Yet I can’t help but see software developers becoming 10x more productive thanks to AI. It is completely changing software. I would have expected it to have a much bigger impact in writing, and I’m upset it hasn’t. I would love 10x banger blog posts if I could have them.

Part of the reason we don’t is that writers just aren’t experimenting enough with AI. They believe the hate. They see it produce AI slop and write off every use case.

This isn’t a good thing, so this post is my attempt to change this, at least a little, by sharing how I write with AI.

Writing with AI or using AI to write

These are not the same. I write with AI, using it, but the actual words are all my own. This distinction is important.

The content AI creates is too bad to even anchor on in terms of editing. Even for a sentence, it is often easier to ask AI what to write and then write it yourself than to have it write the sentence directly and then edit it. Use AI to write and no matter how much editing you do, the skeletons of slop will still be with you.

It is simply a strawman to reduce writing with AI down to having AI produce the final prose for you. Is asking AI a question writing with AI?

The medium I write in makes it more possible for me to write with AI too. I write blog posts about enterprise SaaS software. It’s not high art. I’m not aiming to make people cry with my beautiful prose. I’m looking to learn things, discover information that is interesting, and share it with others in a way that is understandable and entertaining. AI helps with this more than writing that is deeply personal, poetic, or stylistic.

The average blog post in my medium is fairly poorly written but people still enjoy it because the information within it is interesting. If I can use AI to help me find and deliver more interesting information, it has done its job.

Researching for examples

One of the most important bits of my work is finding examples. This is the purpose of research. It is examples that prove or disprove some part of a post. You can write whatever you want on the internet, but only examples can really reveal whether it is true or not.

I know it’s not the most scientific, but I usually start a post with a hunch, idea, or opinion. It is critical to not let AI form this for me. I add notes to this, fill out my thoughts, and at some point, it becomes significant enough to write a post about.

LLMs speed this process up significantly. Instead of waiting for sources of information to come to me or spending time seeking sources one by one, an agent can do this for me.

I have a researcher skill whose goal is to find real, quotable, sourced examples. I found that if I just ask for examples, it hallucinates plausible sounding examples from real companies. I get deep in a conversation and after asking for a source, it says “this is just a theoretical example btw” and I slam my computer. Sources and quotes keep it in check.

Another important part of this is source quality. Just using Claude’s default web search tool leads to sources that are gamed (SEO) or give cliché examples. I want sources I would use if I was doing the research myself, ideally from other startups or posters in the space.

So I give the skill the tools I would use if I was doing it myself. Specifically:

• Exa. A search tool for agents that does a lot better than normal Google.

• Hacker News. This is my target audience so what they share and how they talk is important.

• Local PostHog repos. We’ve written about a lot of topics on our blog already, so there might be ideas I can point back to. We also have hundreds of RFCs that can provide insights about how we make decisions.

• PostHog Slack. Although a bit messy, sometimes internal discussions lead to the best ideas.

• Semble. A new tool in my research stack that is a network of links uncovering related ones to the ones I already have.

Claude’s native web search tool fills out the quotes for the links, but I can also go check for myself that something exists. It then creates an entry that looks like this:

• Success signal: PostHog first-party, published on the blog (Vincent Ge). Real production agent: 1,000 orgs onboarded in 90 days.

• The example: PostHog audited its install Wizard with AI observability and found each run costs $6.67, with the trivial conclude step (just building dashboards + a report) eating $1.47 while carrying ~140K tokens of useless context. The fix backfired: splitting into fresh query() calls cut accumulated input 89% but cost more, because every new call rebuilds the whole cache.

• Quote: “As counter-intuitive as it sounds, the Wizard running everything in one giant loop and carrying around all that context is actually very efficient.” — Vincent Ge

• Quote (cache economics): “you need to save 12x more tokens to break even for every token rewritten to the cache.”

• Source: https://posthog.com/blog/optimizing-agent-cost

• How to use it: The single best “know when NOT to token-min” example — the obvious optimization (clear context) lost money once cache reconstruction was priced in. Also proves the notes’ “you need per-workflow attribution first.”

In terms of surfacing relevant and interesting reads for the topic, AI does a better job than me. The example, quote, and success signal all give a better overview of the article than a title (or summary) does. I can then dive into interesting articles and take notes or ask AI more about these articles to fill out my post.

Making use of that research

Between my notes, versions, and research, my drafts end up being many thousands of words, most of which I won’t use. AI works as advanced in-document search across all this. I often ask questions like:

• What else should I add here?

• Can you help me find an example of this?

• Can you find a source for this idea or example?

• Where in this link does it talk about X?

Maybe 8/10 of its recommendations are bad, but 2/10 are good and that is all I need. If none of its recommendations are good, it’s a sign for me to do more research or dive deeper into one of the links that are relevant.

AI is a much better writing partner when it has a lot of relevant context it can search through and you can ask pointed, specific questions of it.

All of this prevents me from needing to have an elaborate knowledge management system. I moved from Notion to Obsidian specifically because I wanted to use Claude Code and connect code context with what I’m writing. I don’t use all the elaborate tagging and backlinking features. I can get a lot of the serendipity of a knowledge management system, without the actual system.

It also helps to have enough knowledge to push back on models when I think they aren’t right. I have been burned enough times and have a strong enough knowledge of my area to be able to call it out.

For example, when researching a competing product, I feed Claude points and ask “Is this true?” One time, it told me product A “supports fast full-text search across all log fields without building a full index” but product B does not. I knew another part of the post contradicted this, so I pushed back and found that both do.

AI as new eyes

Sometimes I’m so deep in a post I can’t see. I can’t see what’s missing, what I’ve already covered, or if I’m explaining something in a way that is too complex or unfair.

In these cases, AI acts as a new set of eyes. I’m leveraging all of the accumulated context to ask questions like:

• What’s missing?

• What could be stronger?

• What’s the common view of this idea?

• Is this a fair interpretation of this idea?

This plays to AI’s strengths. It’s not so good at taking things away, but it is helpful for adding things. I’m again rejecting a majority of recommendations, but it is useful to build confidence in what I have.

For example, for this post, was this actually how I used AI or just how I imagined it? The best way to tell this was to ask Claude. I had the /insights report and I could ask questions of it. I could have it fact check me.

AI is not good at a lot of things yet

AI is probably bad at more writing tasks than it is good. This is a landmine for writers using AI because they run into a bad use case and write off AI’s ability to help.

As I said earlier, I don’t use it for writing the actual words. I specifically ask it to just report findings, not apply them (I put this in my CLAUDE.md). I don’t let it produce prose, I’m not letting it fill out an outline.

One of the most popular use cases for AI is writing summaries, but I don’t actually think this is that helpful. This is a revealed preference; I never find myself reading AI summaries. I can get more from reading a couple of interesting quotes or skimming the piece myself. Summaries seem like they are compressing information, but it feels like so much is lost in compression that it’s not worth it. If I’m looking for interesting and unique ideas, summarization basically kills those.

Beyond basic errors and typos, I don’t think it is a very good editor either (again revealed preference, as my review-blog skill often goes unused). It reflects back what you want to hear. Tweaking the prompt slightly leads it to edit in a completely different direction. Give it a sense that something is wrong and it will say something is wrong.

For example, I had a rule that intros should be short and avoid repeating themselves. Nearly every blog it reviewed, it would say I could shorten the intro, no matter how short it already was.

Because editing often means stripping things away and rewriting tighter with fewer words, AI struggles at this. It is good at writing a lot but will squirm and struggle when you want it to write less. I don’t bother asking it to make my writing more concise anymore.

But despite all these complaints, I find AI extremely valuable in the right places (it’s spiky). My writing workflow has changed more in the last 6 months than the previous 6 years and I fully expect this to continue to improve. I want to live in a world with 100x more banger blog posts and continued experimentation is essential to this.

Blog posts are a genre of their own. They are not college essays, scientific papers, letters to friends, or books. I’m sure many ideas of what makes good writing carry over from those, but I’m not pretending I know anything about editing in those areas.

Blog post editing is a distinct skill. There has been a lot written about blog posting, but not so much about blog editing. To help you get better at it (and think through how I do it), here’s what I think about when I edit blog posts.

Why write this particular blog post?

I don’t believe in blogging for the sake of blogging. Writing for the sake of writing is journaling. Simply existing is not enough.

Your post should have some purpose for existing and that means answering these questions before you write anything:

• Why should someone read this?

• Why this over something else they can read?

• What puts you in a unique place to write it?

Your post should be entertaining, informative, or revealing. You should have the authority to write it, or build that authority in the work you’ve done to write the post.

This isn’t to say your reason needs to be validated. I’ve seen many posts with a strong reason to exist do poorly, and posts with weak reasons to exist do well. You can’t predict which of your posts will be popular.

Say your best stuff first

Once you’ve decided a piece should be written, there’s not a lot of editing to do until at least some of it is written. Once written, even an outline, the next step in editing is to check if the structure makes sense.

My biggest bit of structure advice for blog posts is just say your best stuff first. Get to the point.

Many times the best bit of a post is the title. You can write a great post with just a title and almost nothing else.

• Reality has a surprising amount of detail

• A blog post is a very long and complex search query to find fascinating people and make them route interesting stuff to your inbox

• Violence and the Sacred: College as an incubator of Girardian terror

• It’s Time to Build

• Speed matters: Why working quickly is more important than it seems

The body of the post backs up the title, but it is the title that sticks with you the entire time.

The best stuff should be up front because the internet is competitive. You are fighting against every other blog, MrBeast, TikTok, Slack, everything. You need to hook people somehow.

