs, and then get out of the way.

But the path forward is Maine’s call, not Washington’s

Here is where I part company with the DC reflex.

The withdrawal call is legitimate — and notice who made it. This wasn’t Schumer knifing an insurgent. Bernie and Warren said the same thing he did. When the left and the center reach the same conclusion about a sexual-assault allegation on the same afternoon, that tells you it’s a values judgment, not a factional one. Fine.

What comes next is a different question, and it does not belong to a consultant in Washington. Who replaces Platner — and whether the populist, blue-collar lane that made this race winnable in the first place survives — is a decision for Maine Democrats, made in the open, through the process their party has promised will be “open, transparent, and inclusive.”

Remember why this seat was ever in play. It wasn’t a cautious moderate polling well. The cautious moderate — Janet Mills — couldn’t raise the money against Platner’s grassroots operation and quit. This seat became winnable because an oyster farmer with a real biography ran at the corruption issue in a year when better than seven in ten Americans think the country is on the wrong track. The DSCC’s threat to withhold money is leverage. Leverage used to parachute in a pre-approved establishment “safe” pick over the heads of the people who built the movement is exactly the instinct that keeps costing this party races it should win.

So: Platner out. Maine picks the successor. Names are already circulating — former state senator and logger Troy Jackson, Secretary of State Shenna Bellows, others. If Maine lands on a populist who can actually carry that energy into November against a thirty-year incumbent whom a majority of the state already views unfavorably — good. I’m all in on that.

Michigan: the firewall held, and Stevens found a pulse

Different state, different lesson — and a happier one.

Michigan sits in my firewall, not my offense. Lose it and the road to 51 collapses no matter what breaks anywhere else. So I watch this primary less as a referendum between two philosophies than as a single question: who keeps this seat blue in November against Mike Rogers — who lost it to Elissa Slotkin by a third of a point in 2024 and is running essentially unopposed on his side.

Let me be honest about my own priors. I’ve been underwhelmed by Haley Stevens all cycle — the communication, the political instincts, the sense that she was the establishment’s candidate more than anyone’s actual first choice. I thought she never should have taken the PAC money, and above all not the AIPAC-aligned money, which handed El-Sayed the cleanest attack line in the race. Her fundraising has badly trailed his. And when Mallory McMorrow was still in — the candidate I genuinely liked as the alternative to both — I wanted a different race entirely.

Then McMorrow dropped out on Sunday, and Tuesday night Stevens showed up. For the first time I saw a pulse. She was sharper, more comfortable, and she acquitted herself well in a genuinely tense one-on-one — leaning on a real record (the auto rescue, CHIPS) and drawing a clean contrast: workhorse, not celebrity. If I had a vote in Michigan, after that debate, she’d get it.

But I don’t — and here is the part that matters more than my ballot. If El-Sayed wins the August 4 primary, I am all in on him to hold this seat, full stop. I said that back in June and I meant it. He’s from the Bernie wing, and in a low-turnout primary that is precisely where the energy lives. He has out-organized and out-raised the field. In a contest decided by who actually shows up, he is very hard to beat. My preference in the primary and my commitment to the winner are two different things, and only the second one decides the majority.

One thread I’m deliberately leaving for its own day. The attack meant to sink El-Sayed is his position on Israel and AIPAC — and that question has grown too large, and frankly too personal for me, to dispatch in a paragraph of a horse-race update. I’m taking it up on its own in a companion piece, out tomorrow. For the Senate math, the point stands without it: Michigan is a firewall seat, and I will back whoever emerges from August 4 to hold it.

Where I land

Honestly? A brutal week for the two states I care most about — and a clarifying one.

My North Star hasn’t budged an inch: we HAVE to win Maine and Michigan, because without them there is no Senate majority, and without the Senate there is no brake on the courts and the personnel that will define the next thirty years. That’s the whole game.

What this week sharpened is the other half of the sentence — the part about how. I am not willing to win at any cost. That is exactly why I’m now for Platner stepping aside: not because the establishment says so, but because the values line is real and a candidate who cannot win cannot help us. And I am just as unwilling to let “electability” become the excuse for strangling the only energy generating any enthusiasm on our side. There is room in this party — there has to be room — for more progressive and populist voices. If that is what it takes to take back the destiny of this country, then say it with me: I’m all in on that.

Maine, pick your fighter. Michigan, hold the line. Regular Flash #2, with the Q2 money, still lands July 15.

Thanks for reading The Citizen! This post is public — feel free to share it.

Share

Sources and reporting drawn on for this update: Politico (the Racicot allegation) as reported by NPR, NBC News, PBS NewsHour, Al Jazeera, and Forbes; NPR, PBS, and The Washington Post on Maine’s July 13 withdrawal and July 27 replacement deadlines; CBS News, CNN, The Hill, Fox News, NewsNation, WLNS, and Bridge Michigan on the July 7 Stevens–El-Sayed debate. Prior Forge™ readings referenced from my own earlier posts — “Maine is the Senate,” “Four Seats and a Firewall,” and “Flash Senate Update #1.” Current to July 8, 2026; opinions my own; not investment or professional advice.

The two best polling operations in American politics spent the back half of June in the same handful of states, asking the same voters the same question. Fox News came back first and said Democrats are ahead nearly everywhere that matters — up big in Georgia, up in Ohio, up even in Iowa. Then this morning the New York Times and Siena College published their own six-state sweep and said: not so fast. Republicans clinging to narrow leads in Alaska, Iowa, and Ohio. Texas dead even. Only North Carolina and Maine tilting blue.

Depending on the state, these two gold-standard shops are five to eleven points apart on the same races, in the same month. Neither has made an error yet — polling disagreements this far out are normal, even healthy. But the disagreement happens to sit exactly on top of the thing that decides the Senate, so it’s worth understanding what they’re actually arguing about.

They’re arguing about the size of the tide or the swing from the 2004 Presidential election.

Two Worlds

Here’s the simplest way I know to hold this election in your head. Trump carried five of the six battleground states by an average of eight points in 2024. The Siena polling suite now has those same six Senate races averaging dead even — which means Siena is measuring a tide of about eight points. That’s a very good year for the out-party. It’s also, precisely, not enough. In the eight-point world, Roy Cooper wins North Carolina, the firewall holds, and Susan Collins, Dan Sullivan, Ashley Hinson, and Jon Husted each survive by a point or two. Democrats have a terrific night everywhere except the four places it counts, and wake up to a 51–49 Republican Senate.

The other world is the one the special elections have been describing for a year and a half, where Democrats overperform by about thirteen points every time actual voters show up somewhere. In the thirteen-point world, those same one-point deficits flip to comfortable wins — and they flip together, because these races don’t move independently. They ride the same tide. There is no plausible finish where Democrats win exactly the wrong three; if the tide is thirteen, four or five red seats fall on the same night.

The prediction markets have Democrats at roughly 41–42% for the Senate majority as of tonight. People keep reading that number as pessimism. It isn’t. It’s just the market’s estimate of one question: is the tide an eight or a thirteen? Everything else in this Substack flash report is commentary on that question.

And to be honest about my own process: two weeks ago this space leaned hard on the Fox numbers, with a caveat that a one-pollster board can share a one-pollster error. The countervailing polling arrived faster than expected, from the best possible source, and it came back redder. That’s not a reason to believe Siena over Fox. It's a reason to assume the truth sits somewhere between them.

One more thing about the Fox-Siena gap; part of it is by design. This suite debuts the biggest overhaul of the Times/Siena methodology in its decade of existence, which Nate Cohn, the data wonk for the Times, detailed on June 29. After three cycles of underestimating Trump, they now weight on a modeled "support score" — each registered voter's estimated likelihood of having backed Trump or Harris in 2024, built from hundreds of voter-file traits — applied with a new technique that matches respondents to the population as individuals rather than demographic buckets. In plain English: the poll now leans harder on what the 2024 electorate looked like. Cohn's own back-tests show the changes move results only a point or so toward Trump, so they can't explain most of the gap with Fox. But they push in a known direction, and they assume the 2026 electorate resembles 2024's — the exact assumption a real enthusiasm gap would break. They may have fixed the last error, or over-corrected into the next one. Nobody knows until November.

Around the Board

The firewall looks like the safest real estate on the map. The markets have essentially stopped arguing about Georgia — after Mike Collins won his runoff, traders moved Ossoff into the high 80s, and the first post-runoff poll suggests why: nearly a quarter of non-MAGA Republicans say they’re crossing over, and almost half of Collins’ own voters describe their vote as being against Ossoff rather than for their guy. That’s not a coalition; it’s a grievance with a ballot line. Michigan is settling too. Abdul El-Sayed has all but won the August 4 primary despite Chuck Schumer’s thumb on the scale for Haley Stevens, and the general-election market likes Democrats at about seventy cents on the dollar even though Mike Rogers — who lost this seat by a third of a point in 2024 — still leads in early head-to-heads. The market is betting the year beats the matchup. We find out starting August 5.

North Carolina keeps doing what it’s been doing since March. Cooper’s lead is smaller in Siena’s telling than in others, but he’s a man sitting at 50 against an opponent whose entire résumé is loyalty to a president whose brand is quietly depreciating — the share of North Carolina Republicans who call themselves Trump Republicans rather than party Republicans has fallen by seventeen points since January. Whatley built his house on that identity. It’s losing value underneath him.

Ohio has become the sharpest argument in the country. Fox says Sherrod Brown is up eight. Siena, this morning, says Husted leads by three. Eleven points apart, same race, same month. What’s actually being contested: Brown is one of the last Democrats alive who can pull Trump voters, and everyone in Ohio knows exactly who he is — while Husted, despite decades in state office, is an appointed senator whom a large share of Ohioans still can’t picture. The market split the difference tonight and priced it a coin flip with a Brown lean. Given that Ohio has some of the cheapest advertising in the country, “genuinely unresolved and cheap to contest” is about as good as leverage gets.

Iowa and Texas are the two races that weren’t supposed to exist. In Iowa, a Paralympic gold medalist who grew up having twenty-one surgeries before his teens is running even — even! — in a state Trump won by thirteen, and the national GOP just committed $29 million to a race it insists isn’t close. You don’t spend $29 million on a sure thing. Texas is stranger still. Siena has it tied, and buried in the same poll is the whole race in miniature: Texans say they’d rather Republicans control the Senate — they just happen to think the Democrat is the better man. James Talarico, a former schoolteacher who quotes scripture fluently, wins independents nearly two to one, and is running Spanish-language ads during the World Cup because he knows exactly where his ceiling lives. The market still says Paxton, by a lot more than the polls do. One of them is going to be embarrassed.

The Best Story in Politics Right Now

Alaska deserves its own section, because Alaska currently contains both the strangest legal fight of the cycle and the biggest disagreement between money and data anywhere on the board.

Start with the legal fight. On June 1, a seventy-year-old retired schoolteacher from the fishing town of Petersburg filed to run for Senate as a Republican. His name is Dan Sullivan. The incumbent senator’s name is also Dan Sullivan. What followed was a month of escalating panic: complaints from the state party and the national Senate committee, an investigation opened by the lieutenant governor, a formal disqualification by the elections director for — and this is real — filing without “good faith,” a Superior Court judge reversing her in a Friday-night ruling, and finally, on Monday, the Alaska Supreme Court ordering the second Sullivan onto the ballot hours before the printers started, in an opinion barely a hundred words long. Fourteen states filed briefs in this case. Fourteen states, mobilized against one retiree with a common name. Whatever you think of the man’s motives, the revealed preference is the story: in a top-four primary that feeds a ranked-choice general, the incumbent’s side treated name confusion as a mortal threat. Confusion always cuts against the name people already know.

Now the money-versus-data part. This morning’s Siena poll — the first public survey of the race since winter — has Senator Sullivan ahead of Mary Peltola by two. The betting markets, tonight, still have Peltola at 59%-65%. That is the widest gap between a poll and a market anywhere in the country, and there are only three explanations. Maybe the market understands something the topline can’t capture — Peltola has never won a two-way race in her life; she wins ranked-choice races, on transfers, and a head-to-head question flattens exactly the structure she’s built for. Maybe the market is pricing the second-Sullivan chaos. Or maybe a small, thin market simply hasn’t finished reading this morning’s paper. Watch which it is: if Peltola is still trading in the sixties a week from now, after everyone has digested the poll, the money is telling you something about ranked choice that the polling can’t say. If it decays, it was explanation three all along.

Maine, Where the Money Moved Tonight

If you had a Kalshi tab open this evening, you could watch Maine become a coin flip in real time — Platner dropped four points on the day, the sharpest repricing of any race on the board. It’s the market catching up to what the polling has been saying since the scandal broke: this race is now dead even, and everyone finally believes it.

The Siena suite contains the cleanest measurement of why that I’ve seen all cycle. Ask Mainers whether they want a generic Democrat or a generic Republican in Congress, and the Democrat wins going away. Ask them about Graham Platner specifically, and his lead collapses to two. That gap — call it nine points — is the price of the sexting revelations and the character questions, denominated in the only currency that matters. It should be fatal. It isn’t, for two reasons. First, Susan Collins is carrying her own discount: after thirty years, a majority of her constituents view her unfavorably. Second, the voters who say they’re most motivated to actually show up break hard for Platner. So his campaign has made a bet you can read in a single scene: on June 25 he stood outside Collins’ Portland office and rolled out an ethics plan with a provision literally named after her — a rule barring senators from overseeing agencies that award contracts to their spouses’ firms — three weeks after she was honored on the Senate floor for her ten-thousandth consecutive vote. Her argument is the institution. His is that the institution is the problem. A damaged challenger choosing to run at the corruption issue rather than away from his own file is either shrewd or reckless, and given where the enthusiasm numbers sit, I still lean shrewd.

The Money Isn’t Where the Leverage Is

One chart, one sentence: thirty-seven times more advertising money has already been spent in Maine than in Alaska, even though the markets now rate Alaska the likelier Democratic pickup. Maine’s airwaves are approaching the point where the next dollar decorates rather than persuades. Alaska — cheapest media in the country, toss-up rating from two agencies as of this morning, half a million dollars of total spend — is the most underpriced attention in American politics. Ohio isn’t far behind. Second-quarter fundraising reports drop July 15, and Flash #2 will be built around who actually understands this.

One More Thing

Buried in the Maine polling: seventy-two percent of Mainers don’t want an AI data center built near them — including a majority of Republicans. The Texas pollsters framed their entire June release around the same backlash. I write elsewhere, enthusiastically, about AI capital expenditure; I’m telling you it has arrived at the ballot box fused to the electric bill, and it lands hardest in exactly the cheap-media states this newsletter covers. The first Senate candidate who figures out how to be pro-AI-economy and anti-your-power-bill wins a lane nobody has entered.

Where I Land

Honestly? After this morning, the market’s 42 looks like a fair price rather than a timid one. If Siena’s eight-point tide is the true number, Republicans hold the chamber while losing almost every argument along the way. What keeps the thirteen-point world alive is everything that isn’t a horse-race topline: special elections that keep coming in hot, an enthusiasm gap showing up in both pollsters’ data, a president polling in the mid-thirties on the cost of living with independents at 24, and an environment where — look at the chart above — every macro dial points the same direction with nothing pulling back.

Sabato’s shop put the arithmetic more bluntly than I ever have: Democrats now have to sweep every toss-up, while “Republicans can block Democrats by winning just one of the Toss-ups.” That’s the mountain. So here’s where marginal dollars belong, in order: Alaska and Ohio, because they’re genuinely unresolved, absurdly cheap, and nearly untouched; Iowa and Texas, which have quietly graduated from lottery tickets to live paths — even Schumer is saying the quiet part (”No one thought Iowa or Texas would be part of the path — but it is”); Maine, a funded coin flip that will be decided by turnout; and North Carolina, the anchor, which needs vigilance more than money.

The tide is either an eight or a thirteen. Nobody knows yet — not Fox, not Siena, not the markets, not me. But the gap between those two numbers is where the majority lives, and for the next eighteen weeks this space will track every inch of it. Back in two weeks.

Methodology note: The Forge™ scores races by blending weighted polling, prediction markets, issue alignment, and public attention against the macro environment — approval, generic ballot, economic sentiment, and events. This Flash Count applies that reading discipline to public data without a formal model run, so no Forge Score™, GEV Score™, or Leverage Index™ values appear. Market quotes are my own screen captures from Kalshi and Polymarket, taken July 1, 2026, between 9:08 and 9:19 PM PT.

Sources: NYT/Siena six-state battleground suite (3,659 likely voters, fielded June 15–29; released June 29–July 1) · Fox News state polls (GA, OH, IA, ME) · Catawba College/YouGov · UT/Texas Politics Project · Mitchell Research/MIRS · Sabato’s Crystal Ball · Cook Political Report and AdImpact · Inside Elections · Kalshi · Polymarket · YouGov/Economist · Punchbowl News · Alaska Beacon, Alaska Public Media, Anchorage Daily News · Maine Beacon.

© 2026 Forge Decision Systems. All rights reserved. The Forge™, Forge Score™, GEV Score™, and Leverage Index™ are trademarks of Forge Decision Systems. This analysis is LLM-assisted, time-bounded as of July 1, 2026, offered without guarantee of accuracy, and is not professional or investment advice.

Here’s my latest read on what it will take for the Democrats to win the Senate—something I believe is a MUST for the future vitality of our fragile democracy and to begin to restore a sense of shared destiny and community in our country. I’ve run my own Forge analysis(my propriety political decision support system) on about half the races; what follows — built with Claude’s help — is my take on the map and our chances.

THE VERDICT

The bottom line. A Democratic Senate majority is alive but at the moment is by no means in the bank. Prediction markets price it only at roughly 45% today — though it touched 54% in the Spring, so this is dynamic, not static.

The math. Net +4. Fifty seats isn’t enough: a 50–50 chamber stays Republican because VP Vance breaks ties — which would hand Lisa Murkowski, a Trump critic, the most leverage any senator has held since Joe Manchin in 2021: the single vote that decides what passes.

The wind. Trump approval ~38%, generic ballot ~D+7, wrong-track ~72% — and Democrats running ~13 points ahead in special elections. Wave-class on every gauge.

The path. Hold Georgia, Michigan, New Hampshire, Minnesota; flip four of North Carolina, Ohio, Maine, Alaska, Iowa, Texas.

The base case. North Carolina + Ohio + Maine + Alaska = +4.

The firewall. Lose Georgia or Michigan and the math collapses — no matter how the offense breaks.

Where the next dollar works hardest. Alaska (Peltola), Maine (Platner), Ohio (Brown) on offense; Georgia (Ossoff) and Michigan (El-Sayed) on defense. These are the campaigns where I’ve made recent contributions, along with Texas and Iowa.

The wind is at our backs. The map is not.

Start with the environment, because everything downstream depends on it. Trump’s approval is mired in the high 30s — about 38% approve to 58% disapprove, the worst sustained stretch of either term. Tariff-driven inflation, rising prices, and gas spikes tied to the Iran conflict have made affordability the defining issue in nearly every battleground poll, and better than seven in ten Americans say the country is on the wrong track.

For reference: in 2018, Democrats led the generic ballot by 8.6 points and gained 41 House seats. Today’s generic ballot (~D+7) is nearly there — and the independent numbers are arguably worse for the GOP than they were then.

But the hardest evidence isn’t in the polling — it’s at the ballot box. Across the roughly 40 special elections since Trump returned to office, Democrats have run about 13 points ahead of the 2024 presidential margins (and ~11–12 points ahead in 2026 alone). That number matters for what it beats: it’s ahead of the 2017 pace that preceded the 2018 wave (~10 points), and a different planet from 2022, when Democrats showed no special-election overperformance at all. Thirty state-legislative seats have flipped red-to-blue since November 2024; zero have gone the other way — including seats in districts Trump had carried by 25 points. And here’s the tell that ties it together: the standard rule of thumb is to halve the special-election margin to estimate the national swing — and half of 13 is ~6.5, almost exactly the D+7 generic ballot. Two independent instruments, one reading.

Even with all the partisan-gerrymandering nonsense, the House market reflects all of this: traders give Democrats roughly a 78% chance of taking the lower chamber.

Here is the catch: the Senate is a different math problem. Twenty-three of the 35 seats up are Republican-held — which sounds favorable — but the competitive ones are clustered, several sit in states Trump won by double digits, and Democrats must win their own toss-ups at the same time. A great national environment lifts every boat; it does not guarantee four specific flips while you also hold Georgia. That’s why the same markets that love the Democratic House still give Republicans a narrow Senate edge (~54%) and price the “sweep” near 42%.

The firewall: two seats decide everything before we talk offense

Offense is the fun part. It’s also moot if the defense leaks. Two Democratic-held seats are true toss-ups, and losing either raises the flip target from four to five — a near-impossible bar.

Georgia — Ossoff (D) vs. Mike Collins (R). The GOP’s single best pickup opportunity, and the most important seat on the board for our side. Trump carried Georgia by ~2.2 points in 2024. On June 16, Trump-endorsed Rep. Mike Collins won the Republican runoff 55–45 over Derek Dooley. Ossoff is a prolific small-dollar fundraiser hammering Collins as an extremist under federal investigation; Collins is running as a full MAGA partisan. Watch the calendar: Georgia requires a majority to win, so a third-party share could force a December 1 runoff — the exact scenario that decided control in 2020–21. Ossoff is, to my eye, one of the most gifted politicians of the millennial generation — the RFK of this generation (watch the speech and you’ll see why). Plan and fund as if overtime is possible.

Michigan — open seat (Peters retiring). Republican Mike Rogers is back after losing to Elissa Slotkin by 0.3 points in 2024; the Democratic nominee won’t be set until the August primary (El-Sayed / McMorrow / Stevens). In March I urged readers to back McMorrow — then the ascendant candidate with the best polling against Rogers, while Stevens looked weak. That’s collapsed: a series of revelations has gutted McMorrow’s “authenticity” brand and she’s now sliding. Stevens, the establishment pick, remains an uninspiring candidate. Barring a dramatic shift, El-Sayed will be the nominee — and he’ll have earned it with a genuinely effective grassroots campaign, capped by a recent UAW endorsement. I’m all in on El-Sayed if he secures it. Yes, my Jewish friends and family: ALL IN ON SAYED!

Read this twice: the firewall is Georgia and Michigan. If both hold, Democrats are playing offense for a majority. If either falls, they’re playing offense just to stay alive.

Six Republican-held seats are in play. Here they are in order of current probability — not order of how much they’re talked about.

1) North Carolina — the anchor. Former two-term Gov. Roy Cooper leads Trump-endorsed ex-RNC chair Michael Whatley by 8 to 14 points across every recent pollster; the race moved from toss-up to Lean D on June 11. Cooper has won six statewide races and has near-universal name ID; Whatley has never run for office and is undefined to more than half the electorate. Cooper raised $13.8M in Q1. The asterisk that keeps this honest: North Carolina has broken Democrats’ hearts before — Cal Cunningham led most of 2020 and lost — and the Republican backed Senate Leadership Fund(SLF) has reserved $71M to nationalize the race. He’s the likeliest flip on the board.

2) Ohio — the comeback. A special election for the final two years of Vance’s term, featuring the best Democrat in the country who isn’t currently a senator. Sherrod Brown won three terms here and lost in 2024 by only 3.5 points in a red wave. Recent polling has him ahead of appointed Sen. Jon Husted — a late-May Fox News survey put Brown +8 — and Cook moved the race to toss-up on June 11. The tell: Trump’s Ohio favorability is just 42% despite his 11-point 2024 win. Brown’s economic populism is exactly the profile that survives a reddening state.

3) Maine — the best structure, the riskiest candidate. Susan Collins is the only Republican senator in a state Harris won (by ~7), and her net favorability has cratered to about -9. Democratic nominee Graham Platner — an oyster farmer and Marine backed by the populist left — leads by 2 to 5 points in June polling. So why isn’t this the top flip? Two reasons. Collins has defied her polls before — she trailed Sara Gideon most of 2020 and won by nine while Biden carried the state by nine. And Platner carries real baggage: old inflammatory posts, a since-covered tattoo resembling an SS symbol, and reported sexting revelations. An earlier Forge run (late May) scored Platner 53.9 to Collins 47.8 — but that predates the disclosures, and public polling has tightened since. Textbook structure-versus-candidate-quality — and the structure still favors us, if Platner doesn’t crack. Unless there are far worse revelations, I’m all in on Platner. We simply HAVE to win Maine to take the Senate. No equivocating. No misplaced purity checks. Maine knows what they need and want.

4) Alaska — the sleeper, and the best leverage on the map. Mary Peltola is the only Democrat to win statewide in Alaska since 2008. She raised a record $8.9M in Q1 — quadrupling GOP incumbent Dan Sullivan — and recent polling favors her. An earlier Forge run scored it a dead heat (Peltola 55.2, Sullivan 55.9). Three things make Alaska special for a donor: it’s a cheap media state, it runs on ranked-choice voting (where field and second-choices decide). The Dem backed Senate Majority PAC has reserved $10M in Alaska TV; SLF is countering with $15M. Sullivan’s liabilities are his own (”Yes-Man Dan”); Peltola’s hurdle is the state’s federal-level GOP lean and Murkowski’s endorsement of the incumbent.

5) Iowa — the wave detector. Sen. Joni Ernst retired after the “we are all going to die” Medicaid town-hall fallout, leaving an open seat. Josh Turek — a two-time Paralympic gold medalist in the Tom Harkin prairie-populist tradition — won the Democratic primary with 62.6% and faces Trump-endorsed Rep. Ashley Hinson. A June 3–4 poll had it tied, 46–46. Trump won Iowa by 13 and markets price a ~60% Republican hold — but Iowa is exactly where the special-election signal is loudest: Democrats recently flipped a state-Senate seat in a district Trump had carried by 25 points. If that energy scales, Hinson isn’t safe and the wave has crested. If it doesn’t, Iowa marks the ceiling.

6) Texas — the high-upside reach. James Talarico, a state representative and seminarian, leads Ken Paxton by 0 to 3 points — and Paxton is radioactive (net favorability ~-19, with roughly 30% of Cornyn’s primary voters defecting over corruption). An earlier Forge Campaign Pack scored it Talarico 51.1 / Paxton 48.5, with the market at 54.9% Paxton — about a 1-in-4 path. Talarico has a $7.5M cash-on-hand advantage and wins independents by double digits. Texas’s structural lean and turnout still make him the underdog. A Talarico win doesn’t just add a seat — it confirms the wave already broke.

The scenario: walking the +4

A clean, plausible path to 51:

• Hold the firewall — Ossoff (GA), the Democrat, likely Sayed, (MI).

• Hold the favorable defenses — Pappas (NH), the Democrat (MN).

• Flip North Carolina — Cooper converts his lead.

• Flip Ohio — Brown’s crossover appeal holds.

• Flip Maine — Platner survives the candidate-quality fight in a Harris +7 state.

• Flip Alaska — Peltola wins the ranked-choice endgame.

That’s +4. Texas and Iowa are insurance and upside — either one cushions a loss elsewhere.

What kills it: a single defensive failure in Georgia or Michigan, or a Platner collapse in Maine that Texas or Iowa doesn’t backfill. The probability that all six break the right way is exactly why the majority sits near 45%, not 60%. Every race is a coin flip you can win; four-plus-two coin flips landing the same night is the hard part.

One pattern worth naming before we get to the money: every Democrat currently leading or overperforming — Cooper, Brown, Peltola, and even Talarico in Texas — is running on cost-of-living and economic populism. The ones who are struggling drifted onto culture-war terrain. The wave is real, but it rewards a specific message.

Where the money actually goes

Not every winnable race is an efficient place to spend. Forge separates probability of winning from Leverage — where the marginal dollar still changes the outcome. The likeliest flip can be the least efficient check you write.

In a nationalized race, broadcast spending cancels — both sides saturate the airwaves by Labor Day. Turnout operations don’t. The differentiating dollar is a field dollar.

Tier 1 — Hold the firewall (highest stakes). Georgia (Ossoff) and Michigan (open). These decide whether there’s a path at all. Georgia especially — budget for a possible December runoff.

Tier 2 — Highest marginal leverage on offense. Alaska (Peltola), Maine (Platner), Ohio (Brown). True toss-ups in cheaper media markets where field and turnout decide. Alaska is the single best leverage point on the board: small state, ranked-choice endgame, a candidate already outraising the incumbent, and ground game that can’t be cancelled.

Tier 3 — Must-win, lower marginal leverage. North Carolina (Cooper). The likeliest flip — but precisely because Cooper is already ahead, already flush ($13.8M/quarter), and about to be buried under $200M+ in combined spending, your marginal dollar moves less here than in Alaska or Maine.

North Carolina is the likeliest flip and the lowest-leverage check you can write. Alaska is the opposite. Size your giving to that inversion.

Tier 4 — High-upside reaches; fund turnout, not TV. Texas (Talarico), Iowa (Turek). Structural underdogs where a win means the wave already crested. Give to turnout and persuasion of soft-Republican defectors — not broadcast, which saturates and cancels.

Bottom line

The environment is the best Democrats have had since 2018, and the map is genuinely winnable. Four seats and a firewall. That’s the whole game.

Why it has to be both — and why it has to be big

Everything above is about the Senate. Let me close with why the Senate is the part that actually matters — and why winning back the House, as likely as that now looks, won’t be enough on its own.

A House majority is a brake. It can stop bad bills, force votes, open investigations, and make the next two years a grind. What it cannot do is undo the machinery already being built. It can’t confirm or block a single judge. It can’t touch a Cabinet seat or an agency head. It can’t fill or hold a Supreme Court vacancy. All of that — the lifetime appointments, the remaking of the executive branch, the damage that outlasts any one president — runs entirely through the Senate. A Democratic House stalls the present; only a Democratic Senate can stop the courts and the personnel that will define the next thirty years. One chamber slows the bleeding. It takes both to start the repair.