Don’t save your best content for the conclusion. If you are writing “lessons learned,” consider just making lessons learned the piece. Weave the lessons and examples together.

Conclusions are not a good place to put information. Realistically, people get bored and drop off. Most people don’t scroll at all. If you really want to tell them something, tell them at the beginning.

Let the examples guide you

The most important part of blog posts is the examples. This is the evidence to show whatever you are saying is true.

Because anyone can write whatever they want, it can be tempting to just write down all your thoughts and opinions and assume others will agree with you, but what’s obvious to you is not obvious to others.

Instead, let the examples guide you. If you don’t have examples for something, it might not even exist. You need to do more research to find out. There have been countless times when I’ve edited a post, pushed for an example, and it hasn’t existed. This weakens not only that point, but the entire post.

The thing is that examples take a lot of work. They are the research. The good part is that blogging lets you rely on your own experiences as evidence a lot more than other mediums. Examples back up your authority, they are ways for ideas to stick in people’s minds. Readers may not understand a concept, but they can if you give them an example.

When you look at the very best bloggers, what makes them so great is their obsession with examples. The writing of James Clear, Morgan Housel, and Ryan Holiday puts the example at the core of everything they write.

Be coherent

It’s obvious, but blog posts need to make sense. They need to make sense to me as an editor if they are going to make sense for a reader. Despite knowing this, there is always a part of a post that doesn’t make sense (whether it’s my own or others). We writers are just blind to this.

Some common ways blog posts don’t make sense:

• The titles, headings, and body need to work together. For example, if the title is “What we’ve learned about X,” the headings should be lessons.

• You need to deliver on the promise. A title and intro create expectations and the rest of the body should fulfill these expectations.

• Readers get lost. Writing can go too fast. It skips over details that actually matter and suddenly some sentence, paragraph, or section doesn’t make sense. Restructuring or reordering can help with this.

An easy pushback to this is always “but you are just an amateur, an expert reader will understand all of this.” My counter is that, first, posts should be accessible to curious amateurs. This is me in lots of cases. I love reading posts about things I know very little about.

Second, your ideal expert reader doesn’t exist. Most of the time, you know and care much more about the subject than anyone else will. Your expert is also busy, tired, and lazy. Even they can benefit from you improving the coherence of your writing.

Make it skimmable

A consequence of the internet’s attention deficit is that, whether you like it or not, people will skim your post. Help them do this well. This means:

• Have descriptive and interesting headings. Don’t say “Introduction” or “Conclusion.” Look at these as a preview of the whole post.

• Texture your writing with lists, bolding, and quotes.

• Split walls of text. One or two sentence paragraphs are fine.

• Use graphics and memes.

Skimmability helps readers find what’s interesting to them in a piece and dive in further. It also helps them orient themselves in a long piece.

Simplify (once you’ve done everything else)

Many of people equate editing with fixing specific words and sentences (line editing), it’s not. Many editors focus on line edits too early, but there’s no point in doing that if the subject, structure, or examples aren’t right.

Once you get these right, my basic line editing approach is just to simplify:

• Remove repetition. Posts are usually short enough that you don’t need to repeat yourself. Repeating yourself is boring too. Sometimes this means removing entire sections.

• Simplify sentence structure. Use fewer commas. Put your subject closer to the front of the sentence.

• Avoid hedging. Blogging is opinionated. Saying “it depends,” “might,” and “sometimes” weaken your writing. You don’t need to say “I think” either, it is implied.

• Clarify what “it” is or who “they” are.

• Use less jargon, clichés, acronyms, forced humour, and regionalisms.

• Write concisely. I built a whole app to do this.

Some amount of this is fine and there are always exceptions to the rule. I don’t try to force some rigid style on all writing. People’s writing tendencies are what make them unique and interesting.

That said, simplification can be a savage process. Writing you’ve spent weeks on gone in a single keystroke. Just remember that this is in service of making what people want to say shine. You sometimes need to cut big chunks to leave only the best bits.

Labs also can’t can’t know the complete functional capability of models. People use them in ways they didn’t expect and won’t understand. This doesn’t mean most people will do this though. Most people will use models in ways they know that work.

When a new model comes out, people are happy to just be told “it’s better.” The average person doesn’t experiment or evaluate new models for new capabilities. Basically no one changes from the default models. At max, they are trying 2-3 models at a time, but there are always more models, effort levels, context engineering, and workflows they could be trying.

Instead, people rely on:

1. Vibes

2. What other people are saying (AKA marketing)

Actual model capability is formed from what the labs are doing, but adoption significantly lags the frontier. This means functional capability is largely a product of marketing.

Fable: A case study

You could not ask for a better piece of marketing than the government banning your product because it’s too powerful. Mythos felt like the most hyped model ever, but did Fable functionally live up to its expectations?

Fable as planner

The functional capability of Fable was more shaped by its constraints than its potential. Specifically, usage was guided by the fact that you could only use 50% of your Claude plan on Fable.

Because usage was constrained and the API was expensive, people focused more on making the most of that usage limit than experimenting with the limits of capabilities. It created a scarcity mindset.

It seemed (at least in my feed) that this meant using Fable as a planner and have other models execute. Dozens of popular tweets like this one captured my attention:

Fable was probably the most capable model at planning, but this doesn’t mean there weren’t other capabilities to explore. Planning makes sense given the constraints. It was even further encouraged by Anthropic through the pattern of using Fable as an advisor.

In an alternative universe where Anthropic didn’t put the 50% constraint on Fable, we wouldn’t have seen such as focus on Fable as a planner. Even if API costs were expense, people would have treated it like any other model. New models of the past were also better planners, but never didn’t have the constraint limiting their functional potential. Fable as planner became an echo chamber that drowned out other potential capabilities.

Pet problems

What does the opposite of letting marketing guide capabilities look like? An example is pet problems. Things that past models couldn’t solve that current ones can, accomplished at the individual level rather than benchmarks.

Why these? Because reality has a surprising amount of detail. Every pet problem a new model solves is a new capability.

I never saw anything about what Simon Willison calls Fable’s “relentless productivity.” This is how it goes to great lengths to accomplish tasks: building its own debugging tools, using every bit of software available to it, trying non-obvious angles when obvious ones fail. For me, this capacity seems like a newer, bigger deal, but it’s only really discovered through someone’s individual usage and reflection.

The problem is that most people don’t have pet problems to solve. They don’t keep track of what current models can’t solve and don’t even think about problems they might want models to solve. They are mostly content with their existing workflows and this limits the functional capabilities of AI.

What kind of marketing is most important to AI capabilities?

Marketing means a lot of things, but there are three areas that are most specifically relevant to the creating of AI capabilities:

Don’t underestimate the impact of creators

If you in the camp like me who thinks AI is good and want functional capabilities to expand, don’t underestimate the impact creators can have.

Creating content, reviewing models, and sharing examples pushes the frontier a little further. Diversity of use cases matters too. Software development gets a lot of focus, while an area like writing gets very little.

Although there is a lot of bad creators, ones like Every and Theo are doing a great service with their model reviews and opinions. They deserve the attention they are getting and more.

Thousands of people are evaluating the model and building their mental models of what they do because of people and companies like these. Their reviews inspire people to apply new use cases to their work and try models in new ways, expanding functional capabilities.

Model announcement as capability creation

The model release announcement is one of the best ways labs can actually expand the functional capabilities of models. This is because they get a lot of attention and people expect them to talk about capabilities.

People mostly want to know this model is better than the last one. This is done in two ways.

1. Show that this is the best model that you should be using (graphics, benchmarks)

2. Telling you how you should be using it

For example, in the Opus 5 announcement you get a paragraph like this:

An engineer at a trading firm used Opus 5 to build a market data feed for a new exchange in a single session. Previous models could not complete this task at all, even given extensive plans from the engineer. Finding no live feed to validate against, Opus 5 even built its own test harness to check that its code parsed the exchange’s data correctly.

The key phrase is “previous models could not complete this task.”

A trap a lot of model announcements make is spending too much time on benchmarks and graphs rather than examples. There is a balance between convincing people to use your model and expanding functional capabilities. Too often (like the Opus 5 announcement), it falls far on the convincing side, focusing too much on benchmarks and graphs. In the 2,821 word post, only 244 words from Anthropic and 410 words from customers were dedicated to new capabilities.

Benchmarks are good marketing, but less good at creating capabilities

Benchmarks are maybe the most important piece of model marketing. They provide a more robust and believable proof that models are getting better which is basically what people want to know. Benchmarks have even created their own brands in support of this, you think the name “humanity’s last exam” is anything other than marketing?

The other major benefit of benchmarks, at least for their creators, is that labs optimize towards them. For example, Zapier has AutomationBench which Opus 5 “topped ... without spending more tokens than prior Claude models.” You can infer that Anthropic used AutomationBench in training, which provides the most benefit to Zapier.

This is also the downside of benchmarks, models can be overfit to them by optimizing directly on them. For example, the recent Grok 4.5 model “unintentionally” included a snapshot of the Cursor codebase in training, which was likely a factor in its good performance there.

And although benchmarks show that models are getting better, it’s unclear how that translates to functional capabilities. I’ve never looked at the details of benchmarks and I doubt the average AI user has either. An improvement to a benchmark does not mean an equivalent improvement to functional capabilities.

As useful as marketing is, it’s still marketing

Marketing is meant for the masses, but how you use a model should be unique to you. Model capabilities are spiky, it might be good at the benchmarks but bad at the actual capabilities you care about. The quirks of how Theo uses models are not the same as you.

This means you should have your own way of evaluating models like pet problems that models can’t solve yet or a personal benchmark, like PelicanBench or the ability to write a banger tweet:

This is especially important if you aren’t a software developer. A lot of marketing, benchmarks, optimization is done for software developers. They do a lot of experimenting and have clear metrics to show if they succeeded. It’s not so easy to evaluate capabilities in other roles.

And if you do this, share it! Reflecting on how you use LLMs and sharing that with others is probably the best way the average person can increase functional capabilities of LLMs. It has the potential to help others more than you realize.

Why? It’s partially because startups can’t have a good history and partially because they won’t. There are structural reasons for it not existing, but even when it does, startups often treat history as a liability and historical amnesia as a competitive advantage.

Here’s how it got this way.

The book is not the medium of startup history

When you think history or canon, you think “book.” Any discussion of the startup canon is no different. It revolves around books:

• The Silicon Valley Canon: On the Paıdeía of the American Tech Elite

• The Education of the Broligarchy

• Silicon Valley Canon & History

But books are not the medium of startups. Startups invented their own mediums of communication like the tweet, blog, podcast, newsletter, and YouTube. Only the film industry is in a similar situation.

It is unsatisfying to say a vague collection of Paul Graham blogs, YC talks, How I Built This podcasts, and Naval tweets are key parts of startup history, but it is true. This is where the culture is formed from. It’s where myself and thousands of others learned about startups.

Books fail startups in multiple ways:

• They are too slow. It can take years to write a book, longer than some startups lifetimes.

• Best practices change faster than books can capture. For example, the Google algorithm might change and make SEO best practice completely out-of-date (we’re onto AEO now). It is often these cutting-edge trends that books are too slow to capture that are where the alpha is. Is a book about Google in 2011 really relevant now?

• If the idea from a book was so valuable, a startup could capture it better. Many startups are just an idea created into a company, this gives the idea leverage that a book nearly always can’t match.

Startups reinvent themselves

Startup history is only as valuable to startups as it is right now. If what has happened in the past isn’t convenient to the current narrative, it gets dropped.

An important culture of startups is “the pivot,” basically keeping the people and capital you have built but completely changing your product and positioning. Slack was a video game company. YouTube was a dating app. You don’t make any more money being consistent and failing.

A big part of that pivot is that the company history gets erased. The history gets condensed down to “we pivoted” and companies are praised for this. They never talk about the fact that their founding story was much messier than the “official” version.

There is another aspect to this, even when companies aren’t doing something as dramatic as a pivot, they still reinvent their positioning. They rewrite their history and any official version is a mythologized version of it. You see this often now in companies becoming “the AI platform for X.”

When you look at how startups talk about their history, there is an inevitability to the current position they are in. Everything was building up to this. They then pour marketing dollars into making this new history a reality. The real but inconsistent parts of the history are removed, skewing history further.

Startup history rots fast

Related to the above point, the startup industry prides itself on getting over failure quickly. A business failing and investors losing money happens. Dwelling too much on this is not a way to move towards future success. Time spent looking back is not time spent looking forward.

Many startups that failed in history just timed themselves wrong. If you cared too much about history, you might read into it that “this idea sucked” but it was really the context around it. Because of this, startup people learn that something can fail but that doesn’t mean it can’t succeed in some other context.

For example, an investor tried to convince Apoorva Mehta, the founder of Instacart, not to do the startup because a similar business, Webvan, failed. The investor literally gave him a floppy disk of the business plan trying to dissuade him. Although Webvan failed, Instacart is an $11B business 25 years later.

There was so much history saying it was impossible to build another huge social network to compete with Facebook, YouTube, Instagram, Snapchat, etc. Many were saying their network effects were too strong. And then, TikTok came along like history didn’t matter. Same thing for Google and ChatGPT.

Another way startups history rots is that much of it exists online in blog posts, tweets, news articles, emails, internal memos, Slack messages. Something as simple as an employee leaving or a website changing can destroy it. A company moving from Slack to Teams or getting acquired or even going through new security or privacy changes erase a significant amount of history. If you’re looking for a specific number:

Pew found that 38% of webpages that existed in 2013 were no longer accessible by 2023, and a quarter of pages that existed at some point between 2013 and 2023 were inaccessible by late 2023.

The startup industry is just too young

Maybe I am asking too much. The recent generations of founders are young and still in the game.

The best books about startup history are about the computer industry, which is slightly older than the startup industry. You could point to Sequoia and Kleiner Perkins founding in 1972 as the start of the modern startup era. We actually have some good history for companies like 3 generations ago like Apple and Microsoft.

The best parts of the history of other industry didn’t develop until much later. Alfred Chandler wrote the definitive history of railroads as businesses in the 1970s, almost a century after the formative period. Daniel Yergin’s The Prize came out in 1990, nearly 150 years after the beginning of the oil industry.

Many of the current and past generation companies were started by people in their 20s and 30s meaning it will take time before they are in a position to reflect on everything. Like I mentioned earlier, they probably have better things to do than writing history.

Critiques are more popular than praise

Startups cause change and disruption. They said “move fast and break things” and meant it. This is painful to many people and causes them to distrust startups at best and dislike them at worst.

Because of this, a lot of the most popular writing about a given company is critiques of it. For example:

• Careless People

• Reflecting on one very, very strange year at Uber

• Hatching Twitter

• The most popular pieces of “startup” history overall are ones about failure like Bad Blood and WeCrashed.

On basically every startup, there are huge Wikipedia sections on “Criticisms and controversies” or “Legal issues.” OpenAI’s “Governance and legal issues” section is 2,462 words, huge compared to the 4th biggest company in the world, Saudi Aramco, and their 327 word “Controversies” section.

This creates an antagonistic relationship between startups and the people who would be writing history, which makes a “positive” history harder to develop.

Why startups should care a bit more about history

Everything above makes ignoring history look rational, but “ignoring history and “being historically ignorant” are different things and often mixed up.

Startup history produces weather reports rather than laws. As law, Webvan failing means grocery delivery is dead. As weather, it tells you why Webvan failed and what changed since.

The edge history provides becomes more valuable as execution becomes cheaper thanks to AI. History can give you a stronger idea of what to build. Much of the “taste” people talk about is built from history.

History is also part of how an industry earns society’s trust, which continues to be a challenge for startups, especially recent AI startups. Creating positive history creates a positive reputation. Beyond making business easier, this also helps attract talented people to the industry.

The good news is that this is improving. There’s more pro-startup media than ever: Founders, Acquired, TBPN, Substack, Stripe Press, YC. Today’s coverage becomes tomorrow’s history. Do it well enough and the next generation won’t have this complaint.

• At their core, Vercel is a frontend hosting platform, but there are a bunch of similar ones like Netlify, DigitalOcean, AWS along with no-code tools like WordPress, Webflow, and Bubble. How does Vercel manage to stand out? They weren’t the first ones to make deploying and hosting web apps easy.

• Even if they did stand out compared to the last generation of competitors, how do they stand out against the next generation? The AI app builders like Lovable, Bolt.new, and Replit?

• The answer is probably Next.js which is cool, but it’s a big expense, a lot of work to evangelize and convince people to use, and there are plenty of other frontend frameworks that work well enough. Why did this beat React + your state management of choice, Vue, or any of the other options?

• Also, they keep acquiring (or creating) open source projects like NuxtLabs, Tremor, and Turbopack but these seem to just increase expenses and only have marginal connection to their core product (hosting) that makes them money. Why?

As someone who uses Vercel every day, this was a lot of question marks. It wasn’t until recently that it finally clicked: Vercel is the Red Hat of frontend.

What is Red Hat?

Red Hat is a software company that builds and provides open source software. They are largely associated with Linux having created one of the first distributions in Red Hat Linux in 1993.

Since they couldn’t sell Linux licenses (it is free), their version takes thousands of packages, freezes the code, and creates an enterprise-focused version. This version provides the long-term stability, security, and compatibility enterprises wanted, along with a promise of support (for a subscription).

This enabled enterprises like banks, airlines, governments, and telecom companies to get the functionality of Linux with less risk.

Beyond its business model, Red Hat has built capabilities around developing and supporting open source communities. It is one of the top contributors to Linux and has catalyzed several other open source projects.

For example, Red Hat has acquired and developed open source projects for automation (Ansible) and orchestration (OpenShift). This was especially important as the world moved from on-prem to virtualization, cloud, and containers.

If you’re wondering if this path is a good one, yes. Red Hat was acquired by IBM in 2018 for $34 billion. It had annual revenue of $6.4b last year, employs 19,000 people, and is a core part of IBM’s business now.

The trouble with the Red Hat model

In 2014, a16z partner Peter Levine wrote that there would never be another Red Hat. In 2015, Dan Woods added to this saying:

“Will any company ever get to $1 billion or even $500 million in revenue from open source subscriptions and have a chance of being profitable? To this the answer is no. There will never be another Red Hat.”

Between the two of them, their argument was:

• No other open source company could have a product as widely applicable.

• The support-driven business model of Red Hat sucks and is unable to support adequate funding of ongoing investments. Anyone else trying to sell services for open source software won’t make it.

• Linux is a unique piece of software that companies didn’t need to be convinced to buy and Red Hat was in a uniquely good position to sell because of their first-mover advantage and reputation.

Although these are all fair points, their argument reveals the problem with “saying never” about any situation. History may not repeat, but it does rhyme. In this case, Red Hat rhymes with Vercel.

How Vercel rhymes with Red Hat

Far from “another Red Hat never existing,” Vercel has seemingly followed a similar path intentionally or not. Both:

• Built an enterprise version of a popular open source project that has strong adoption. For Vercel, Next.js is basically an enterprise version of React, which is the most popular frontend framework.