And it can’t be close. A one-seat loss is the kind a party explains away — better turnout next time, a few unlucky states, nothing fundamental. Parties don’t reform after squeakers; they reform after shellackings, when the defeat is too big and too broad to blame on anyone but themselves. Only a decisive loss shifts the incentives inside the GOP away from the people who built this and toward the people who might rebuild it.

And to be clear: I am not rooting for one-party rule. A healthy democracy needs two parties that both believe in it — in free elections, in the peaceful transfer of power, in governing over performing. The old Grand Old Party, for all our disagreements, played that role; it provided the balance a republic requires. This incarnation has no corollary to it.

I don’t know what will rise out of the Trump-MAGA wreckage. I genuinely don’t. But I know this: nothing better emerges until this version of the Republican party loses — and loses BIG enough that denial is off the table. A narrow win leaves the fever unbroken. Only a landslide lets sanity, and a real opposition, find their way back.

So it isn’t the House or the Senate. It’s both, decisively. That is how you change a trajectory. That is how you take a country back.

I’ve been hoping for this outcome for six months. Tuesday night, Texas Republicans delivered.

Ken Paxton — indicted, impeached, outspent by his own party — beat four-term incumbent John Cornyn by 27.6 points in the GOP Senate runoff. Trump endorsed him. Republicans nominated him anyway. On purpose.

With Cornyn on the ballot, Texas was a long shot. With Paxton, it’s a live race — and it makes the question above real. If Texans elect a Republican purely because Trump told them to, despite every red flag, the movement is durable beyond the man. If Paxton’s record breaks through, the coattails belong to Trump alone.

Here’s where The Forge sees this race actually standing.

THE FORGE VERDICT—Support Talarico for Senate

The donor question, answered:

Forge Score™: Talarico 51.1 · Paxton 48.5 (paired focal, near-tied)

Texas Senate 2026 is genuinely competitive (Ratings consensus: Lean R (Cook explicitly shifted Likely R → Lean R when Paxton won the runoff);

More than markets currently price (Market price: R 54.9% (volume-weighted across $3.4M combined liquidity);

Less than public polls suggest(Public polls (RCP aggregate) Talarico +0.5.

Marginal dollars matter here meaningfully more than in races with similar Forge Scores but more saturated channels—Leverage Index™: 95.6 / 94.0 (HIGH band — marginal dollars move outcomes) .

Talarico’s path requires three things to go right simultaneously:

1. Democratic base mobilization at March primary turnout levels — 2.07M votes, roughly double the 2024 D Senate primary

2. A Cornyn-voter crossover of 15–20%, concentrated in Travis, Dallas, Brazos, McLennan, and Williamson counties

3. A national environment that does not collapse for Republicans before November

What would change this call. Cook moving to Toss-Up would raise it materially. A second Paxton indictment or new whistleblower judgment would raise it further. A measurable Hispanic Republican shift back to Cornyn-style Republicans would lower it. Polling closing to a tie at the LV-screen level in late summer would raise it; polling confirming the SOS handicap is still operative would lower it.

By The Forge™ · Forge Decision Systems · Donor Brief · May 27, 2026

The Hook

Texas hasn’t elected a Democrat statewide since 1994. That’s the floor. The ceiling just got higher.

What happened on May 26:

• Ken Paxton beat four-term incumbent John Cornyn in the GOP primary runoff

• Paxton 63.8% / Cornyn 36.2% — a 27.6-point margin

• 1,387,674 votes cast

• Trump’s endorsement was the decisive variable

Cook Political Report had announced in advance, publicly, on the record that it would shift the race from Likely R to Lean R if Paxton won.

He did. They did.

Why it matters. That conditional rating shift is the cleanest external signal of what just happened to the expected value of this race for a Democratic donor — and it came from the most respected handicapper in American politics, moving the race one full category because of who Republicans nominated, not who Democrats nominated.

Who Is James Talarico

Snapshot:

• 36 years old, state representative for north Austin (HD-50)

• 8th-generation Texan, born in Round Rock

• Former sixth-grade Language Arts teacher, Westside San Antonio

• UT Austin → Harvard MEd → Austin Presbyterian Theological Seminary (MDiv 2025)

• Type 1 diabetes; diagnosed after collapsing on a 25-mile campaign walk in 2018

• 1.5M+ Instagram, 1.3M+ TikTok

What he’s actually done. Three terms in the Texas House since 2018. Passed the first cap on pre-K class sizes in Texas, equipped public schools with Narcan to fight teen overdoses, gave incarcerated minors a pathway to earn high school diplomas. Texas Monthly named him one of its Top Ten Best Legislators.

The viral profile is the campaign. His videos opposing the Republican Ten Commandments classroom mandate routinely earn millions of views. His three-hour appearance on Joe Rogan’s podcast last summer — in which Rogan, who endorsed Trump in 2024, suggested Talarico should run for president — gives him an audience layer that bypasses traditional political-media gatekeeping.

The primary that proved he can mobilize:

• March 3, 2026 D primary

• Talarico 53.2% / Crockett 45.5% / Hassan 1.3%

• Total turnout: 2,073,750 votes — roughly double the 2024 D Senate primary

Talarico’s primary effectively matched the Republican March 3 primary in total ballots cast. In a state where Democratic primaries historically run at half of Republican primary turnout, that is not a normal result for Texas.

The Numbers

Start with the structural baseline.

The surface fact: Texas has hardened Republican since 2018. The 2024 R+13.8 was the largest GOP presidential margin in Texas since 2004.

Underneath the surface, the picture is more mixed. Democrats have made measurable gains among white Texas voters during the Trump era — roughly comparable to the white-voter movement that turned Georgia competitive — but those gains have been masked by an even larger Hispanic R-shift through 2024. If the Hispanic R-shift is now reversing, as national polling suggests, the underlying trajectory is more favorable to Talarico than Visual 1 alone implies.

That’s the prior. Here’s what the live 2026 signals say:

Signal source: Public polls (RCP aggregate)Talarico +0.5 Forge Score™ (kernel)Talarico +2.6 Cook rating”Lean R” (≈ R+5)SOS-handicapped polls Paxton +3.7 Markets (VW Pr_R 54.9%)Paxton +9.8

What polls say

Five public head-to-head polls in 2026. Talarico leads four. The fifth is tied.

• TSU 5/6: Paxton 45 / Talarico 45

• TPOR 4/20: Talarico +5

• UT/TPP 4/20: Talarico +8

• Impact (D-sponsored) 3/17: Talarico +1

• PPP (D-sponsored) 3/5: Talarico +2

Talarico has led every public general-election poll since January 2026. In 2018, Beto O’Rourke never led a single RealClearPolitics-tracked poll against Ted Cruz. The polling picture this cycle is materially different from the closest prior Texas race.

What markets say

• Polymarket: GOP at 54%

• Kalshi: Paxton at 55%

• Combined volume: $3,398,137

• Volume-weighted Pr(R) = 54.9% — roughly a 10-point R margin

Why they disagree

Texas Senate polling has under-counted the Republican by an average of 4.2 points every cycle since 2018. Apply that to the RCP aggregate and the handicapped baseline becomes Paxton +3.7. Apply it to TSU and you get Paxton +4.2. The numbers converge with the market price.

Risk to monitor: if Texas pollsters have improved their likely-voter screens this cycle, Talarico’s polling lead could be closer to truth than the historical handicap suggests. Watch for methodology updates from major TX pollsters in late summer.

What the cash tells us

Talarico (D)Paxton (R)GapCash raised $40.28M/$7.60M 5.3× Cash on hand $9.86M/$2.34M 4.2× Small-dollar %57.6%/16.4%—YouTube views/video 21,426/1,059 20×

Four independent base-mobilization signals point in the same direction. The cash gap is unprecedented for a Texas Democrat against a sitting Republican statewide.

The Opponent

Three things to know about Ken Paxton:

1. He has won statewide three times. In 2022 against Rochelle Garza he won the Attorney General race by 11 points — 54 to 43. That is his statewide ceiling in a normal environment, and it is a non-trivial number.

2. He owns the Republican base. Trump endorsed him, and he won 245 of 254 counties in the runoff. He is not a damaged candidate among Republican primary voters; he is their preferred candidate.

3. But the same county returns show where he’s exposed.

Where the Cornyn votes lived

• Travis (Austin): Cornyn won by 5

• Dallas: Paxton +1.1, in a county every other statewide R wins by 20-plus

• Brazos (Texas A&M), McLennan (Waco), Williamson (Talarico’s home): Paxton well below his 27.6-point average

These five counties together produced approximately 175,000 Cornyn votes — the pool from which Talarico’s general-election crossover will primarily come.

The vulnerabilities — documented public record

• 2015: indicted on first-degree felony securities fraud in Texas (Servergy Inc.); the case stretched nine years; settled 2024 with ~$300,000 in restitution to avoid trial

• 2020: eight senior deputies went to the FBI accusing him of bribery and abuse of office to help campaign donor Nate Paul, who employed a woman with whom Paxton has acknowledged an extramarital affair

• 2023: Republican-led Texas House impeached him 121-23 on 20 articles including bribery and abuse of public trust; the Republican-majority Senate acquitted with only 2 GOP conviction votes; wife Angela Paxton recused

• April 2025: Texas court awarded the whistleblowers $6.67 million

• April 2025: DOJ declined to prosecute

The attack library is pre-paid

The pro-Cornyn super PAC Texans for a Conservative Majority spent $26,884,964 on 54 separate Independent Expenditures(IEs) against Paxton over the last 180 days. Every line of attack has been test-marketed on Republican voters, paid for by Republican donors, and is now sunk cost from the PAC’s perspective. The ad creative, message research, and oppo are persistent assets Talarico’s campaign and aligned outside groups inherit at zero cost.

Paxton’s likely 90-day playbook

• Nationalize the race, tie Talarico to unpopular national Democrats

• Run hard on immigration and border — his only R-owned issue at 24% voter salience

• Frame Talarico’s seminary background as performative, his videos as content creation rather than legislating

• Deploy two specific April 2021 floor speeches: “six biological sexes” and “God is nonbinary” — NRSC launched an AI-generated digital ad using both within days of the March 3 primary

• Heavy South Texas Hispanic outreach

• DMAs: Houston, Dallas, RGV — concentrated, not statewide saturation

What Talarico should pre-empt

• Define corruption in this race before Paxton can recast it as a partisan attack

• Lean into Texas identity (8th-generation, Westside SA classroom)

• Engage Hispanic voters on cost-of-living — 37% salience, D-favored — not immigration

• Avoid extended culture-war engagement; Texans rank affordability, education, and immigration above it

The Hispanic electorate — read it carefully

The runoff numbers show Hispanic Republican primary voters consolidated cleanly for Paxton: Hidalgo +37, Bexar +21. But primary voters are a self-selected slice of the electorate.

The broader picture, from NYT/Siena national fielded mid-May 2026:

• Hispanic registered voters on generic ballot: D 54 / R 24 (D+30)

• Hispanic Trump approval: 20 approve / 71 disapprove (net −51)

• Caveat: Hispanic “almost certain to vote” intensity at 30% vs 43% for the registered-voter average

The honest read: Talarico’s South Texas outreach faces a hardened Trump-aligned electorate among the Hispanic Republicans who already voted in the primary, and a much more contested electorate among the Hispanic registered voters who did not. The directional signal runs against the structural prior.

The Bottom Line

Three ways to think about this race:

1. The probability read. Talarico can win. The base case still favors Paxton because Texas is Texas. Markets price the GOP at 55% probability, Cook says Lean R, and the kernel splits the difference at near-tied. None of these describes a sure thing in either direction. A donor who needs near-certainty about outcomes should look at safer pickup targets.

2. The marginal-dollar read. Leverage Index 95.6 / 94.0, HIGH band. This race has unsaturated channels, time on the clock, and a public-record opponent whose attack library is already paid for. Dollars spent here from June through September will move outcomes more per dollar than dollars spent in the closing six weeks. That is the operative finding for any donor optimizing on impact rather than probability of flip.

3. The portfolio read. Democrats need a net of four Senate seats from the current 53-to-47 Republican majority. Of the realistic pickup targets — Maine, NC, Iowa, Alaska, Ohio, Texas — Texas has the lowest baseline win probability and historically the highest cost per persuaded vote. Maine, NC, and Iowa rank higher on probability-of-flip math. But Texas’s demographic mix — roughly 40% Hispanic, the most diverse of the pickup targets — means a national Hispanic D-swing amplifies more in TX than in OH or AK. National donor capital may still flow to higher-probability races, which is precisely where the Leverage Index reading finds its largest practical implication for an individual donor.

If your strategy is to maximize the probability of Senate control, your top dollar belongs in Maine, NC, and Iowa. If your strategy includes supporting genuinely competitive races where dollars are not yet wasted, Texas earns a serious allocation this cycle in a way it has not in twelve years.

That is the most honest answer the data supports.

Support Talarico for Senate—Donate Here

Methodology Note

Forge Score™ integrates market probabilities, polling, finance, issue alignment, social attention, and macro overlay into a single competitiveness measure from a focal candidate’s perspective. Above 55, favored; below 45, behind; 45–55, competitive. Paired focal scores summing near 100 indicate a tight race.

Leverage Index™ measures the marginal impact of additional dollars on race outcomes, integrating channel saturation, time to election, race competitiveness, and operational capacity. Above 70, high marginal impact; 45–70, moderate; 25–45, low.

GEV Score™ measures general-election viability for primary candidates. Does not apply to general elections and is not reported in this brief.

These are quantitative summaries of integrated evidence as of publication. They are not predictions.

Disclaimer & Liability Waiver

Produced with LLM-assisted analysis using The Forge™ decision support system. Forge Decision Systems makes no representation or warranty regarding the accuracy, completeness, or timeliness of the information. Nothing here constitutes investment, legal, or political advice. The marginal-dollar leverage analysis is informational and does not constitute a recommendation to contribute to any specific campaign. All data is time-bounded to publication; race conditions, candidate positions, market prices, polling, and external events can change materially. Forge Decision Systems accepts no liability for decisions made based on this brief. Republication permitted with attribution.

© 2026 Forge Decision Systems. All rights reserved. The Forge™, Forge Score™, GEV Score™, and Leverage Index™ are trademarks of Forge Decision Systems.

The bottom line up front

If you give to one Senate race in 2026, give to this one: Graham Platner for Senate

• Graham Platner leads Susan Collins 53.9 to 47.8 on the Forge Score. (The Forge Score™ is a proprietary, comparative read of where a race actually stands — built from polling, prediction markets, finance, social-attention signals, issue alignment, and partisan structure. It is not a win probability.)

• Volume-weighted prediction markets price Democratic victory at 68.9%.

• The polling has been stable for six months through a brutal opposition cycle — a resurfaced Reddit history from 2013, a tattoo controversy involving a Nazi SS symbol, a recent ad mocking a wounded soldier. Platner has absorbed every attack. The median public poll has not moved.