• Monetized it by providing the best services around that open source product. Developed the expertise to provide services for that open source technology in the best possible way.

• Bolstered their reputation and improved their offering by developing and acquiring other open source products.

Although Peter Levine was right that the “support” business model probably sucked, this doesn’t mean an open source product can’t be a successful one. If you adjust it slightly to be “adjacent services” like hosting, it seemingly works well. Vercel, with a reported $200M in ARR, is well on its way to the “impossible” $500M amount Red Hat hit.

Looking at Vercel through the lens of Red Hat cleared up a lot of my confusion about why it was successful.

1. None of the other frontend hosting platforms or AI app builders take enterprise as seriously as Vercel does. Intentionally or not, others appeal much more to vibe coders and hobbyists.

2. Although builders of all kinds use Next.js, it can be seen primarily as a product built for enterprises. They make this pretty obvious with their focus on performance and security on their homepage.

3. Next.js builds on React’s popularity and Vercel doesn’t need to carry the entire development load. They benefit a huge amount from all the investment in the React ecosystem (from Meta and others). Vercel, through Next.js, can capture a significant amount of benefits from this.

4. The acquisition and development of other open source projects is because it improves their reputation for caring about performance and developer experience while strengthening their core offering of “enterprise React.” For example, Turbopack is all about improving bundle speeds for large (AKA enterprise) apps.

A collision course in the age of AI?

The similarities don’t end in both companies’ histories. They both face similar circumstances in the age of AI. Simply, they both care a lot about it, and it shows with all the AI buzzwords they plaster all over their website.

Neither Vercel nor Red Hat are in a position to develop their own models (although Red Hat’s parent company, IBM, is) and doing this doesn’t fit with their enterprise open source model anyway. Instead:

• Vercel aims to become the “AI Cloud.” They explain this on their site as helping companies “deploy AI in seconds. Access all major models through a single, unified interface and shared AI credit wallet.”

• Red Hat aims to let enterprises “tune small models with enterprise-relevant data, and develop and deploy AI solutions across hybrid cloud environments.”

Basically, they both are aiming to be the middle layer between model providers and enterprise apps. This makes sense as enterprise apps are already hosted on their respective platforms, making it easier for them to move up the value chain.

What this also means is that they are on a collision course. This middle AI layer is going to end up somewhere and it’s going to be valuable. Vercel, Red Hat, and many other companies (like OpenRouter) are battling for it. Right now, there likely isn’t much overlap, but as the contracts get larger, there will be. It will be interesting to see how a potential battle between such similar companies plays out.

Something that caught my attention recently is Nick Bare, Bare Performance Nutrition (BPN), and specifically, their recent Go One More Backyard Ultra. I’ve been following Nick Bare for a while and he has influenced my training (more runs), but the recent Go One More Backyard Ultra seemed like a culmination of their efforts and clarified what’s so smart about them.

They sell an identity, work with influencers to model that identity, and then channel attention towards a defining event. This breaks through the noise and competition of the fitness and supplement space and enables them to build a moat in the massively competitive supplements industry.

Here’s how they do it:

1. Sell an Identity Rather Than a (Supplement) Product

Supplements are one of the most competitive consumer markets. There are tons of companies selling products very close to the raw ingredients, basically entirely differentiated by brand and marketing.

One of the areas of differentiation they do find is features. Some products are tasty, vegan, natural, made from different (clean) ingredients, seed oil free, promote gut health, and countless others. Convincing someone that your feature is better than others is a lot of work and there seemingly aren’t many new “features” left untapped.

The other main way of selling supplements is influencers and endorsements. This has been the case for about as long as people have been famous. All kinds of people, from bodybuilders to reality show stars to podcasters, have their own brand of supplements.

This is an indirect way to sell an identity. The promise is that if you buy a supplement, you’ll be more like the influencer you admire.

BPN is the next evolution of this. They take a more direct approach and sell the identity itself.

I see that identity broadly as the “hybrid athlete.” This is someone who lifts (heavy) weights and runs far. I would say the expanded definition also includes people who challenge themselves to do extreme physical activities (”do hard things”, “go one more”).

Selling the hybrid athlete identity is core to what BPN does and their supplements are along for the ride. It turns out, when done well, identity is a much better differentiator than the abstract features of supplements. You don’t need to understand what the supplements do, you just need to aspire to be a hybrid athlete.

The added brilliance of the hybrid athlete identity is that it expands the market. BPN can provide products to both runners and lifters, where other supplements only really appeal to one or the other.

2. Provide Models for Hybrid Athletes

In a classic influencer structure, Nick Bare would act as the model of the hybrid athlete lifestyle (and he does), but BPN takes it a step further. They include and promote other athletes in multiple ways to create a sort of “hybrid athlete cinematic universe.” They do this in multiple ways:

• Sponsoring athletes who are either hybrid athletes or “do hard things” like ultra runner Sally McRae or powerlifter Russel Orhii.

• Featuring other athletes in Nick’s videos like Jake Dearden, Dan Churchill, and Ian Fonz.

• Having a separate BPN YouTube channel that creates content for other athletes (and performs shockingly well for a secondary channel).

• Being the model and providing a blueprint for content in the form of “challenges” like fast marathons, HYROX, The Speed Project, and many more.

This provides a multitude of benefits:

1. There is constantly content about being a hybrid athlete, validating the lifestyle. The definition (and reach) expands further and further. People who want more hybrid athlete content, get more.

2. Rather than people following an individual like Nick, they feel like they are part of a movement. This again increases the potential reach.

3. Content creators can also ride this wave by identifying as hybrid athletes and following the template Nick provides. Searching “hybrid athlete” uncovers many others with a similar style and structure of video. This ultimately leads to more hybrid athlete content.

All of this powers the hybrid athlete cinematic universe (and sells more supplements along the way).

3. Channel Attention Into Crossover Events

Like any good cinematic universe, crossover events are what really excite people. When channeled, these break through the noise and reach people they never would have reached.

Although Nick sometimes includes other athletes in his videos, hybrid athletes train together, and some may attend the same events, getting the full power of the crossover requires manufacturing. BPN did just this with the recent Go One More Backyard Ultra.

The race featured 125 runners from around the globe completing a 4.2 mile loop every hour on the hour until one person remained. Yes, this means people ran hundreds of miles over multiple days. The race actually ended in a tie after 2 days and 235 miles, finishing in a huge storm.

Multiple members of the hybrid athlete cinematic universe like Truett Hanes, Lucy Davis, and Jonny Davies participated as well as people adjacent like Zach Pogrob. Both the participants and the brand got content, attention, promotion, and followers both leading up to the event and coming out of it.

But ultimately, BPN is a business and they care a lot about the money. So how did they follow up this burst of attention? The launch of a new product, a pre-workout for endurance athletes. This perfectly aligns with the event and with hybrid athletes more broadly.

This is how a brand becomes a moat. Few are going to have the patience to build up a brand over years, attract such a dedicated following (who will run 100s of miles), and then coordinate a huge event like this (and do it repeatedly). It shows that the BPN athletes aren’t the only ones who are committed to doing hard things, the business is too.

This is a bad idea, and simplicity is a bad goal. Products that seemingly succeed because of simplicity are hiding their real advantage. Keeping this advantage hidden is in their interest; it creates a false model to follow and fewer real competitors.

Understanding the problem with “X but simpler” and avoiding it as a tarpit will help a product find ways to build more sustainable value.

The problem with “X but simpler”

There are plenty of truisms about simplicity, but taking too many lessons from them is risky. Stealing a phrase from Sandi Metz, simplicity is the wrong abstraction. Here’s why:

1. Simple products are cheap

LLMs have made it easier than ever to develop a simpler version of every product. Anyone can vibe code their way to a working prototype in a weekend.2

This means any moat a simple product will have will quickly evaporate. There are plenty of copycats out there ready to eat your lunch by competing with you and charging less.

You see it in the indiehacker world all the time. Someone achieves success with a simple piece of software and multiple copycats pop up.

This also means big companies (where the $$$ is) can build a simple version of a product themselves too.

2. It’s a short term strategy

“X but simpler” is not a long term strategy, it’s a dead end.

Once you build out the simple use case, what are you supposed to do? Just leave it?

Even the simplest of products with users will get feature requests. If you don’t develop those features and the users mature, they will just go to the more mature “complicated” product. You have basically done all the hard work of attracting a user, getting them familiar with the product’s concept, but not capturing the lifetime value. Sowing without reaping.

The “complex” products you are basing yours on were once simple too. The complexity was added to address a variety of use cases, and eventually requirements like security, regulation, functionality. A good product often becomes what its customers want it to be, and when they have many customers, that means a lot of features.

3. It’s a worse business

Often what people are saying when they say “X but simpler” is actually “X but a worse business.”

One thing startups (or individuals) can do that big companies can’t is have no business model. They have the excuse that they need to get the product working before charging for it. They also don’t need to pay anyone if it’s something they are working on themselves.

This means a lot of work can go into a product before validating that it works as a business. When it comes to testing this, simple products are often in a tough spot:

1. They can’t charge more money for them than more expensive products. They’d need to be a better, differentiated product for that.

2. They can’t capture the larger customers. The advanced features larger, more valuable customers want might not exist in your product.

I feel like indiehackers and “business thought leaders” are largely to blame for this. These people might be known for the products they build, but actually make money from consulting, courses, or templates. For example, Marc Lou makes ~3% of this revenue from the products he’s built and 97% from a course and template.