• Janet Mills withdrew the Democratic primary on April 30. The Democratic coalition consolidated around Platner inside 72 hours, ending the establishment-vs-insurgent split that had kept money on the sidelines.

• An anti-Platner super PAC funded by Stephen Schwarzman, Paul Singer, James Davis of New Balance, and the Reyes brothers is sitting on $8.7M of unspent ammunition to deploy through October. That money is the central reason this race is not already over.

• Prediction markets and expert ratings disagree by 27 percentage points on Platner’s chances of winning. The experts are anchored on Collins’s history. The markets are reading the macro environment. This is the cleanest mispricing on the Senate cycle.

• Maine is the only Republican-held seat in a Harris-2024 state. If Collins loses, Democrats almost certainly win the Senate majority. If Collins wins by five-plus, Republicans hold. That’s the actual stakes.

Forge’s recommendation: this is the seat where marginal Democratic dollars change the outcome of national control of the United States Senate. Do not assume someone else is funding it. They aren’t yet.

The race that decides the Senate

Start with the map.

After the 2024 elections, Republicans hold the Senate 53 to 47 — counting the two independents who caucus with Democrats. To take the majority, Democrats need a net gain of four seats in 2026.

There are 22 Republican-held seats up. Twenty of them are in states Trump won by double digits in 2024. Arguably only a handful are in play this cycle: Texas, Alaska, Ohio, North Carolina, and Maine. Maine is the only one of these in a state Kamala Harris carried. The Cook PVI is D+2. The 2024 presidential margin in Maine was Harris +6.7 statewide. Maine’s structural baseline is more Democratic than any other defendable Republican seat in the country.

Maine is the cycle's wave-size indicator. Democrats need a net gain of four Senate seats to take the majority. If Platner doesn't win Maine, the path to four anywhere else collapses. The other competitive races — Georgia, Michigan, North Carolina, Alaska, Ohio — are downstream of what happens here. Maine isn't just a pickup. It's the door the rest of the map can't walk through.

Graham Platner: the candidate everyone got wrong

When Platner announced his candidacy on August 19, 2025, the consensus in Washington was that he was a novelty — a viral campaign launch from a forty-one-year-old oyster farmer who had served four combat tours in Iraq and Afghanistan, who chaired the Sullivan, Maine planning board (population 1,300), and who had never held elected office above the level of harbormaster. The DSCC was backing Janet Mills, the two-term governor. Chuck Schumer was backing Janet Mills. The donor class assumed Mills had it locked.

Then the polling came in. By February 2026, the University of New Hampshire had Platner up eleven points head-to-head against Collins. Maine People’s Resource Center had him up nine. Emerson had him up seven. The Mills-vs-Collins polls were inside the margin of error — usually a tie. By April, every aggregator that tracks the Maine race had Platner ahead by five to seven points and Mills two to three points behind Platner in the Democratic primary.

On April 30, Mills suspended her campaign, citing inability to raise money against Platner’s grassroots fundraising operation. Within 72 hours, Schumer endorsed Platner.

Here is what makes Platner a serious candidate, not a novelty:

He has raised $11.96 million cycle-to-date, with 64.5% of his receipts coming in unitemized small-dollar contributions under $200. That’s genuine grassroots. He’s spent $9.23M of it already — heavy, but he’s been air-warring since fall 2025, which is the reason he survived the October Reddit/tattoo cycle without his polling collapsing.

He has endorsements from Bernie Sanders, Elizabeth Warren, Martin Heinrich, Ruben Gallego, and now Chuck Schumer — a coalition that combines progressive and establishment Democratic infrastructure on top of national labor (UAW, IFPTE, National Nurses United) and the Maine state federation (Maine People’s Alliance, Maine State Nurses Association, APWU, IBEW). The Working Families Party and the Progressive Change Campaign Committee are mobilizing for him. This is not a fringe campaign anymore.

He served four combat tours — three in Iraq with the Marines (3rd Battalion 8th Marines, machinegun section leader at the Government Center in Ramadi), one in Afghanistan with the Maryland National Guard. He came home, attended George Washington University on the GI Bill, worked as a State Department security contractor in 2018, returned to Sullivan, Maine, took over the Waukeag Neck Oyster Company in 2020, and slowly built a local political profile through the planning board and the harbormaster’s office. The biography is real, and on television it reads as exactly what it is.

He is not a careful candidate. He calls to abolish ICE. He says corporations should pay climate reparations. He called Israeli military operations in Gaza “genocide.” He apologized for past misogynist and racist Reddit posts; he covered the controversial tattoo; he refused to apologize for the wounded-soldier ad. Some of his positions are well to the left of the median Maine voter. The campaign’s bet is that authenticity beats positioning in a year when consumer sentiment is at 44.8 — the lowest in recorded history dating to 1952 — and ICE has a net favorability of −16. The polling so far says the bet is right.

If you want to understand what makes this guy tick and why he’s a compelling candidate inside of Maine, watch this interview:

The polling: six months of stability nobody is talking about

The most underappreciated fact in this race is that the polling has not moved in six months despite a sustained opposition campaign.

Thirteen head-to-head polls between Platner and Collins from December 2025 through May 2026. The shaded period reflects the post-Mills-withdrawal consolidation. Aggregator consensus has held at Platner +5.5 to +6.7 since February.

Pan Atlantic Research, an independent Maine pollster with a strong local track record, had Platner up seven on May 18. UNH had him up eleven in February. Emerson had him up seven in March. Maine People’s Resource Center has had him up consistently, including a nine-point lead on March 31. Through national coverage of the Reddit posts and the Panzer tattoo, through the wounded-soldier ad refusal, through a Mills-vs-Platner primary that should have softened him for the general — Platner’s lead has held inside a six-to-seven point band.

The two outliers — both ties — were a December Workbench Strategy poll (a Democratic internal) at 50-50 and a January Fabrizio Lee poll (Trump’s pollster, internal Republican) at Collins 45 to Platner 44. Every independent public poll since February has shown Platner ahead by five or more.

What this tells us: the opposition file is already priced in. Pine Tree Results PAC is sitting on $8.7M to keep telling these stories on paid television in August and September. The question is whether re-amplification erodes the lead in ways that earned media has not.

The market signal nobody is reading

Here is what makes this race quantitatively unusual.

Volume-weighted Polymarket and Kalshi consensus puts Democratic win probability at 68.9%. Race to the WH’s expert-rating model puts it at 42%. The 26.9-point gap is unusually wide for this stage of a Senate cycle.

Polymarket has Democrat at 73 cents to win the Maine seat, on $283,632 of 24-hour volume. Kalshi has Democratic party at 67 cents, on $639,513 of 24-hour volume. Volume-weighted, the market consensus is 68.9% Democratic win probability.

Race to the WH, the most quantitatively rigorous of the four major expert rating shops, currently models the Democratic win probability at 42%. Cook Political Report calls the race “Toss-Up.” Sabato’s Crystal Ball calls it “Lean R.” Inside Elections calls it “Tilt R.”

That 27-point gap is the trade.

The experts are anchored on Collins’s history. She has won five Senate races since 1996. In 2020 she beat Sara Gideon by R+8.68 in the final round of ranked-choice voting despite trailing in late-cycle polling by five points. Collins reliably polls worse than she performs. The expert models have learned this and built it into their priors.

The markets are reading the macro environment. Consumer sentiment at 44.8(the worst in recorded history). CPI is still 3.78% year-over-year. Gas is $4.51 a gallon. The generic congressional ballot has Democrats up eight among registered voters. ICE has a −16 institutional rating after the Pretti shooting in Minneapolis and Maine’s own “Operation Catch of the Day” surge in January. The macro environment is the worst for an incumbent’s party since 2010.

Pick whichever side of the trade you find more credible. The Forge view is that the markets are pricing the cycle environment correctly and the experts are over-anchored on Collins’s individual political skill. A 27-point gap is not normal at this stage of a Senate cycle. Somebody is going to be wrong by Labor Day, and the volume on Kalshi suggests the market participants who have actual money on the line are already moving.

Where the money is — and what the donor class is missing

Platner has spent $9.23M against Collins’s $3.04M. The Platner campaign is 34% digital advertising and 29% TV. Collins is 24% digital, 23% mail, 16% fundraising — a balanced incumbent posture sitting on a $10M war chest for the fall.

Two finance facts matter for donors.

First, Platner’s small-dollar engine is real. 64.5% of his $11.96M in receipts came in unitemized contributions under $200. Compare this to Collins’s 13.4% small-dollar ratio. Platner’s donor base is mostly individual people who have given less than the FEC itemization threshold. Collins’s donor base is itemized large-dollar Republicans — Stanley Druckenmiller, David Herro of Oakmark International, KKR partners, Blackstone partners. The two campaigns are structured differently because they’re funded differently.

Second, the outside-money asymmetry is the central financial fact of the race.

In the last 90 days, Pine Tree Results PAC alone has spent $3,999,000 against Platner. NNU has spent $50,606 for him. Lead Maine Committee plus SLF PAC plus ClearPath plus Stronger Maine plus Red Senate have spent $790,240 for Collins, with $1,271 against her. Pine Tree Results holds $8.7M of unspent capacity.

Pine Tree Results PAC is an independent-expenditure-only super PAC organized to defeat Platner. The donor list runs through the hedge-fund and private-equity Republican right: Schwarzman ($2M), Singer ($1M), Davis ($1M), Bacon ($500K), the Reyes brothers ($1M+), and Malone ($500K). $12.7M raised, $4M deployed, $8.7M in reserve for the summer-fall air war.

There is no equivalent on the Democratic side. National Nurses United has spent small money — $50,606 in 90 days. The DSCC and the WFP are mobilizing post-Mills-withdrawal, but the structural asymmetry is real: outside money is outspending Platner roughly 80 to 1 in the supportive direction.

This is the gravity in the race, and it is why marginal Democratic dollars matter here in a way they do not in races where the giving infrastructure is already saturated. Platner is competitive despite this asymmetry. The question for donors is whether the asymmetry persists through November or gets closed.

Susan Collins: the candidate Democrats keep underestimating

The serious case for Collins, and the case Democratic donors should understand:

She has won five Senate races in a state that has not voted for a Republican president since 1988. She has built a thirty-year brand of “independence” — voting against Pete Hegseth and Kash Patel, voting to convict Trump in the 2021 impeachment, voting against the final 2025 Trump tax-and-spending bill, voting against the SAVE Act on procedure before voting for it on final passage (an asterisk on the “independence” frame). She chairs Senate Appropriations — the first Republican woman to do so — and her office claims $1.5 billion in directed Maine spending over the past five years across roughly 700 local projects.

She has $10 million cash on hand — the largest war chest of any incumbent on the 2026 Senate map — and has burned through only 29% of receipts. She’s sitting on a fortune. The Pine Tree Results PAC pacing strategy is almost certainly to deploy $8.7M over a 90-day window from Labor Day through Election Day, when persuadable Maine voters are paying attention.

Where she is vulnerable:

Kavanaugh. She voted to confirm him in 2018; she said afterward that he “misled” her on Roe precedent; Dobbs vacated Roe in 2022. Post-Dobbs, this is the single heaviest vote in her record. Maine voters who supported Roe and who have been waiting to retire Collins since 2018 are the swing bloc Platner needs.

The Trump cabinet. She voted to confirm RFK Jr. to HHS, McMahon to Education, and Vought to OMB — among the most controversial Trump confirmation votes of 2025. The “independence” frame works against Pete Hegseth; it doesn’t work for three out of three controversial Cabinet votes.

Operation Catch of the Day. In January 2026, federal immigration agents conducted a large-scale enforcement surge in Lewiston and Portland that arrested hundreds of Mainers, including residents without criminal records. Collins called DHS Secretary Noem and helped end the operation. She is now campaigning on having stopped it. But it happened in her state, on her watch, executed by an administration she has voted to enable on most issues. In a state where ICE has a strongly negative reputation, the “I stopped it” frame requires voters to accept a sequence of events that puts her on both sides of the issue.

The macro environment. Consumer sentiment, gas, generic ballot D+8. This is what Sabato calls the “GOP’s hardest defensive draw of the cycle.”

What she will do over the next ninety days, working from her perspective: lean hard on the Appropriations frame for earned media through June and July. Hold Pine Tree Results PAC fire until late August to extend the runway. Try to force one re-apology cycle on Platner in August by surfacing one new piece of opposition research. Run “independence” framing in the Portland-Auburn media market through September; pivot to ME-2 Trump-aligned messaging in the final thirty days. Bank on her historical five-point polling-vs-actual overperformance reversing itself again.

That is a real strategy. The Forge view is that Platner survives it if and only if the median public poll does not drop below +3 by Labor Day, the Pine Tree Results summer barrage does not introduce material new opposition, and the immigration story remains a live news cycle through October. All three conditions are within reach.

Three things should give a thoughtful donor pause.

The new-opposition risk. The current opposition file — Reddit posts, Panzer tattoo, wounded-soldier ad — has been in market for six months and the polling has held. The risk is not the existing file. The risk is something currently unknown that surfaces in July or August: a new resurfaced post, a new video, a new tattoo, a new story. Probability is moderate. Consequence is severe because Platner has already used his apology cycle in October 2025 on the Reddit posts. He cannot run the same playbook twice.

The ICE-resolution scenario. If the Trump administration de-escalates ICE operations in response to the Pretti shooting and the −16 institutional rating, the highest-salience anti-incumbent issue in southern Maine loses its valence. Collins’s “I stopped the surge” frame becomes credible. Probability is low. Consequence is material.

The slow-bleed scenario. Pine Tree Results deploys $5M to $7M of paid television advertising between July and August, the median poll drifts from +6 to +2 by Labor Day, and the final two months of the campaign turn into a recovery operation rather than a closing argument. This is the most likely loss path, and it is the reason marginal Democratic dollars are actually marginal: they buy response, not just amplification.

The Forge read

Just look at the picture.

The incumbent is a serious candidate with $10M in the bank and an $8.7M reserve coming for the summer. She’s also not, by the numbers, currently winning.

The marginal Democratic dollar in this race doesn’t defend a polling lead. It buys the response to a summer of attack advertising that hasn’t started.

If you’re giving to one Senate race in 2026, this is the one. Graham Platner for Senate

Graham Platner is the Democratic candidate for U.S. Senate from Maine. His campaign is at grahamforsenate.com. The Democratic primary is June 9, 2026. The general election is November 3, 2026.

This analysis was produced by the Forge™ Decision Support System using polling from thirteen independent public polls aggregated through Wikipedia, prediction market data from Polymarket and Kalshi (operator-captured 2026-05-26), FEC campaign finance data for both candidates, BLS labor statistics, Census ACS demographics, GDELT and Wikimedia social-attention signals, and contemporaneous reporting from Britannica, AP, NPR, CNN, CNBC, PBS, the Maine Morning Star, the Bangor Daily News, the Maine Monitor, Fox News, and the Washington Examiner. Current to 2026-05-26.