4. Simple does not solve distribution

Simplicity is a hook, not a long-term distribution plan.

When you focus on simplicity, you are permanently anchoring yourself to the product you are a simpler version of. Your market becomes a small portion of their market. Eventually, you will run out of people who are simply interested in a simpler version of the existing product. You have not solved for distribution yourself.

This means you’ll always be competing with them. They will have:

• More margin and resources to use in their battle against you.

• More capacity to improve their product (including making certain parts, like onboarding, simpler).

Simple apps also have a harder time with retention. It might be easy to activate someone, but because you provide less value, it is trickier to keep them around. For example, there have been a massive number of Twitter clones relying on “simpler” to attract users, but the users they do attract haven’t stuck around. The simple hook is not enough to create the network effect needed.

What should you do instead?

As Sandi Metz says in The Wrong Abstraction, the “fastest way forward is back.” You need to reconsider why you chose simplicity in the first place, deeper thinking may reveal a better path.

I don’t have the secret method to “build a successful product,” but common alternatives include:

X but targeted

Stripe is payments for engineers. Canva is Photoshop but for social media peeps and casual designers rather than hardcore editors. Pitch is PowerPoint for startups. Substack is Mailchimp for writers. Signal is like Whatsapp with better security, same with ProtonMail vs Gmail.

At some point, all of these could be considered simpler versions of the companies they are competing with, but their main frame is as an alternative that targets a specific subset of users.

X but with a different business model

Often this means cheaper. When so many companies put pricing behind a “quick call,” being transparent with pricing is a breath of fresh air. When so many tools are built exclusively for enterprises, building for small businesses can gain a lot of traction.

PostHog, Supabase, Coolify, Netdata. Cal.com, and Hubspot are all examples of this. Products that already exist but with a completely different business model backing them.

X but opinionated

These products might be simpler, but the source of that simplicity comes from a careful understanding of the structure and purpose of a product.

Linear might be a “simpler” version of Jira but that’s built out of a deep understanding of user needs along with strong opinions on how to address those needs. It focuses on the functionality and corner cases high-performance product teams want.

Fey, Superhuman, Basecamp, and Hey are all examples of this. It might be better to think of them as “more elegant” products rather than simpler ones.

Simple is a false

None of these are as easy as simple, but this is my point. Bragging about the simplicity of a product often reveals that the creators don’t understand what makes it unique or different. Eventually, this leads to demands for complexity, poor business results, waning retention, and a lack of structure to deal with it.

Just like any other industry, the get rich quick scheme of building a simpler product isn’t as real as it seems.

1. She speaks Spanish with her family and many of her friends. Not knowing Spanish would mean missing out on a significant number of conversations across our life.

2. How can you fully understand and connect with someone if you don’t know at least some of their native language?

Like many, I immediately started THE modern way to learn a language: Duolingo. After doing hundreds of lessons and building up a healthy streak (maybe a year, I forget), I reflected on what I learned: functionally nothing.

Being a part of Spanish conversations revealed I didn’t actually know any Spanish. Duolingo tricked me like it has with many other people. You feel like you are learning a language, but when actually go to speak it, you are only slightly better off. This is a problem.

Duolingo doesn’t care if you learn a language

Luis Von Ahn, the founder and CEO of Duolingo, in an interview with the Verge, revealed the company’s priorities:

Do you think that there’s a conflict between gamification and engagement — the things that you’re historically successful at — and education?

Yes, there is.

How do you manage that conflict?

Very easily. Always go with engagement.

Really?

Yes.

I mean, presumably, you’ve heard both sides of the argument. Why have you made this decision?

I’ll give you many arguments, but the one that works the most is this: It doesn’t matter how effective you are. You can’t teach somebody who’s not there.

When the CEO of the company says that he does not care about you learning a language, you should listen to him. A similarly concerning sign is how much they talk about building an engaging product rather than helping people learn a language.

Instead of thinking of Duolingo as a language learning app, you should think about it as a game. It’s like crosswords or Wordle. It requires you to use your brain a bit, but it mostly focuses on fun. The seriousness required for you to learn a language is not there.

I don’t blame them for this. Their incentives as a business are misaligned with you. Like a dating app, they make money by ensuring you make progress but don’t succeed in your ultimate goal. If they were churning out fluent speakers, they’d have many fewer power users willing to pay for their product.

What language learning requires

It is only after trying other methods of language learning I realized how much Duolingo lets you down. The requirements to learn a language are nearly the opposite of what Duolingo provides.

It takes a lot of time

Duolingo is meant to be completed in 5 minute chunks. I often found myself repeatedly doing the bare minimum before bed to keep my streak.

In reality, it takes hundreds of hours (at a minimum) to learn a language. At 5 minutes a day, a year of Duolingo only nets you 30 hours of language learning time. Even over the course of years, this is not nearly enough to learn a language.

As an opposite example, take Isaak Freeman learning Chinese. He did 1500 hours of practice in less than a year.1 This is ~4 hours a day on average.

Duolingo may claim it is about “quality” of hours, but others have tested this too. Take the story of Steven Sacco, a retired language professor:

Intrigued by Duolingo’s “34 hours” claim, he put it to the test, studying Swedish on the app for a total of 300 hours. (Most introductory university Swedish courses amount to 150 or so hours of coursework, he reasoned; 300 would be more than safe.) He then convinced the professor of UCLA’s Elementary Swedish course to let him take the final exam. He got an F.

You need to push your limits

Duolingo is meant to be engaging and not feel like work. They know if it feels hard, you will quit. This means you can’t really fail on Duolingo.

In real life, you can definitely fail. That is the whole point. Someone speaks to you in another language, you forget the words, and choke up. You derail an entire conversation trying to contribute. You use the wrong word or verb tense and people laugh at you. This is what reality looks like.

Von Ahn recognizes this and says:

The other thing is, somewhere between 80 and 90 percent of language learners don’t want to talk to another human. They may tell you they do, but they don’t. It’s because when you’re learning a language, you’re pretty shy about it, and only the extreme extroverts are okay talking to a stranger on video in a language that they’re not very good at. The majority of people won’t do it.

To learn a language is both knowing the words and being able to use them. At best, Duolingo only teaches you one of these. You’ll eventually need to do the second, and being able to do that is a muscle.

It requires a lot of input

The theory of language learning I’ve come around to is comprehensible input. The summary is “that linguistic competence is only advanced when language is subconsciously acquired, and that conscious learning cannot be used as a source of spontaneous language production.” You should be working at the edge of your competency to learn new things.

Duolingo almost never overwhelms you with new words. It seemed like I was learning the word for apple for months. You spend a huge amount of time reviewing past words rather than pushing the limits of your comprehension (again, because this is hard).

Even if Duolingo teaches you a language, it does it at way too slow a rate.

Duolingo isn’t real

This is one of the most common complaints about Duolingo: that it doesn’t reflect how people talk in real life. For example, I’ve basically never heard someone say “adios” in Spanish, “nos vemos” or “chao,” sure. Even if Duolingo wanted to teach you “nos vemos,” it doesn’t translate well into English (“we see each other”).

Like many language learning courses, Duolingo starts from the start, teaching you how to greet people, introduce yourself, and common words. Unfortunately, a majority of conversations do not start from the start. They are made of “the middle parts” and all the verbs and different conjugations that come with it.

One of my favourite language learning experiences so far was an audio course called Language Transfer. What made it special was that it focused on the ability to recall a language, rather than memorize it. It did this by providing memory hooks, like voyage → I go → “voy,” and context on the linguistics of a language. Duolingo rarely, if ever, does this.

Language Transfer also focused on the important part of learning a language, listening and speaking. This is, again, the opposite of Duolingo. You may think you are learning how to listen, but Duolingo is so far off from real listening experiences that it is worthless.

If not Duolingo, then what?

Now when I meet people who use Duolingo, I try to convince them of one thing: stop using it. I realize this is sort of a social faux pas, live and let live or whatever, but I (obviously) feel strongly about this (enough to write a long blog post).

Duolingo is a part of an epidemic of unseriousness. When I think of a Duolingo learner, I think of a 20 something who likes to travel and does Duolingo on their way to work in an effort to become a more worldly person. “Una cerveza por favor” type stuff. Never works unfortunately.

No matter how long a streak someone has had, I’ve never met anyone who’s said Duolingo has taught them a language.2 If this isn’t proof, I don’t know what is.

This would be fine in isolation, but Duolingo might be the single most popular way to learn a language in the world.

A fun fact: there are more people learning languages on Duolingo in the US than in all US high schools combined. This is true in most countries in the world. We teach languages to more people than the public school systems.

There are multiple reasons why this is a big problem:

1. All this unseriousness is bad for your soul. It weighs you down. You have to, sort of embarrassingly, repeatedly say “I’m learning a language, but I only know a little.” You are admitting to other people you are not serious. This is demotivating and has knock on effects for the rest of your life.

2. I don’t know if you read the news, but AI is coming. AI will be (is) good at doing a lot of things humans are, especially when people are unserious about it.3 The AI world will reward people who do more than “just show up” because AI is really good at “just showing up.”

The solution to these problems: Be serious.4 Life is short. If you want to learn a language, put in the effort to actually do it. If you don’t, be honest with yourself and do something else.

I would consider a banger a creation that is particularly viral, unique, engaging, and novel. Neither these AI thought leaders nor the general banger writing population have become insanely prolific thanks to LLMs. They haven’t turned everyone into Michael Crichton:

Michael Crichton spent four years at Harvard Medical School, and what does he have to show for it? Seven novels, a movie, several more manuscripts and screenplays and a fast‐paced, provocative book of non‐fiction on the state of the American hospital, which was published last week.

On the other hand, LLMs have both:

1. Made it significantly easier to produce a large number of words about a subject quickly.

2. Increased the rate of bangers in other areas, like software development.

You would think writing, which is core to what LLMs do, would be impacted more. Why hasn’t it?

LLMs are having a huge impact on coding

This lack of written bangers reveals itself when compared to coding. There are many more people saying “I built this in 3 hours” than “I wrote this in 3 hours.” Some examples:

• The world's shortest hackathon (2 hours) hosted by Vercel and Nvidia

• Everything I built with Claude Artifacts this week

• Recreating complex canvas animations in 15 minutes.

• Pixel perfectly rebuilding a UI that 800+ engineers couldn’t nail.

• People recreating entire UIs and apps with v0, Bolt.new, Claude Artifacts, aider.chat.

This personal site that looks like a receipt is a perfect example of a banger coding project. Its viral, unique, engaging, and novel. Thanks to LLMs, it was built in a few hours.

Beyond bangers, millions of people have used LLMs to become better developers. GitHub found that developers complete task 55% faster when they use LLMs. This means a lot more code and projects being written and completed, and a lot more potential for bangers.

Why aren’t LLMs increasing the number of banger posts?

When compared with coding, writing just hasn’t kept up. There’s many potential reasons why:

1. Writing is sand, code is Lego

Writing is like sand. It starts as nothing and will crumble without support and careful planning. Code is like Lego, it gives you the structure and pieces you need to make your vision a reality.

Because of the structure and reusability of code, your vision of a banger is a lot closer to reality for coding projects than it is for writing ones. The LLMs has a better understanding of the pieces available to make that vision real. Reusing writing to the same degree would be considered plagiarism.

The reusability and structure also means there is less to go wrong. Although the quality of code may vary as much for code as it does for writing, end users don’t see the code, they only see the experience it produces. For writing, the words are the experience. Poor quality is much more obvious.

Ultimately, LLMs can produce both passable writing and code, but passable code is much more likely to create a banger experience than passable writing.

2. Writers haven’t figured out how to use LLMs

There is a cultural layer of using LLMs writers haven’t figured out. There are plenty of good blog posts about how developers are using LLMs, but none (I can remember) about writers using them. None of the writers I admire have written about how it has changed the way they do things.

For my workflow, LLMs basically function as a superpowered Google search. I have a question or an idea I want to explore and use the LLM to explore it. Going further, I find they work well to interact and ask questions of documents (or code) using projects. It has made me a better writer, but it hasn’t helped me produce 10x more bangers.

This is similar to someone I consider to be a counter point to this whole post: Dwarkesh Podcast.1 He leverages LLMs heavily and built one of the best and most important podcasts that churns out banger after banger.

While asking Claude questions, Dwarkesh uses the LLM as a sounding board to validate and refine his grasp of his guest’s field of expertise. This back-and-forth with Claude helps him discern if he’s “found a blind spot in their thinking” or if he’s simply “confused by their ideas.”

Why haven’t we heard similar things from writers? Probably because the truly LLM-empowered writer doesn’t exist yet.

3. The tools aren’t there yet

Tools often act as productized versions of workflows people have figured out. Because writers haven’t fully figured out how to use LLMs to produce great writing, there’s little for LLMs to productize.

There are many tools for writers, like Type.ai and Notion AI, but these rarely are part of the workflow of banger producing writers. These tools seem to be used more for simple tasks like formatting, grammar-checking, and synonyms. At best, most writers seem to just use ChatGPT and Claude.

Developers are constantly shaping their tools to improve their effectiveness. This is something we’ve seen yet again with all of the LLM-powered, coding-related products. LLMs quickly became integrated with the products, like IDEs, they were already using. Unfortunately, this is not the same (yet) for writers.

As proof of this, look at the testimonials of writing and coding related tools. I rarely see praise for writing related tools on social, and the testimonials on their site are from people and companies I’ve never heard of. A tool like Cursor, on the other hand, regularly gets praised for both helping non-technical people code and seasoned engineers be more productive. Their site is filled with people and companies I’ve actually heard of.

4. LLMs make you smarter but that doesn’t make you a better writer

LLMs can teach you anything. They are like a tutor with infinite patience that can help you understand hard concepts, ideas, books, languages, and innovations. Knowing these things does not necessarily translate into your ability to write about them well.

Have you ever read an academic paper? These are written by the smartest people in the world about the most important ideas in their lives. Theoretically, these should be bangers, but a vast majority are boring, dense, and basically unreadable unless it is your job. This shows that being smart is not all matters for banger production.

The idea that getting smarter does not lead to more bangers is also a bearish one for LLM progress. A vague “smartness” seems to be the main focus when it comes to LLMs. Sadly for me, there’s no LLM developer optimizing for a model that writes banger blog posts.

Maybe Facebook, TikTok, or X (sigh) will figure this out. After all, creating an algorithm that optimizes for engagement is core to each of their businesses. They are in a position to combine this engagement focus with LLMs to create my imagined banger blog post wireheading machine.

5. Distribution is ruined (and so is my banger radar)

The last potential answer is that the world is producing more banger blog posts, I’m just not seeing them. This could be caused by LLMs or just be coincidental.

Twitter used to be the go-to place where I saw posts and could judge if they were bangers, but the algorithm changes have ruined this. X now downranks links, especially to Substack, where many bangers live (and has made a large impact to the banger blog post numbers 🙏). Maybe I’m nostalgic for a time where a blog post would take over Twitter for days at a time, but that’s not realistic anymore.

Even if Twitter’s algorithm didn’t change, everyone isn’t in one place anymore. They’re on Substack, Beehiiv, Ghost, Mastodon, Bluesky, Threads, TikTok, and many more. A banger doesn’t capture everyone’s attention like it used to.

The spirit of writing will always matter

Like Homer invoking the muse, there is some creative spirit of bangers that LLMs have not captured. The areas LLMs aren’t good at might reveal where this creative spirit lies: topic selection, audience understanding, context, sources, and structure. Avoiding being automated away means tapping into your creative spirit and improving your skills in these areas.

There might come a day where LLMs do manage to create banger after banger, but that will still require someone to curate and lead them (taste matters, as tech bros like to say). This same creative spirit will still matter in this future.

Likely because, from a financial perspective, service makes software businesses look uglier. Service isn’t zero marginal cost. “What about my margins bro? Can’t you just automate it away?”

As much as we try to squeeze service into data, APIs, and LLMs, the human part of it still matters. After all, it is humans creating the software and humans using it. Service as a differentiator in software is dramatically under explored, especially when compared to the role it plays in other industries.

SaaS companies are like restaurants

Another industry with a similar split between product and service is restaurants.

People think about restaurants in terms of food, but service has a massive impact (but hidden) impact on people’s experience. Often, the greatest restaurant experiences are the product of amazing service.

In Will Guidara’s “Unreasonable Hospitality,” the importance and magic of service is particularly clear. He was the general manager and co-owner of Eleven Madison Park, a restaurant that earned 3 Michelin stars and ranked first in The World's 50 Best Restaurants list. He writes about what it takes to be the most hospitable and service-oriented restaurant on earth. His ideas have a lot of potential to transfer across domains and apply to software.

Whatever you do for a living, you can choose to be in the hospitality business. Intention means every decision, from the most obviously significant to the seemingly mundane, matters. - Will Guidara

A similar pattern exists between software and restaurants. The abundance of bad service makes good service seem like a miracle, but good service still isn’t world-class service. You must innovate to reach that point, Guidara did that, and software businesses can learn from his path.

The service in software status quo

The type of restaurant service that was the status quo when Guidara took over Eleven Madison Park was what you think of a luxury restaurant to be: a luxurious setting, formal attire, perfectly set tables, and knowledgeable staff.

Service in software, at its best, is in a relatively similar position. Responsive support, helpful salespeople, onboarding or implementation guides to help you get started. In a world where so much bad support exists, these are great, and many companies are missing them, but are they really world-class service? Is it magical?

To be the best in the world, to have service as a differentiator, requires moving beyond this. Guidara moves to the definition of hospitality as not perfect service, but creating exceptional moments of connection and service.

Service is about people and their experiences

Everyone goes through life with an invisible sign hanging around his or her neck reading, “make me feel important.” - Danny Meyer

Does software ever make individuals feel important? Because everyone gets the same experience, making someone feel important is almost the opposite of what happens. Software is built for everyone and anyone to use, instead of being built to make people feel special. It is impersonal. There is no sign saying “this is for you.”

The impact of impersonal software continues to play out. To use another restaurant example, Howard Schultz, long-time CEO of Starbucks has spoken about how the Starbucks app and ordering ahead has negatively impacted the customer relationship and business.

No software starts by being impersonal. Many people build something for themselves or their friends. Much of the most popular startup advice focuses on people. What does “talk to users” or “build something people want” sound like to you? It could translate to “focus more on service.”

As companies and the software they create get larger, it often becomes more impersonal. The average employee gets further away from users. Rather than build for specific people, you build for some blob known as a “persona” or “target market.” The service aspect of software as a service is lost.

Guidara provides a solution for breaking this trend.

Unreasonability to the rescue

One of the best parts of “Unreasonable Hospitality” was around “legends.” Legends were moments or experiences Eleven Madison Park created for guests that were above and beyond.

One example was a group of foodies who had been to all of the most fancy restaurants in New York. Guidara overheard them say the only thing they didn’t try was a hot dog. So Guidara went out, bought a street hot dog, and convinced the chef to plate it fancy. When they were served a hot dog in the middle of their luxurious meal, they freaked out (in a good way).