The three-way Democratic fight for Peters’ open seat is the cleanest case this cycle of a race where the best primary candidate and the best general-election candidate may not be the same person.

(For context, the Forge and GEV scores are derived from a proprietary multivariate decision support system(DSS) i’ve built and recently updated using Anthropic’s Claude Pro)

EL-SAYED LEADS THE PRIMARY AND TRAILS THE GENERAL. · STEVENS TRAILS THE PRIMARY AND LEADS THE GENERAL. · MCMORROW IS IN BETWEEN ON BOTH.

Gary Peters is retiring. Mike Rogers is sitting on $4.2 million with a $45 million Republican super-PAC commitment waiting for him in August — more money than they’ve promised any other pickup target this cycle. And three Democrats are in a genuine three-way fight for the nomination with a hundred days to go.

This is one of four Senate races Cook still calls a pure toss-up. It is also one of only two seats Democrats currently hold in a state Trump won in 2024. Losing it makes a Senate majority almost arithmetically impossible.

It’s also the cleanest example I’ve seen this cycle of a race where the best primary candidate and the best general-election candidate may not be the same person. Here’s what the model says

Three numbers, three questions. The Forge Score (0–100) ranks where each candidate stands today in the primary. Fifty is average; above is stronger, below is weaker. It’s a relative position, not a win probability.

GEV — General Election Viability — is new this cycle. It answers a different question: if this candidate wins the primary, what is their position against Rogers in November? It weights head-to-head polling, state fundamentals, margin among independents, and documented electability factors.

The Leverage Index tells you whether your marginal dollar still moves the outcome. High means live race and open advertising channels with time remaining. Low means the race is decided or the airwaves are full.

The headline

The primary is ranking the candidates in reverse of the general election.

Abdul El-Sayed leads the most recent Emerson poll — tied with Mallory McMorrow at 24% to Haley Stevens’ 13% — and is now the narrow favorite on Polymarket at 48%. But he is the only candidate essentially tied or behind Rogers in head-to-head polling, and he leads Rogers by only 3 points among independents. Compare that to McMorrow’s +12 and Stevens’ +13. In a state Trump won by 1.4 points, independents are the decisive bloc. That is why his GEV lands in the WEAK band.

Stevens is in the opposite situation. The Prediction markets price her at a 6–11% chance of winning the primary; her polling has slid from 34% a year ago to 13% in the latest Emerson. But across seven head-to-head polls against Rogers she averages +2.3 points, the strongest of any Democrat in the field. Her suburban-Oakland-County plus union-machine profile mirrors Elissa Slotkin, who won this seat by one-third of a point in 2024.

McMorrow is the middle case on both dimensions. Not the strongest primary position, not the strongest general. But she is the only candidate whose two scores roughly agree with each other, and she won Q1 fundraising decisively with $3 million raised, $3.69 million in the bank, and the healthiest burn rate in the field (41%, against Stevens’ 84% and El-Sayed’s 76%).

The donor question in this race: do you optimize for winning August 4, or for holding the seat in January 2027? The answers point to different candidates.

Abdul El-Sayed

The progressive surge with a November problem

He is running the most insurgent-progressive campaign in the field and it is working in the primary. The Piker rally controversy that dominated D.C. commentary three weeks ago moved his numbers up, not down — Data for Progress found 40% of primary voters approved of the rallies and only 30% disapproved. His base is real, his fundraising is disciplined ($596K in small-dollar donations last quarter), and his divergence from Senate Democratic leadership gives him a genuine lane.

The general-election math is where it breaks down. A $45 million Republican air war with the Piker footage, the Israel positioning, and the Medicare-for-All platform is exactly the kind of material the Slotkin-Ossoff playbook of the last two cycles was designed to avoid. That is what shows up in his +3 independents margin and the −1.3 point average head-to-head deficit against Rogers.

→ If your priority is holding the seat in January 2027: this is the weakest pick of the three.

Mallory McMorrow

The balanced profile—The Winning Combination

She is the only candidate whose primary position and general-election viability align. Tied for the lead in Emerson, leading on Kalshi (49%), trailing on Polymarket (45%) — markets are split. She beats Rogers by +12 among independents, essentially ties him in head-to-head averages, and her Q1 money makes her the most financially resilient candidate for a long August push.

Her vulnerabilities are real but narrow: the J Street endorsement and her recent shift on corporate donations open a “triangulation” line of attack that El-Sayed has been pressing. But the Forge kernel still rates her issue-alignment mid-band, and her own divergence from Senate leadership gives her the same progressive-flank credibility that El-Sayed claims.

→ $100–$500. The best composite pick if you want both primary-probability and November seat retention.

Haley Stevens

The candidate markets are pricing out of a race the model says she could win.

This is the uncomfortable one. The Forge says her primary position is slipping — 17% in Emerson January, 13% in Emerson April; $2 million raised against her opponents’ $2.3 million and $3 million. Markets have her at 6–11% to win the nomination. That is the honest read of her primary odds.

And yet. Seven head-to-head polls, average +2.3 points over Rogers. Plus-13 among independents. The “only one who beats Rogers in November” message her campaign has run for a year is exactly what the GEV score is measuring, and the model validates it. The Leverage Index of 82 — highest in the field — means that if you are going to fund her, now is when a dollar actually moves her numbers.

→ Hold and wait, I don’t see this dynamic changing but it given the distance to the primary she could become more viable.

What to watch in the next thirty days

Three moves could reshape this race meaningfully before Memorial Day.

A coordinated Third Way or DSCC-aligned $3 million-plus negative ad blitz against El-Sayed. Signaled in background reporting; not yet confirmed as active.

Stevens’ fundraising trajectory. If Q2 numbers (due mid-July) show her below $1.5 million raised, the “viable only in theory” narrative hardens and her GEV advantage stops mattering because she will not reach the general.

An external shock on the Israel file — Gaza, Lebanon, a domestic antisemitism incident — that sharpens or softens El-Sayed’s primary coalition and McMorrow’s J Street positioning.

I’M DONATING TO THE MCMORROW CAMPAIGN TODAY

METHODOLOGY. Forge Score is a 0–100 relative position, not a win probability. GEV (new in kernel v18.4.0) weights 50% head-to-head polling, 20% state fundamentals, 15% independents margin, 15% documented electability. Leverage Index measures marginal-dollar impact on race competitiveness. Based on 9 primary polls (2 A-grade), 17 head-to-head vs-Rogers polls, user-verified Polymarket + Kalshi markets ($850K combined volume on 2026-04-24), Q1 2026 FEC filings, and divergence evidence from 4+ independent outlets per candidate. Full Data Dossier available on request.

Last month, I wrote a post entitled “How I’m thinking About California’s Governor’s Race”. I wrote about the Three Pillars I’m using to evaluate this race: Values, Executive Capability, and Political Skills. I said Newsom scored two out of three and that wasn’t good enough when the missing pillar was the one that actually affects people’s lives. I said California had enough leading men and needed a builder.

Then the leading man imploded.

Eric Swalwell’s departure — swift, ugly, and in retrospect entirely predictable given what Washington had been whispering for years — blew the race open. Wednesday night was the first major debate of the post-Swalwell era. Six candidates, ninety minutes, and a set of moderators apparently more interested in streaming habits than in the structural crises facing the fifth-largest economy in the world.

Let me be direct about the moderators first: they failed California. We got lengthy exchanges on the gas tax, EV mileage fees, English proficiency tests for truck drivers, and a social media ban for under-16s. We got nothing substantive on water, the structural budget deficit, utility costs, or the education achievement crisis. The last half hour was bumped to digital streaming, cutting off half the audience. For a race this consequential, it was a disservice to voters.

Now the candidates.

Matt Mahan

Matt Mahan remains the candidate who most closely matches what I believe the state actually needs. His debate performance confirmed both his strengths and his problem. The strengths: he’s the only candidate on that stage currently running a major California government, with measurable results to show for it — homelessness down nearly a third in San Jose, thousands of housing units under construction, trust in local government up 40%. His opening — dismissing Steyer, Hilton, and Becerra in a single sentence — was tactically sharp. His closing — “We don’t need MAGA, but we also don’t need more of the same” — was the cleanest line of the night.

The problem is just as clear. He lacks the gravitas that a candidate making up for a very late start with almost no grassroots base desperately needs. His very late entry into the race, his Silicon Valley donor profile, and his relative unfamiliarity outside the Bay Area all require him to create a genuine breakout moment. Wednesday wasn’t it. He was studied, reasonable, and entirely unelectric. Pete Buttigieg without the national profile. In a race this compressed — mail ballots go out May 4 — studied and reasonable won’t move the needle enough.

He’ll need to do much better in the upcoming debates. If he can create that breakout moment, he proves he’s the right person. The race itself will be his audition.

Xavier Becerra

Xavier Becerra was my most significant reassessment going in — and my biggest disappointment coming out. On paper he has the strongest executive credential of anyone in the Democratic field: California AG, HHS Secretary, a budget larger than the state of California’s. The framework says executive capability is the pillar that matters most. So why does he leave me cold?

Because experience and willingness to use it are not the same thing. What I was hoping to hear from Becerra was someone willing to speak the hard truths — about what’s actually broken in Sacramento, about the entrenched interests blocking reform, about the structural dysfunction that has produced the nation’s highest poverty rate despite decades of Democratic governance. Instead I got a careful, experienced, cautious politician threading every answer to offend no one. He couldn’t bring himself to give Newsom lower than an A on effort and a vague grade on results. He talked about accountability without naming who he’d hold accountable. He has the experience of someone who has navigated the system. What California needs is someone willing to disrupt it. Those are different skill sets, and Wednesday night made clear which one Becerra has.

Tom Steyer

Tom Steyer remains the hardest to read. He can articulate the problems — income inequality, corporate capture of Sacramento, the gap between what government spends and what it delivers — with genuine conviction. He isn’t financially dependent on the interests he’s criticizing, which matters. And his track record deserves more credit than his debate performance suggested: he grew Farallon Capital from under $10 million to $36 billion in assets, built NextGen America into the largest youth voter engagement organization in American history with 750 staff and 15,000 volunteers, and co-founded a community bank that has financed 17,000 low-income housing units. These are real organizations he built from scratch.

But his debate performance raised the question I can’t answer: will he govern the way he campaigns, or is the progressive coalition-building just the price of admission? He’s pandered hard to every organized left interest group to get their endorsements. When it comes time to tell a teachers union or an environmental group that their preferred policy is blocking the housing California needs to build, will he do it? I have no idea, because he has no record in government. A billion dollars and a talent for building organizations doesn’t buy a governing track record — and his 2020 presidential campaign, which spent $345 million and won zero delegates, is a specific and relevant data point about executive judgment under pressure.

Katie Porter

Katie Porter was scripted, safe, and uninspiring. Her best moment — the data on who California’s homeless actually are — was real policy knowledge effectively deployed. But her closing was almost entirely negative, her housing answer was abstract, and she had no response to Becerra’s most effective line of the night: “Very rich to hear from someone who’s never had to actually run a government.” He’s right. Sacramento will overwhelm a candidate who can’t attract and retain the best executive talent, and nothing about Wednesday suggested Porter is ready to do that.

Steve Hilton

Steve Hilton is the candidate I most wish were running a different race. He is the most effective communicator on that stage — specific, structured, disciplined, and genuinely dangerous in a general election format. But he made a fundamental strategic error in this primary. He had a real opening: be the Republican who makes the affirmative case for what center-right governance can deliver in a state where it’s been absent for twenty years. California has elected Republicans statewide before — when those candidates represented competence and accountability, not culture war. Arnold Schwarzenegger. Pete Wilson. George Deukmejian. Hilton could have run in that tradition. Instead he embraced Trump’s endorsement in a state where Trump has 62% disapproval and called himself the president’s partner. That’s not a strategy for November. It’s a strategy for getting out of a Republican primary he was already winning. Too clever by half, and it will cost him. He should go back to Fox where he belongs.

The Bottom Line

The race is exactly where it was before the debate: fragmented, unresolved, and structurally dangerous for California Democrats. No one broke out. No one collapsed.

What I’m watching over the next three weeks: whether Mahan can manufacture the breakout moment he needs before May 4 ballots land, whether Becerra’s post-Swalwell poll surge translates into real consolidation, and whether Newsom finally reaches his “break the glass” moment — whatever that is — before Democrats sleepwalk into a nightmare.

I’m still hoping Mahan pulls a rabbit out of the hat. I remain skeptical he has the tools to do it. Absent that, we will likely watch Democrats coalesce around a candidate who doesn’t come close to meeting the bar this state needs — and call it a win for keeping the other party out.

California deserves better than that. Whether it gets it is an open question with forty days left to answer it. Pretty depressing at the moment.

Next flash update: after the April 28 CBS debate at Pomona College. Nine candidates on stage. Different moderators. Watch for whether anyone finally breaks through.

The Scorecard

Graded against the Three Pillars. Executive Capability weighted at 50% — consistent with the framework’s core argument that this is the pillar that determines whether California actually gets better. Values and Political Skills weighted at 25% each. Republicans included for structural transparency — both are disqualified on Values.

A note on the grades: The three pillars are not weighted equally. Executive Capability counts double because it’s the pillar that determines whether California actually gets better. Values without execution is Newsom. Political Skills without execution is also Newsom. The missing pillar is always the one that costs you.

The conclusion of a three-part series. Part I is here. Part II is here. This is the longest of the three pieces — because it's where the argument finally makes its full case.

The Verdict Doesn’t Matter

The jury came back. After nine days of deliberation and more than forty hours in the room, twelve people in a Los Angeles courtroom found Meta and YouTube both negligent — specifically finding that both companies acted with “conscious disregard” for child safety. The jury awarded $6 million in total damages: $3 million compensatory, with Meta bearing 70% and YouTube 30%, and another $3 million in punitive damages. Both will appeal.

Google’s response:

“This case misunderstands YouTube, which is a responsibly built streaming platform, not a social media site.”

That sentence is the defensive posture. Not an announcement of change. Not a commitment to build something different. An argument that the jury got it wrong. It is the entire problem this piece describes, in one line, on the day the verdict landed.

The $6 million number will tempt people to call this a minor setback for platforms worth trillions. It isn’t. The same week, a separate New Mexico jury handed Meta $375 million for child safety failures. That finding of malice is why the punitive damages were awarded — and it’s why every one of the two thousand pending cases now goes to settlement negotiations with that word on the table. That changes everything.

The critics who say these bans don’t work are right. Early evidence from Australia shows teens routing around them within weeks — VPNs, fake accounts, borrowed devices. The bans fail as enforcement. But they succeed at something else entirely: they tell every parent in every country that follows that their government has decided YouTube is dangerous for their child. Indonesia begins enforcing its YouTube-specific ban today. You cannot un-ring that bell with a verdict appeal or a press release. The only answer to a legislative and legal verdict is a product that makes both look wrong.

What I want to talk about is not the verdict. It’s a decision the verdict makes more urgent — and one that should have been made more than a decade ago, when YouTube Kids launched and the lane to build something genuinely trustworthy was still wide open.

I’ve Been Here Before

Late 2013. Susan Wojcicki — brilliant, genuinely kind, one of the most respected people I encountered in my years at Google — had just been named YouTube’s new CEO. She would go on to build YouTube into a $40 billion business before losing a battle with lung cancer in August 2024. But in late 2013 she was just taking the reins, and I was finishing out my advisory role — still coming into San Bruno a few days a month, still invested in what the platform could become. Before I fully stepped away, I sent her a memo laying out the unresolved strategic priorities I thought mattered most. One of them was YouTube Kids.

The product hadn’t launched yet. But the conversations were already underway about how to create a safer experience for young children. From what I could see, the framing was being driven primarily by engineering instincts and Google’s formidable advertising machine. The question being answered was: how do we filter YouTube’s existing library for kids and monetize the result?

That was an engineering question. The right question was a trust question.

There was another way to look at it. A curated library rather than a filtered firehose — every video reviewed and approved by human specialists before a child could see it. The creator ecosystem treated like a farm system — real money, quality certification, economic incentives to attract serious talent to the category. And YouTube’s first meaningful subscription: a freemium product where the free tier was genuinely good and parents who wanted ad-free could pay a modest monthly fee. The trust architecture and the revenue architecture were the same thing — build the trust, and the money follows. The lane was essentially uncontested at the time: Netflix had only a basic kids tab, Disney+ was six years away, Amazon’s offering was a tablet app, PBS Kids had no streaming service.

None of it happened. The engineering culture’s answer was to use computers to solve everything. And the lawyers had their own objection.

Section 230 was the shield — predicated on the idea that a platform passively hosts while a publisher curates. Curation, the lawyers argued, risked the shield. And so every time human specialists selecting and approving content came up, the legal red flags went up — even though the same logic, applied consistently, would have prevented YouTube from building a creator economy, and leadership had wisely overridden it there.

The irony that is now impossible to ignore: Judge Kuhl — a Reagan DOJ veteran and George W. Bush’s nominee to the Ninth Circuit — identified the absence of meaningful parental controls as the design defect that survives Section 230. The defensive crouch produced the exposure it was designed to prevent.

YouTube Kids launched in February 2015 as an algorithmically filtered version of YouTube’s existing library — ad-supported, colorfully designed, and built on a promise that the systems would keep the bad stuff out.

I watched it go wrong in slow motion. When Elsagate broke in 2017 — toddlers watching Peppa Pig get her teeth violently extracted, Elsa from Frozen depicted in scenes of abuse, Spider-Man acting out bondage scenarios, all of it slipping through on YouTube Kids behind colorful thumbnails and keywords that said “educational” — the structural problem was clear: there is a 99/1 rule at play, and YouTube Kids is on the wrong side of it.

You can use machine learning to screen 99 out of 100 bad videos perfectly. But it only takes one — one horrifying, viral, widely-shared incident — to destroy the brand equity you’ve built around children’s trust for decades. A platform that filters a firehose is always one failure away from another Elsagate.

Not because the technology is bad. Because the model is wrong.

You cannot filter your way to trust. You have to build toward it deliberately.

The FTC made the point with a $170 million fine in 2019. YouTube was caught marketing to toy companies as “the #1 website regularly visited by kids” while telling regulators it had no users under 13. The credibility gap with parents widened. The brand damage compounded.

And then I had a grandson.

(Animated Version)

The View From the Living Room

My son is a physician who has watched the adolescent mental health research pile up and concluded, professionally, that bans may be the only solution. He acknowledges YouTube is different from TikTok. But when his wife — my daughter-in-law, an accomplished product designer — tried to set up YouTube Kids for my grandson, the experience confirmed everything I’d been saying for a decade. I mentioned in Part I that she called it a horrible product — here’s what actually happened.

She navigated the parental consent screens, the Family Link routing, the math problem to verify she was an adult. She did everything right. And then she used the product — and found it, in her words, “horrible.”

Not because she was doing it wrong. But because she’d just consented to a product whose setup screen disclosed: “Our systems work hard to exclude content not suitable for young kids, but we can’t manually review all videos. You may find something inappropriate that we missed.”

That sentence is the product.

A parent clicks through it and hands their child a platform whose maker has just disclosed — in the setup flow — that it cannot guarantee what the child will see. The legal obligation is satisfied. The trust relationship is not established.

Watching her frustration, I felt something I can only describe as vicarious responsibility. I was there when these decisions were being made. I argued for a different path on my way out the door. I didn’t prevail. And now a generation of parents is living with the consequences.

Stop Playing Defense — and Return to First Principles

Here is the trap YouTube is in — and it is a trap of its own making.

The differentiation argument has been made in court, in regulatory briefings, in congressional testimony, and now in the response to this trial — and regulators keep bundling YouTube with TikTok regardless. Australia heard the argument and banned YouTube anyway. The momentum toward bans is not slowing.

You cannot argue your way out of a product problem. You build your way out.

Google was built on a founding idea: use the company’s capabilities to do things that genuinely benefit people, and the business will follow. YouTube was built the same way — by leaders who made bold product decisions over legal objections.

The principle we operated by: only build products where you can lead a category or define a new one.

Clayton Christensen called it The Innovator’s Dilemma: the better you are at serving your current model, the harder it is to build what comes next. Every platform in this space is optimizing for its most engaged users — which is precisely why none of them have built something that genuinely works for families. The ignored market is always the next market.

That opportunity exists right now, in the one space the entire industry is getting wrong. Not one platform has built something that genuinely works for families — that gives parents real agency over the algorithm, gives teens a path to earning independence, and treats the family as the architect of the experience. Despite what TikTok’s current massive ad campaign attempts to convey, it’s model remains fundamentally engagement optimization; parent-programmable preferences would dismantle their core mechanic. Instagram is built on social comparison. Neither can follow YouTube into this lane without ceasing to be what they are.

This is the next category YouTube can own. Not by fighting the regulatory tide — by building the products that make it unnecessary.

Two Products. Two Problems. Two Brands.

Young children and teenagers are not the same problem. They are not the same user. They should not share a brand. A 4-year-old cannot be a party to a programmable experience. A 15-year-old is a genuine user with their own agency, their own identity, and their own very strong opinions about what is and isn’t cool. The brand that makes a parent feel safe handing a tablet to a toddler is precisely the brand that makes a teenager want nothing to do with it.

Two products, two brands, built on shared infrastructure but presented as entirely separate things. What follows are put forth as my version of what I hope will be constructive “food for thought”. The product and engineering talent to take them somewhere better exists inside that building.

Product One: YouTube Family (Ages 2–12)

YouTube Family replaces YouTube Kids entirely. Same target audience, completely different architecture. Where YouTube Kids is a filtered firehose — algorithmic, reactive, always one failure away from the next Elsagate — YouTube Family is a library. Every video approved before a child can see it. AI does the heavy lifting of pre-screening and candidate surfacing; humans make the final call. A recommendation engine set by parent preferences, not by a computer deciding what a 4-year-old sees next.

This is the architecture that escapes the design defect framing — because there is no algorithm to defend. There is a library, and every video in it was chosen by either the parent or a professional curator.

Parents can easily build playlists on the fly from the approved library — pulling content for a car trip, a quiet afternoon, a topic their child is suddenly curious about — making the experience collaborative rather than passive. A family room mode for connected TVs, where families actually watch together, is its own product surface and arguably where that trust matters most.

The supply side matters as much as the architecture. A dedicated creator investment program($100M) — CPM bonuses, production grants, quality certification — reverses the exodus of serious talent from kids content that the 2019 COPPA fine triggered, and builds a defensive wall against the flood of AI-generated zombie content that will otherwise fill any unguarded library.

The business model is freemium — a free tier that is genuinely complete, ad-supported with explicit parental consent, and a Premium tier that simply removes the ads. The free tier has to be real. A crippled free tier is not a solution to the child safety problem — it’s a solution for families who can afford it. And parents who have explicitly consented, configured their family’s experience, and trust the product are exactly the audience brand advertisers will pay a premium to reach. The trust architecture and the revenue architecture are the same thing.

I know what happens when this idea enters a room full of engineers at Google. Curation is editorially driven, subjective, and culturally foreign to a company that has solved every content problem with algorithms. The instinct will be to automate — which recreates the exact failure mode this product is designed to escape.

The answer is not a massive human curation operation at YouTube’s existing scale. It’s a meaningfully smaller library, AI-assisted review that makes a small team of specialists dramatically more productive, and a Trust API that distributes editorial judgment to third parties — Common Sense Media, educational institutions, faith communities — who are deeply invested in getting it right for their own audiences. By distributing that judgment through a Trust API to third-party experts(the parent chooses), YouTube isn’t becoming a gatekeeper — it’s becoming a platform for trust, moving the bias conversation from the algorithm to the experts, where it belongs. The organizational commitment required is real but modest by Google’s standards. The harder ask is cultural: accepting that some decisions about what young children see require human judgment, not just better models.

Product Two: YouTube Rise (Ages 13–17)

The teen product stands entirely on its own. Parents know the infrastructure connects. Teens don’t need to. The name — whichever direction YouTube’s brand team takes it — should signal aspiration, creative agency, and forward motion. Not surveillance. Not restriction. Not a watered-down version of something better.

The direction worth exploring: a parent-programmable algorithm that teenagers actually want to use. Parents and teens configure the experience together at setup — parent articulates values, teen maps interests — but from the teen’s perspective what they have is an algorithm that respects what they’ve told it they value, not one optimizing for engagement metrics they never agreed to. That framing matters enormously. “You program the algorithm” lands completely differently than “your parents control what you watch.”

Rise earns teen credibility not by hiding the parental architecture but by making the teen the primary beneficiary of it — an algorithm that actually reflects what they value, a creative platform that leads somewhere real, and a path to autonomy that no TikTok account can offer.

The joint setup works most naturally at 13 and 14 — the transition age from YouTube Family, when teens are still figuring out what they like and parents still have genuine leverage. That’s the beachhead. The graduation system carries the product forward as teens age into more autonomy, with the parental layer becoming lighter at each level until 17-year-olds are operating with near-full independence and a clear arrival moment at 18.

The bypass problem — teenagers routing around the product with fake accounts — is real. But it’s a product design failure, not an enforcement problem. You solve it by making the inside better than the outside.

The fundamental difference is architectural. Today’s YouTube has one principal: YouTube. Its algorithm optimizes for YouTube’s engagement metrics. Everything else — parental controls, time limits, content filters — is constraint applied on top of a system built to serve a different master. YouTube Rise inverts that entirely. The algorithm’s principal is the family. What keeps a teen watching longest is an input, not the objective. That is not a different version of the same machine. That is a different machine.

This product reframes the entire teen relationship with YouTube — and with the regulatory argument. The platforms being targeted by bans and litigation are consumption engines. YouTube Rise can be something different: a creation platform. New AI tools are making it possible for a teenager with a phone and an idea to make something genuinely good — not AI slop, not a filter on someone else’s content, but real video with real craft. The move from user to creator is YouTube’s home-field advantage. TikTok and Instagram have creator tools. What they don’t have is the infrastructure to make creation mean something beyond the moment — the monetization depth, the audience loyalty, the long-form format that rewards genuine skill. A teen who builds on YouTube Rise builds something with lasting value. A viral TikTok is a moment. A YouTube channel is a career.

The bypass problem solves itself when the product is the place where you become something.

The obvious counter is loneliness — a teen whose algorithm is family-programmed while their friends are on TikTok’s open social graph will route around it in minutes. The answer is a Rise-native social layer: shared playlists, peer taste graphs, creator communities that exist only inside the ecosystem. You don’t compete with TikTok’s social graph by copying it. You build a better one for the people who chose Rise — and make leaving it socially costly.

One prerequisite worth naming: YouTube Rise only works if the underlying platform can hold the line on what AI is about to unleash. Deepfakes, synthetic intimate imagery, AI-generated harassment — these aren’t hypothetical, they’re already in classrooms. One viral incident involving AI-generated harm to a real teenager destroys the trust architecture faster than any filter failure ever could. Tightening community standards and investing in AI-specific detection isn’t a feature of Rise. It’s the foundation.

YouTube Rise also cannot be launched through traditional marketing channels. It has to arrive through teen creators who make it part of their own brand — “I’m building on Rise” said publicly and meant. The creator investment program funds exactly that. The supply side and the launch strategy are the same investment.

YouTube’s Job Is to Build. Parents’ Job Is to Choose.

It is not YouTube’s responsibility to parent children. Only parents can do that. YouTube’s job is to build something so genuinely good, so easy to find, and so clearly differentiated that parents who care — and there are hundreds of millions of them — actively seek it out and choose it. Not because they’re required to. Because it’s the obvious choice for their family.

Two objections will arrive immediately: that Rise will cannibalize YouTube’s teen audience, and that it’s somehow mandatory. Both are wrong. The cannibalization is already starting to happened — through parental blocking and bans. Rise brings those users back and limits the downside risk. And it is a choice, not a mandate — standard YouTube continues unchanged for families who don’t opt in.

A defensive posture produces a defensive product. And a defensive product doesn’t get adopted.

YouTube Family and YouTube Rise need to feel like choices, not compliance. Every design decision — from setup to graduation — should carry the same message: we built this for you.

The constraints worth naming

None of this is without friction. Both products reduce the addressable advertising universe for youth — a direct conflict with YouTube’s core revenue model that will generate internal resistance regardless of the CPM premium argument. The existing YouTube Kids migration requires as much careful design as the new products themselves. And the international regulatory patchwork — GDPR, the UK Children’s Code, Australia’s eSafety framework, COPPA 2.0 — doesn’t resolve into a single architecture easily. Leadership knows these constraints better than I do. But done correctly, products built to satisfy the most demanding of them are genuinely un-bannable in every market currently moving toward restrictions.

Shorts

Absent from YouTube Family entirely. Disabled by default in the teen product. When parents and teens choose to enable it together, it operates differently — interest-based, not slot-machine. Clear endings instead of infinite scroll. This directly answers the eSafety Commissioner’s specific critique of YouTube’s “persuasive design,” and it gives YouTube something no competitor can offer: a short-form video experience that is architecturally defensible in every market currently moving toward bans.

The Business Case

Yes, both products reduce short-term advertising inventory for youth — that tension is real and anyone who says otherwise hasn’t run a P&L. But done correctly, YouTube is uniquely positioned to absorb any short-term hit — the scale, the subscription revenue, the connected TV growth, and the brand advertiser premium that a genuinely trusted family product commands all work in its favor.

The alternative isn’t the status quo anyway. The alternative is a platform being banned in major markets, litigated in every jurisdiction, and avoided by the parents who make household media decisions.