The staff at Eleven Madison Park tried to pull these off as often as they could. They even had staff, with the title “Dreamweaver” dedicated to it. These moments helped create a world-class service culture that pushed them to be named the best restaurant in the world.

The beauty of this is that it isn’t exclusive to restaurants. “Legends” (or whatever you want to call them) are acts of radical service that break a natural drift towards the impersonal.

Examples of legends in software

You might say “that isn’t possible in software, it is all hard-coded” but that’s not true. Of course, Paul Graham has said this better in Do things that don’t scale, but examples of legends in software include:

• Shipping a fix within minutes of someone reporting an issue.

• Shipping a feature tailor-made for a specific user.

• Make your website, content, onboarding, and transaction emails unique, funny, clever, and useful.

• Adding a banner to the account when they do something great

More specifically, the companies behind software have even more room to pull off legends:

• Send pizzas like Antimetal or champagne like Brex

• Send hand-written notes like Wufoo.

• Host dinners, events, and conferences for your users, customers, and community.

• Do a Collison install and completely hand holding users through installation.

• Recruiting users IRL, door-to-door.

• Permissionless co-marketing. Shouting out or featuring other products.

Service in software is often about giving them what they want. Legends go above and beyond to give people experiences they didn’t know they wanted. It breaks both you and them out of the impersonal experience of their day-to-day interaction. Software can create “wow” moments, you just need to care about doing it.

Distributing legend responsibility

The connotation of “do things that don’t scale” in software is with founders, not employees. Similarly, unreasonable hospitality seems difficult to scale, but might not be. It’s about solving two blockers:

1. Budget. They cost money and time, but probably less than you expect. Guidara says: “Manage 95 of your business down to the penny; spend the last 5 percent foolishly. The last 5 percent has an outsize impact on guest experience.”

2. Ownership. The ability to do it needs to be distributed. Because interactions happen everywhere, a good idea can come from anywhere. Everyone needs the ability and permission to execute them.

3. Culture. Legends became part of the lore at Eleven Madison Park. They were chronicled, people talked about them. Pulling one off makes you want to do it again and encourages everyone to do the same.

Guidara notes that a great part about executing legends is that they can be repeated. Many of the examples above aren’t unique for one user. Although a legend might not be unique to you, it is unique to the user. You can send merch to 20 users, but for each of them, it is the one time they’ve got merch.

Cool, but what about my margins…

Again, looking at the examples, you’ll see a lot of tangible and massive successes that come with pulling legends off (Antimetal did $1M in revenue on $15k in pizza). It’s almost like caring about your users is good for business. The opportunities are real, but it takes a culture of unreasonable hospitality to capitalize on them.

There is a massive gap between what it takes to write and what it takes to publish. It requires a different way of thinking, one they don’t teach you at school. After all, you learn “how to write” not “how to publish.”

Before you get mad at me:

1. I define publishing as sharing with others. Publishing an internal memo or on a blog no one reads is still publishing to me.

2. Writing without publishing is fine, just call it what it is: journaling, researching, procrastinating 😉.

Fostering a mindset where publishing happens

The difference between writing and publishing can be summed up by urgency. People who publish are in a rush to do so. There is a reason why people are so productive under a deadline. A deadline generates urgency and motivation.

Your motivation to publish any piece wanes over time, especially when you don’t have a deadline. Motivation will eventually disappear completely. When this happens, your draft ends up sunk with the collection of unpublished pieces all writers have.

To avoid this, monitor your motivation, keep your urgency up, and go as far as you can right now. Can you finish an outline, draft, or edit today? Better, can you publish today? You don’t know what tomorrow will look like. It is full of new obstacles and ideas ready to drain away your motivation for your current draft.

Prioritizing publishing means cutting scope

Scope weighs writers down. The gap between what you planned for the post and what is written demotivates you. It is a mountain you need to climb. What no one tells you is that you can choose to climb a smaller mountain.

Reject your expectations of what a “published post” is supposed to look like. It is completely made up. Be less fancy. Be less clever. Use fewer, simpler words. Word count doesn’t exist in the real world. No one is marking your work.

Smaller, published posts are better than larger, unpublished ones. Not every piece needs to be the best, most important thing you've ever written. There is tons of wisdom in the small and specific. You can always update, rewrite, or expand on it later (but you’ll be shocked by how infrequently this happens).

Focus on the most important point and cut anything that doesn’t support it. Keep the main thing the main thing.

Feedback should work towards publishing

Feedback is a core part of every “professional” writer’s workflow. The problem with feedback is that it rarely works towards the goal of publishing. Feedback is often someone applying their ideas of “good” to your piece. If you don’t completely trust this opinion, it hinders your ability to publish.

• Bad feedback muddles your writing with irrelevant ideas, focuses on nitpicks and opinions, feels like weights being added to your post

• Good feedback clarifies your writing towards publishing, focuses on the core ideas and structure, and feels like a weight being lifted off your post

Feedback is not a miracle worker. No one can publish a piece for you. If you are unmotivated to continue working on a piece, feedback is probably not going to help.

The paradox of publishing

A lot of resistance to publishing happens at the final step: deciding you’re done and pressing “publish.” There are two paradoxical points you can use to solve this:

1. No one will remember what you wrote. People can barely remember something they read five minutes ago. We’re all dust in the end. A dud post will be forgotten but frees you to work on your next one.

2. Your writing can have a massive impact. Complete strangers tell me my writing impacts them. I have no idea how they found my writing, but they wouldn’t have found it if I didn’t publish. Writing has changed people’s lives and created billions in enterprise value, but that doesn’t happen without them publishing it.

Publishing does require courage, but remember that others published who were less courageous and at higher risk of ruin. What they published shaped the world we have today. Real writers publish.

At one point, this culture (and the success it created) was thought to only exist in San Francisco and to a limited amount in other “startup hubs” across the world. Many people still believe in this “SF or bust” mindset (they are largely incentivized to but that is a story for another day).

Fun fact: One of the most idealized depictions of builder culture, John Carmack in the early days of id Software, took place in Shreveport, Louisiana.

On the flip side, COVID broke a lot of the chains that tied builder culture specifically to SF. More and more effort has gone into distributing builder culture and its benefits to the rest of the world. This includes culture, ideas, meetups, hackathons, content, and more.

The question remains: Will it work? If you believe Twitter X, “SF is back.” Will distributed builder culture be able to take on SF builder culture and win? Does that even matter?

What are the best parts of SF builder culture anyway?

What makes San Francisco and builder culture so special? What are the characteristics that have helped create so much success?

• Speed: The pace is just faster. Move fast and break things. Many startups die because they are too slow. As Elon says “I have this mantra. It’s called, ‘If a schedule is long, it’s wrong. If it’s tight, it’s right.” Good ideas have competitors and they will try to defeat you in any way possible. Do things that don’t scale. When you aren’t around people working hard, it is very easy to underestimate how hard people are working.

• Experimentation: Validating startup ideas is a science. It helps when so many of your potential customers are so close to you. Building MVPs and trying to solve real problems. A/B testing, growth hacking. Less planning, more doing. Action reveals information.

• Acceptance of failure: Failure is part of a longer-term journey. Failures are not cut from the scene. That experience is seen as invaluable. Less tall poppy syndrome when a bunch of people around you are tall poppies.

• Ambition: The failures are paid for by massive successes, but for those massive successes to happen, people need to aim for them. When you see the current success of people around you and the past success of people in a similar position, you believe you can do the same.

• Openness: Detailing your journey is some of the best learning. Blogging, tweeting, sharing culture, advisors. Because there are many legit companies, best practices can spread by people moving between them.

• Extremely supportive: The rewards of helping someone are so high. They could be the next big founder, VC, hire, thought leader. Connecting and funding people is common. “How can I be helpful” is a meme for a reason.

Can the world steal the best parts of SF’s culture?

It takes a lot to build a culture. SF has benefited from 100+ years of culture building from the military, universities, semiconductors, computers, the internet, social media, YC, and now AI. All this compounds and the outcome is the startup builder culture it has now. The rise of AI in SF has shown just how “so back” builder culture remains.

Because of the amount of wealth builder culture has created, governments around the world have tried their best to copy it. They’ve created startup accelerators, hacker spaces, funding for AI, tech weeks, kombucha and cold brew on tap, tech conferences. In 2019, Sam Altman said there are 8,000 YC clones in China alone.

It’s not obvious any of it has made a big impact. SF remains the main draw to people who identify with this culture. Even if they don’t live there, the best founders, startups, talent, and VCs still spend a significant amount of time there.

Although the top-down government efforts aren’t seeing a lot of success (especially for the amount they invest in them), the bottoms-up community-led ones are punching above their weight and succeeding. They are doing this by creating pockets of builder culture wear it hadn’t reached before.

Examples of distributed builder culture

Small groups of people can create a culture of excellence regardless of their surrounding environment. This is proven by the many examples of scenes (or scenius) throughout history. Distributed builder culture is doing this.

The success rides on multiple waves like a community hole left by COVID, remote work, new frontiers in AI and crypto, and societal stagnation. From this, we are seeing a rebuild of these communities in new ways, empowered by new technologies, knowledge, practices, and distribution methods. Examples include:

• Buildspace: People bring their ideas to life over 6 weeks, sharing progress, and get support from a like-minded community of builders. Wraps with a 3-day final showcase that 1000s of people attend.

• Socratica: A network of weekly coworking sessions to work on your passion projects with like-minded people. Casual demos at the end of each session and linked to bigger demo events. Originated at the University of Waterloo but now has associated groups worldwide.