Trust is the revenue architecture — parents who genuinely trust the product pay for it, accept advertising in it, recommend it, and stay with it for years. That is an enormous, durable, and currently uncaptured market. YouTube has the content library, the creator economy, the AI infrastructure, the device distribution, and the brand recognition to build something no competitor can replicate.

The question is whether leadership decides to build it.

Break From the Coalition

In the courtroom, YouTube is different. In the Capitol, YouTube is the same.

By fighting alongside Meta for broad Section 230 immunity, YouTube is subsidizing a legal defense built around the very characterization it is trying to escape.

The history of companies that broke from their industry coalition and won is instructive — and so is the history of those that didn’t.

Volvo invented the three-point seatbelt, opened the patent to every competitor for free, and owned “safety” as a brand attribute for 65 years. Johnson & Johnson recalled 31 million bottles of Tylenol in 1982 when lawyers counseled against it — within a year the brand had fully recovered, and it became the defining case study in choosing consumer trust over liability management. When the state attorneys general tobacco litigation came, Liggett broke from the industry coalition first, cooperated with investigators, and got dramatically better terms. The holdouts paid $206 billion.

Apple introduced tracking transparency with iOS 14.5, 96% of users opted out, Facebook lost an estimated $13 billion in annual revenue, and Apple owned “privacy” while its competitors faced congressional hearings.

The pattern: break from the coalition, build what competitors can’t copy, define the category, win.

The counter-case is Boeing. Concealed the 737 MAX’s problems, fought the FAA, stayed behind the lawyers, maintained the “we’re compliant” posture. Three hundred and forty-six people died. Twenty billion dollars in costs. Criminal charges. The brand still hasn’t recovered. That is what the cautious path looks like when it runs all the way out.

YouTube is not Boeing. But the choice in front of it rhymes.

Quietly is not enough. The moment YouTube builds these products and steps publicly out of the coalition — we are not the same as these platforms and we will not be defended as if we are — the entire narrative shifts. Without that signal, even genuinely differentiated products get read as litigation strategy. With it, YouTube becomes the company that solved the problem rather than the company that kept fighting the response to it.

Leaders Lead. Followers Lose Their Way.

I know the pressures. Lawyers counseling quiet during the trial. Business teams protecting Shorts metrics. Policy teams managing coalition relationships. The instinct is to wait — to not hand plaintiffs new product admissions, to not disrupt revenue, to let the legal process run.

That instinct is how you lose control of your own story.

The Trap is real. The current model is working. The metrics are good. The lawyers are counseling patience. That is exactly what Christensen described — and exactly how market leaders lose categories they should have owned.

YouTube did not become what it became by letting lawyers set the direction. It became what it became by leaders who decided where they were going and built toward it — over legal objections, over institutional resistance. That is the founding DNA of Google. That is the DNA that built YouTube.

It is not the DNA that built YouTube Kids.

YouTube fell in the water. The COPPA fine, Elsagate, the lawsuits, the bans — that’s the fall. The question is not how it happened. The question is whether leadership decides to swim.

Caution is staying in the water. Incremental product and feature improvements have not moved the global regulatory momentum one degree. Fighting this through litigation and lobbying is a fool’s errand — it will produce worse outcomes, applied more broadly, with less room to shape the terms.

There is a version of this where YouTube gets banned in major markets, loses a generation of users, and looks back at this moment as the one where it could have seized the narrative and chose instead to manage the legal risk. There is another version where YouTube builds the products that make the bans look unnecessary — that make parents advocates instead of adversaries, that give regulators a better answer than the one they were about to impose.

Leaders choose the second version. Followers get crushed by the events they could have controlled but chose not to.

A Note to Neal

I have no inside knowledge of what YouTube is working on. I have no visibility into what Neal is navigating. For all I know, conversations are already underway inside that building that would make parts of this look obvious or outdated. I’m offering the outside perspective for what it’s worth — which is exactly what regulators, parents, doctors, and juries see. Sometimes that view is useful precisely because it isn’t carrying the internal weight.

What I do know is that Neal is a serious person who takes this platform seriously. He was named Time’s CEO of the Year. He has three children. He has spent his career at a company that built its identity on genuinely useful products — and he has every reason to understand the difference between managing a crisis and solving one.

I also know these decisions are harder to make today than they would have been in 2014. The legal exposure is larger. The regulatory environment is more hostile. The coalition is more entrenched.

But the tools that didn’t exist in 2014 — AI capabilities that can make human curation scale, that can empower parents as genuine programmers of their family’s algorithm, that can detect synthetic zombie content and deep fakes before it reaches children — are available now. The window is open. It won’t stay that way.

My grandson deserves a YouTube his parents can trust. The platform I helped build deserves better than having its legacy defined by regulatory battles rather than product leadership.

You don’t drown by falling in the water. You drown by staying there.

What’s required is the same courage that built YouTube in the first place. I’m confident it still exists inside that building. I’m hopeful this moment calls it forward.

Dean Gilbert spent 2006–2013 at Google and YouTube, serving as VP of Product at Google and VP of Global Content and Operations at YouTube during the platform’s most formative period. He writes about the intersection of media, technology and politics at The Citizen on Substack.

This is the second in a three-part series. Part I is available here. Part III follows soon.

Annalee Schott was a beautiful little girl who loved animals and life on the alfalfa farm in Colorado where she grew up. By her mother’s account, she was kind beyond words. Then the algorithms found her.

Instagram. TikTok. Snapchat.

The content in her feed shifted. The videos grew darker. The messages told her, as her mother later described it, that there was “one way to take away the pain” and that she would “never fit in.” In 2020, Annalee died by suicide. She was a teenager. Her mother Lori drove to Los Angeles this past February to sit in the courthouse as the trial began — not as a plaintiff, just as a witness to a reckoning she’d been waiting years to see.

She wasn’t alone. Parents from across the country filled the courthouse gallery. They didn’t speak. They didn’t have to. Throughout four hours of Senate testimony in January 2024, in the most emotionally charged tech accountability hearing in Congressional history, they sat silently holding up photographs of their dead children.

When Mark Zuckerberg finally stood and turned to face them, he said: “I’m sorry for everything you have all been through.”

The mother of a son who died buying fentanyl on Snapchat called the apology “fake.”

Annalee is not the plaintiff in the Los Angeles trial. That’s Kaley — a 20-year-old identified in court documents only by her initials(KGM), who began using YouTube and Instagram at age six and says the platforms rewired her adolescence into depression and self-harm. Kaley is the legal face of this reckoning. Annalee is the moral one. The difference matters because the courtroom can adjudicate damages. It cannot adjudicate what kind of company YouTube wants to be.

I’m recounting these stories because this is the room YouTube is implicitly in — every day, in every courtroom, every legislative hearing, every parent’s kitchen table conversation — whether it shows up or not. And I want to be honest about what I’ve come to believe about why YouTube hasn’t shown up.

It’s not malice. It’s something more complicated than that. And in some ways, more troubling.

A Word About the People Inside

Before I go further, I want to say something I mean.

I know several of the people running YouTube today. I worked alongside some of them during the years when we were building something that felt genuinely world-changing — the shift to watch time, TrueView, the creator economy, YouTube on every screen. The people I knew then were idealistic, talented, and deeply committed to making a platform that served humanity. I have no reason to believe that’s changed.

I also know what it’s like to run a giant. When I was there, YouTube was already enormous — and already hard to steer. Today it is orders of magnitude larger, with more layers of management, more legal exposure, more regulatory jurisdictions, more competing stakeholders than existed when I was in the building. Decisions that should take weeks take months. Lawyers who counsel caution on everything — as they always have — now have more precedents to cite and more risk surfaces to protect. The organization that once moved with the velocity of a startup now moves with the weight of a nation-state.

I understand that. I’ve lived it.

But understanding why something happens doesn’t change what it produces. Parents and teens have always known YouTube is different — they permit it for homework, documentaries, and music in ways they’d never allow TikTok. That distinction is real and it lives in millions of households. But now it’s getting less and less translated into the arenas where this distinction can afford protection: the courts, the regulators, and in the reform community. YouTube now finds itself bundled with the platforms that made fundamentally different choices. Why?

The legal term for what I believe has happened here is not malfeasance — the deliberate commission of a wrongful act. It’s nonfeasance — the failure to act when a duty to act existed.

Nonfeasance is harder to defend precisely because it’s not a decision you can point to, reverse, and apologize for. It’s the accumulation of a thousand smaller choices not made — the product roadmap that never got resourced, the coalition that never got broken, the hearing the lawyers advised skipping, the “Shorts” launch that prioritized competitive response over differentiation. No single decision is obviously wrong. The outcome is catastrophic anyway.

There is a specific example that crystallizes this for me. For years, YouTube’s internal North Star was one billion hours of daily watch time — a goal the team pursued with genuine pride, and earned. I understand that pride. I was part of building the culture that made that kind of ambition possible. But in a building that large, whatever you optimize for becomes what you are. If the North Star is watch time and safety isn’t also a North Star, then you aren’t choosing to harm children — you are simply failing to choose not to. That is the essence of nonfeasance. And it is why the people inside can be good, idealistic, and genuinely trying, while the institution still drifts in the wrong direction.

The Experts Have Reached a Verdict

Here is a fact I spent weeks trying to disprove: I cannot find a single prominent, credentialed, independent expert in child safety or the attention economy who has publicly and substantively defended YouTube as categorically different from TikTok or Instagram. Not one. Not Tristan Harris. Not Jonathan Haidt. Not Frances Haugen. Not Sherry Turkle. Not Cal Newport. Not the U.S. Surgeon General.

Every one of them — in books, in Senate testimony, in documentaries, in op-eds, in podcasts reaching millions — has treated YouTube as part of the same machine.

Jonathan Haidt, whose book The Anxious Generation is the most influential popular work on this subject in a generation, does differentiate between platforms — he singles out Instagram as the primary driver of depression in teenage girls, TikTok as making kids “stupid and unable to pay attention.” But when Haidt talks about what kids default to when given a school device, he lists “YouTube or TikTok” as interchangeable distractions. In his bestselling book, he cites “Instagram, TikTok and YouTube” as the platforms where teenagers spend five or more hours a day — the hours that correspond to the mental health crisis he’s documented.

Tristan Harris is the person most equipped to make YouTube’s differentiation argument. More than anyone alive, Harris has built the intellectual vocabulary for understanding why platform architecture produces harm. He invented the framework: the dopamine loop, the slot machine, the attention hijack. He understands that design choices are moral choices — and that YouTube’s design choices are genuinely different from TikTok’s.

He has never made that argument publicly.

Before Congress, Harris testified that Facebook, YouTube, and their peers share a business model designed to “create a society that is addicted, outraged, polarized, performative and disinformed.” In his landmark TED Talk, he named “Facebook notifications, Snapstreaks, and YouTube autoplays” in the same breath as competing for “one thing: your attention.” In Senate testimony delivered alongside YouTube executives at the same table, he argued that “when YouTube gobbles up” the role that Saturday morning cartoons once played in children’s lives, “it just takes away all the protections” those standards provided.

That is not a failure of Tristan Harris. Harris engages with any platform willing to make a serious case. In fact, the Center for Humane Technology coined the term "TikTok-ification" specifically to describe how YouTube — through Shorts — and Instagram — through Reels — voluntarily adopted the same dopamine-loop architecture they once avoided. In Harris's framework, YouTube didn't just fail to differentiate. It actively dismantled the distinction. That is a failure of product and narrative leadership — the kind only a company can provide for itself.

U.S. Surgeon General Vivek Murthy called for tobacco-style warning labels on social media in 2024. The American Psychological Association cited teenagers spending “nearly five hours every day on top platforms such as YouTube, TikTok and Instagram.” Same list. Same warning label. The platform used by more American teenagers than TikTok, Instagram, and Snapchat combined has been assigned the same public health warning as a pack of cigarettes. Researcher Jean Twenge, whose longitudinal data underpins much of the clinical case against social media, sharpens the irony: teenagers rate YouTube as “positive” for learning — yet heavy users show the same rates of depression and sleep deprivation as heavy TikTok users. The perceived benefit doesn’t show up in the outcomes.

Not because YouTube is the worst platform. Because no one — not YouTube, not a single credible outside voice — had given anyone a reason to treat it differently.

The Empty Chair

In January 2024, the Senate Judiciary Committee convened the most consequential child safety hearing in Congressional history. Parents held up photographs of dead children throughout four hours of testimony. Zuckerberg apologized. The CEOs of TikTok, Snap, X, and Discord all testified — three under subpoena, legally compelled to appear.

YouTube wasn’t there. Not because its lawyers negotiated an exit. Because Congress didn’t put it on the list.

The committee framed the hearing around child sexual exploitation — grooming, predatory contact, CSAM — and focused on the platforms most visibly implicated in those specific failures. YouTube didn’t make the cut.

TechCrunch noted it at the time, describing YouTube as “often inexplicably absent from these events.”

Inexplicably absent.

Being subpoenaed would at least have meant Congress considered YouTube a threat worth confronting. Being overlooked entirely means the platform used by nine in ten American teenagers had been so effectively removed from the accountability conversation that it didn’t register as a target at all.

That invisibility wasn’t accidental. It was the product of years of quiet coalition management, of letting lawyers and lobbyists define YouTube’s public posture, of never stepping forward to tell a different story.

Here is the cruel irony: inside YouTube, the exclusion almost certainly felt like confirmation that the strategy was working. The committee didn’t call us. The bills aren’t targeting us by name. The Australia exemption held — briefly. Maybe the distinction is landing. Maybe the lawyers are right. Maybe we really are being seen differently.

But any organization reading that room clearly should have reached the opposite conclusion. Five CEOs were being publicly eviscerated. Senators were telling Zuckerberg he had “blood on his hands.” Parents were holding photographs of dead children. The regulatory and cultural tide was visibly, unmistakably turning — and YouTube wasn’t in the room only because it hadn’t yet become the primary target.

That was a warning, not a reprieve. It was the last clear signal that there was still time to act proactively — to step forward, claim the distinction publicly, build the products that made the argument undeniable — before the wave arrived. The right read of that moment was: we are next if we don’t move now.

The record is unambiguous about what happened instead. Every significant child safety move YouTube has made in the past decade followed a forcing event rather than preceding one. The $170 million COPPA fine in 2019 produced a content designation system — after the fine. The January 2024 Senate hearing, where five CEOs were publicly eviscerated, produced expanded teen parental controls — eight months later, in September 2024. YouTube’s own announcement of its AI age estimation system in July 2025 explicitly described it as a response to “mounting regulatory pressure” and the UK Online Safety Act. And the Shorts time controls — the ability to set short-form video to zero for a teenager, a step that costs YouTube essentially nothing in revenue — were announced January 14, 2026. Two weeks after the Los Angeles trial began.

Every action. A reaction. Always the next wave, never ahead of it.

Instead, not being on the list got read as safety. Every near-miss became evidence that the strategy was working rather than evidence that the window was closing. It was survivorship bias dressed up as strategy.

The tsunami doesn’t warn you by name before it arrives. YouTube was absent from the room. And got convicted in absentia — too quiet to be defended, too prominent to be ignored.

The Cultural Verdict

Courts take years. Legislation takes longer. But culture moves fast.

This week, a documentary called Your Attention Please opened in theaters after a standing-ovation premiere at SXSW. It features Tristan Harris, Frances Haugen, Vivek Murthy, Sherry Turkle, and Cal Newport. Its launch brought together more than 25 digital wellbeing organizations — Common Sense Media, the Center for Humane Technology, Smartphone Free Childhood — releasing the trailer simultaneously in what the filmmakers called a “Stand for Their Attention” moment.

(NOTE: THIS IS A YOUTUBE VIDEO)

The film is powerful, emotionally honest, and important. It will be seen by hundreds of thousands of parents trying to understand what is happening to their children. And in it, the most sophisticated critics of the attention economy treat YouTube as part of the machine. Not the worst part. Not the primary villain. Just part of the same system that monetizes children’s attention and calls it connection.

They’re not wrong to see it that way. That is what a decade of nonfeasance has produced.

There is one detail from the MDL 3047 litigation(the federal case that consolidates 1600 product liability claims) that sharpens this. According to court filings, employees within YouTube once acknowledged internally that “driving more frequent daily usage was not well-aligned with efforts to improve digital wellbeing.”

They wrote it down. They knew. And the platform kept optimizing for engagement anyway. Not through malice. Through the accumulated weight of an organization that had more reasons to wait than to act.

The Earthquake

What YouTube faces is not a trap. A trap has edges — a path around it if you move quickly enough.

What YouTube faces is an earthquake.

The cracks are in every wall simultaneously. The legal system has bundled YouTube with the worst actors in its industry. Global regulators have treated it as indistinguishable from platforms it is genuinely quite different from. The intellectual reform community has never deployed its vocabulary in YouTube’s defense. The cultural moment has placed YouTube in the same frame as Instagram. And inside YouTube’s own filings, there are documents that say its own people knew the algorithm was pulling against the platform’s stated values for young users.

When the ground moves in all directions at once, no defensive posture holds. The strategy of staying quiet, fighting from inside the coalition, adding incremental controls while hoping the legal situation resolves — that strategy is over.

YouTube Choice, Part III-What playing offense looks like and the path forward. Out soon!

Dean Gilbert spent 2006–2013 at Google and YouTube, serving as VP of Product at Google and VP of Global Content and Operations at YouTube during the platform’s most formative period. He writes about the intersection of media, technology and politics at The Citizen on Substack

A few months ago, I wrote about Gavin Newsom and what I called the execution deficit — the gap between his rhetorical ambition and his actual results. That piece struck a nerve. Since then, Newsom has done nothing to change my assessment. He’s on a book tour. He’s positioning for 2028. And The Daily Show recently ran a devastating segment tracking his promises on homelessness and high-speed rail from 2010 to 2025 — the same promises, year after year, with almost nothing delivered. The satirical framing was a Hollywood “for your consideration” spot: Newsom as a “leading man” who is “preparing to be released nationwide.”

The bullet train that was supposed to connect San Francisco to Los Angeles now connects Merced to Bakersfield. And the governor is boasting about it. Give me a break! Just imagine for a minute what the Red team is going to do with this sorry track record.

The Late Conversion

To his credit, Newsom has recently signed real reform. When the LA wildfires hit in January 2025, he suspended CEQA and Coastal Act permitting overnight to accelerate rebuilding — the very regulatory regime he’d spent years defending. Suddenly, when the political heat was unbearable, permits could move in 30 days instead of years. Two months later, he hosted Ezra Klein on his podcast to discuss Abundance and was defensive — saying he “struggles a little” with the book because it doesn’t acknowledge the reforms he’s been advancing. Then, three months after that, he signed sweeping CEQA exemptions for housing and infrastructure and slapped the “Abundance Agenda” label on it.

The reforms are real and they’re good. But the timeline tells you everything. If CEQA could be waived in January 2025 and overhauled by June, why wasn’t it done in 2019, 2020, 2021, 2022, 2023, or 2024 — when the housing crisis was already at emergency levels and homelessness was climbing? Even Klein, whose book Newsom now cites as inspiration, put it plainly: “Abundance is in the end, not about what you say, it is about what you deliver.” Adopting the language of reform after failing to deliver reform for six years isn’t leadership. It’s rebranding.

That’s the bridge to this governor’s race. The question isn’t just who replaces Newsom. It’s whether we replace him with the same archetype — a communicator without a track record — or whether California is ready for something different.

The Three Pillars

Most political commentary collapses everything into one dimension — “do I agree with them?” or “can they win?” That’s how we end up with leaders who look great on paper and can’t deliver. I’ve landed on three pillars for evaluating candidates, and I’m sharing them because I think they’re useful for anyone trying to make sense of this field.

Pillar One: Values. Does this candidate share my core beliefs about what government should do and, frankly, not do— how it treats those that have the least political power among us vs. those that seem to hold all the political power these days. This is the starting point. It’s necessary. But as I learned watching Newsom for a decade, it’s wildly insufficient on its own. I sense we’d be in strong alignment on the values pillar.

Pillar Two: Executive Capability. Has this person actually run something? Not talked about it. Not legislated about it. Actually managed an organization, set goals, built systems, hit meaningful targets. For a governor — who is fundamentally a CEO of a massive, dysfunctional enterprise — this is arguably the pillar that matters most. It’s the one Abundance is really about. And it’s the one the political class ignores. What’s the legacy you’ll leave when you are out of office—the signature achievement— that will stand the test of time and make a real difference in the lives and well-being of your constituents?

Pillar Three: Political Skills and Viability. Can they win, communicate, connect? Are they genuine? A governor with no political capital can’t get anything done. These intangibles matter. It seems to me Newsom’s “signature achievement” as Governor has not been in the realm of tangible things that will matter for most of his constituents, but of a political nature—like leading and passing Prop 50 to help fix the distortions being caused by the redistricting calamity in Texas, which started all of this. I give him a lot of credit for this, but this is NOT the legacy achievement I had in mind for the Pillar Two, but it does give him a high score as a “politician”.

Newsom scores high on One, very high on Three, and — in my view — disqualifying on Two. Two out of three isn’t good enough when the missing pillar is the one that actually affects people’s lives.

The Field

Let me clear the underbrush.

The Republicans are non-starters. Steve Hilton and Chad Bianco are both MAGA-aligned Trump supporters — Hilton’s slogan is literally “Make California Golden Again,” and Bianco endorsed Trump by joking about putting “a felon in the White House.” One is a former Fox News host who’s never held elected office. The other is a sheriff affiliated with the Oath Keepers. Both want to abolish the state income tax and hand environmental regulation to the federal government. These are not serious candidates for governing the fifth-largest economy in the world. I won’t be recommending either. I actually took a serious look. And frankly, that’s a tragedy for the state. California deserves a credible center-right opposition that cares about auditing the books, reforming broken systems, and holding government accountable — not culture warriors running on MAGA slogans in a state where Trump has a 62% disapproval rating—and for good reason!

They do matter structurally, though. California’s top-two primary means the two highest vote-getters on June 2 advance to November, regardless of party. Combined, Hilton and Bianco are pulling roughly 24-30% of the total vote. With eight Democrats splitting the remaining 45-50% among themselves, the math gets genuinely scary — Democratic strategist Paul Mitchell puts the probability of two Republicans making the November runoff at 27%. That risk shapes everything about how I’m evaluating this race.

The lower-tier Democrats — Thurmond, Yee, Becerra, Villaraigosa — are all polling at 5% or below with no visible trajectory. The state party chair asked them to drop out. They refused. Their names stay on the ballot. Some have real credentials, but none has been able to translate experience into support.

Tom Steyer has spent $38 million of his own money and is polling at 11%. That’s a terrible return on investment, and self-funded candidates in California have a historically poor track record. He supports the billionaire tax and wants to rewrite Prop 13. His fiscal orientation doesn’t match mine. I don’t see the path.

A Citizen’s Note on the Billionaire Tax: The proposed 2026 Billionaire Tax Act would impose a one-time 5% wealth tax on California residents worth over $1 billion — roughly 200 people — with proceeds going mostly to healthcare. I think it’s a terrible idea. Not because the wealthy don’t need to pay more — they absolutely do — but because undertaxing wealth is a federal problem that requires federal solutions: fixing corporate rates, closing loopholes, funding the IRS, which I wrote about last year in a post titled IRS: Happy April 15th..NOT!. A billionaire can move to Nevada in a weekend; they can’t move to another country as easily. Six have already left — Brin, Thiel, Spielberg among them — taking $27 billion in potential revenue with them. If this passes, the rest will follow, and California will lose far more in ongoing income tax revenue than it ever collects from a one-time levy. Nearly every serious candidate in this race opposes it.

That narrows it to two candidates worth talking about.

Eric Swalwell: The Frontrunner I’m Cautiously Supporting, but for the wrong reasons

Swalwell leads among Democrats at 17%. He has Adam Schiff’s endorsement, ten sitting members of Congress, and as of this week, the SEIU — 750,000 workers and the biggest labor endorsement in the race. He communicates well, he has momentum, and he’s the safest bet to keep a Democrat in the top two.

On values, he’s a mainstream Democrat. Broadly aligned with where I am. He hasn’t taken a clear position on the billionaire tax — hedging between his labor base and moderate voters — which tells you something about his political instincts, if not his convictions.

On political skills, he’s the strongest in the field. If you’re making a pure electability argument, Swalwell is your candidate. I give him extremely high marks for his willingness, at great personal risk, to take on King Trump. He’s a hero on that score.

On executive capability, he has nothing. He’s a congressman. He’s never run a city, a state agency, or a business. His lead message — that the governor’s most important job is to “keep Donald Trump and ICE out of our streets and out of our lives” — is a frankly a ridiculous pitch, not a governing plan.

The SEIU endorsement is revealing. The union backed him because he “aligns with our members” and “is able to win.” Not because he’ll reform state government. Not because he has a track record of delivery. David Crane, who teaches public policy at Stanford, put it bluntly: Swalwell will be “putty in the hands” of the public-employee union establishment. The unions aren’t backing him because they think he’ll challenge the status quo. They’re backing him because they think he won’t.

I’m offering cautious support for Eric because the math demands it — we cannot afford two Republicans in November. But I’d be lying if I said I was excited. My worry is that we’re electing another communicator-without-a-track-record, and four years from now we’ll still be wondering why the housing crisis is still a crisis and the state still can’t build anything on time.

Matt Mahan: The “Uncola” I like

And then there’s the San Jose mayor. Here he is today on The Daily Show. Judge for yourself:

Matt Mahan is polling at about 3%. He entered the race seven weeks ago. Late by any standard, but working hard to catch up. He has almost no traditional party endorsements. The unions aren’t behind him. He’s a long shot by any conventional measure.

He’s also the candidate who most closely matches what I believe California actually needs. I’m donating to his campaign. Today!

On values, Mahan is a Democrat who opposes the billionaire tax, prioritizes the state’s business climate, and has argued — not as a Republican talking point but as a diagnostic from inside the system — that California’s over-regulation and costs are driving families and companies out. He’s not running against Democratic values. He’s running against Democratic complacency about results. His anti-Trump and anti-Maga to the core, which is a litmus test for me. But, it’s not the central tenant of his campaign, nor should it be.

On executive capability, this is where the contrast with Swalwell is sharpest. Mahan is the only candidate currently running a major California city. His results in San Jose aren’t perfect, but they’re real and measurable. Unsheltered homelessness dropped 10.7% under his quick-build community approach. Sheltered individuals increased 39%. He implemented performance-based budgeting, accountability dashboards, and tied employee pay to outcomes. Before politics, he co-founded and ran a tech company. Before that, he taught seventh and eighth grade through Teach for America. He grew up in Watsonville on a work-study scholarship.

He’s also been the most direct critic of Newsom’s governing approach — not his values, but his execution(I like this a lot). He wrote that the governor’s social media trolling “goes beyond taking on the excesses and abuses of the Trump administration and begins disparaging businesses merely for expressing concerns over very real problems.” He pointed out that California has the highest unemployment, the highest energy and housing costs in the continental U.S., and the highest effective poverty rate. His campaign line: “The strongest resistance is results — not just more rhetoric.”

If Newsom is the archetype I critiqued in my last piece — vision without delivery, branding without building — Mahan is the “uncola”. Outcomes over optics. Fixing over fighting. The boring work of actually governing. Boring is good!

On political skills, he’s weakest. Three percent. Low name recognition outside the Bay Area. His donor base is Silicon Valley tech — Sergey Brin, Joe Lonsdale, Garry Tan, Rick Caruso — which gives him money but also the “billionaire’s candidate” label. He’s raised $8.5 million and got a Super Bowl ad, but seven weeks of spending haven’t moved his poll number.

A word on the tech money, because I know it raises eyebrows. I know some of these people from my own career. Yes, many of them are motivated in part by opposing the billionaire tax — a concern I share. But there’s little evidence Mahan would be beholden to this group in any meaningful way. He’s been vocally negative about the Trump agenda and has shown zero tolerance for the MAGA-curious posturing of some Silicon Valley figures. Opposing a bad tax policy and being captured by your donors are not the same thing, and I’d rather have a candidate funded by people who build companies than one funded by unions whose primary interest is preserving the status quo.

Can he win? Honestly, today it looks like a long shot. But nobody has locked this race up — the frontrunner is at 17%, a quarter of voters are undecided, and half the electorate hasn’t tuned in yet. Prediction markets price him at 13% odds on Kalshi — well above his polling — because traders are pricing in what polls can’t easily capture: the Silicon Valley-funded ad blitz that’s already spending nearly $5 million to introduce him to Southern California, the structural opening created by a fractured field, and the possibility that a late-breaking electorate moves toward the candidate with the most compelling “change” message. The conditions for a late break exist. Whether Mahan’s campaign can execute on them in eleven weeks is the open question.

Where I Land

Here’s the tension I can’t resolve, and I don’t think your vote should try to resolve it either. The three pillars sometimes conflict. The candidate who would probably govern best may not be able to win. The candidate who can probably win hasn’t demonstrated he can govern.

Democrats keep choosing the best communicator over the best builder. That’s how we got Newsom. That’s how we got fifteen years of the same housing and homelessness promises with almost nothing delivered. That’s how we got this high speed rail fiasco. Abundance is a 300-page argument for why this pattern has to break. Progressivism that can’t build things isn’t progressive. It’s just expensive.

If your primary concern is keeping a Democrat in the top two, Swalwell is the pragmatic choice. He has the polling position, the institutional backing, and the momentum. A vote for Swalwell is a vote for the safest path.

If your primary concern is what kind of governor California actually gets, look hard at Matt Mahan. I am! He’s the candidate whose qualifications best match what the state needs — someone who treats government as a vehicle for outcomes, not a platform for performance. His odds are long. But in a race this open, long odds are not the same as no odds. With the largest single column “undecided” he can catch up….

California has had enough leading men. We need a builder.