• Novus: An every Saturday builder meetup in Vancouver with demos at the end of each session. This is one I attend regularly. Also related is Minimum Viable Demos (which you should come to June 30th).

• Neo Toronto: Lumping together New Demos, 535, and the growing builder movement in Toronto.

• Grants: 1517 Medici Project, Slope (and Merge), Emergent Ventures. Low strings attached funding for ambitious people and projects. Often comes with encouragement and communities of others who received grant funding.

• Z Fellows: A community of determined technical founders with funding, guest speakers, office hours, mentorship, and networking.

• Pioneer: A now-defunct internet accelerator run by Daniel Gross. Created a community of 300+ founders in 50+ countries. Shared progress updates, pitched to experts, and got feedback.

• Indiehacking: Not exclusively indiehackers.com, but a group of people who combine startup building, geographic arbitrage, content creation, and solopreneurship. Trying to make money online and sharing their progress along the way.

• Verci: A second home and community for creatives in New York. Hosts workshops and shares weekly updates.

Each of these are relatively new creations that embody some of what made SF builder culture so successful while spreading it well beyond a single physical location.

How does distributed builder culture win?

Distributed builder culture has already done a lot to encourage smart, young, talented people to pursue building. It’s shown people are taking this path, creating support structures for it, and that it’s available to anyone. Spreading this to more people, no matter where, is extremely valuable.

Doing this requires communities to stay alive. This takes a large (and underrated) amount of energy. All communities have ups and downs, inflows and outflows, but need to continue to exist and create interactions between people and ideas. Momentum matters a lot.

To stay alive, communities need to focus on the right things. Distributed builder culture winning does not mean SF needs to lose. There is a lot for builders to learn from SF’s builder culture as well as other distributed builder communities. “Great artists steal.” Learn from startups and make something builders want.

Finally, building should remain the focus. “The main thing is to keep the main thing the main thing.” Every group of people will get tugged in various ways away from their defined purpose until that is lost. Builder culture is battling the status quo and needs to fight for a space for that culture to exist. You may not see them, but some people don’t want builder culture to exist, you can’t let them win.

If distributed builder communities can continue to stay alive and spread builder culture, their future is bright.

• Engineers build an app, reduce infrastructure costs, and increase conversion. You can often see it in MRR bumps or AWS cost reductions.

• Salespeople sell goods, find prospects, and close deals. Look at “on target earnings” for their impact.

• Marketers measure traffic and conversion and attribute it to revenue.

• Founders start companies, raise money, and sell products. Valuation, revenue, and exit value are all clear targets.

Although writing is a part of all of these roles, it’s rarely seen as core to value creation. This is because writing sucks at value capture. The most writing gets sold for is like $50 (book). Substack has helped, but only to raise this to ~$10-15 per month. These other roles have many examples of thousands to millions of dollars of value being created and captured.

The value capture in writing is a second-order consequence. Someone writes and the reader is the one who benefits and captures most of the value based on the action it inspires. It is then difficult to attribute action back to ideas from writing.

Million dollar blog posts are real

Even though many undervalue writing, I still believe in it. Take my favorite genre of writing: the humble blog post. Many blog posts have created millions of dollars worth of value. I’m not talking about side hustle, affiliate sites, but instead posts like:

• Patrick McKenzie’s negotiation guide. It has helped hundreds of highly paid software engineers make tens of thousands of dollars across multiple jobs.

• Mr. Money Moustache influenced a generation of people around savings, retiring early, and the FIRE movement with posts like The Shockingly Simple Math Behind Early Retirement. The FIRE and fatFIRE subreddits are filled with people with millions in savings influenced by him.

• Kevin Kelly’s 1000 true fans directly influenced and set the path entire creator business industry including Patreon which is valued at over $1.5B.

• The effective altruism industrial complex like Eliezer Yudkowsky’s Sequences and 80,000 hours’ career guide encouraged many smart people to take making money and AI much more seriously. As the best example, take Sam Bankman-Fried. He was influenced by effective altruism to make a lot of money (fraudulently), lose it all, but still maybe come out near even because of his investment in Antropic? Many more also influenced by this writing went to work at (and make lots of money in) the companies driving the AI boom.

• Susan Fowler’s strange year at Uber permanently changed the trajectory of Uber by being part of the ouster of Travis Kalanick. Depending on how you feel about that, this either saved Uber or stunted it (currently at a market cap of $146B). Either way, Kalanick went on to found a new company, CloudKitchens, which now has a valuation of $15B.

• Paul Graham’s collected works including How to Start a Startup and Do Things That Don’t Scale are arguably the most influential pieces of “startup culture.” YC’s worth around $10B, how much of that can be attributed to these posts?

  • Sam Altman and Chris Dixon’s blogs like How To Be Successful and Climbing the wrong hill are in a similar category.

• Ben Thompson’s commentary on big tech has had a major impact on their strategies (and stock prices). As a recent example, some think his analysis of Apple AI may have had a $150B impact on Apple’s market cap.

• There are many short-selling funds that publish their research such as Hindenberg Research. For example, their post on Nikola led to a 40% drop in stock price and an ongoing fraud investigation.

• Steve Yegge’s Google Platforms Rant condensed the Amazon way of building products and directly targeted change at Google. Since leaking, it has influenced many companies since then including Google). It shifted many people towards Amazon ways of building (seems to be working out for them). Jeff Bezos’ shareholder letters do the same, although not a blog post, they’re similar enough.

The million dollar blog post was inside you this whole time

More people than realize have a million dollar blog post inside of them. It has never been easier to share writing online that can reach and impact large numbers of people. There are no gatekeepers to sharing your ideas, and because of the leverage technology provides, a single idea can create more value than ever.

How do you write a post like this? The similarities in the above posts reveal some trends:

• Reveal a valuable secret. Often the best secrets come from experiences. These are one-of-one, but the lessons from them can be extremely valuable to others. If something isn’t a secret, it’s probably already priced in.

• Go deep. None of these posts are short or share information acquired simply. To manufacture one of these takes a lot of work, experience, research, writing, and editing. When you think about the potential payoff, it is worth it.

• Stay close to the money. It’s easier to measure the value your post creates when people can see the monetary impact it has. Salaries, negotiations, stock prices, careers are all tied to money.

• Aim at high-leverage people. The most impactful blog posts often aim to do so. Try to influence people in high-leverage positions: young people choosing their career, highly paid people negotiating, AI researchers, politicians, and stock traders.

• Become an expert. Rarely does someone who does not have direct experience with the subject write a million dollar post. Many people above are provable experts in the things they are writing about.

Because of the nature of internet writing, you never know who is going to read your posts or what impact it will have on them. I’m not notable, but I know a small number of blog posts have dramatically impacted my life and some of my posts have impacted the lives of others. Although not as clear as other work, writing can prove to be massively valuable over time.

This idea inspired a recent PostHog blog I helped with: How software salespeople will try to screw you (and how to fight back). It turns out companies can save hundreds of thousands of dollars in software contracts by negotiating better. By sharing our experience doing that, we hope to save companies a lot of money.

The warmth and smells of spring create a dramatic contrast with the dull, greyness of winter. You reorient to the world. Rather than hunching over to protect yourself from the rain and cold, you look up to see the blossoms of trees and the flight of birds. This contrast also opens you up to the potential to experience something magical. This feeling triggers a childhood memory in me of marbles.

In elementary school, the beginning of spring marked marble season. Officially supported by one teacher, an entire society revolving around marbles popped up. This society was complete with an economy, hierarchy, work, gambling, games, and terminology.

Everyone would have a collection of marbles, often kept in velvet bags (Crown Royal purple was a popular choice). Some would spend their recesses trading in quasi-flea markets with the marbles they had. These ranged from “peewees” to “tom bowlers” (I remembered this as “tramboli”) to “kongs” to “king kongs.”

More popular was a carnival of games that would pop up. The one I remember best consisted of:

1. Digging a small hole and placing a marble in front of it

2. Drawing a line in the dirt for participants to shoot from

3. Participants use their marbles to try to win the prize by hitting it into the hole

We would haggle over what line we could shoot from depending on the marble size we were shooting with. Some would throw dozens of small marbles to try for a bigger one, others would use a kong from close up for an easy win.

There was also a sort of golf that would happen where kids would compete to flick marbles with their thumbs into a set of holes. Older students would take younger students on as “caddies” for their golf game and pay them in marbles. This would also include a sort of alliance, patronage, and protection in the carnival games. Older kids would haggle for better deals for their “caddies.”

Marble season was perfectly matched to its environment. It was neither hot nor wet because both would ruin the conditions of play. It happened on your hands and knees in the dirt in the forest where the smells of springs were constantly kicked up. Although I didn’t realize it at the time, it was a pure celebration of spring.

At the end of a week or two, marble season would be over. Everyone would move on as kids in elementary school do. More dramatically for me, I moved from Victoria to Vancouver before I had a chance to participate in my final marble season (and all the status benefits being at the top of the hierarchy entails).

This makes it particularly nostalgic to me. Although no one played marbles in Vancouver, the smell of marbles was still there. Every spring I am reminded of this. The smell of spring (and marbles) rekindles a little bit of magic from my childhood and I’m grateful for that.

What I Did This Month

• For PostHog, wrote a newsletter on how we work asynchronously, tutorials on Astro and Remix, some docs, and a bunch of “alternatives” pieces.

• Added a bunch of new translators to Iliad Translations. Finished reading the Stanley Lombardo translation.

• Met a bunch of cool people. Focused more on studying Spanish. Had KBBQ and did an escape room for the first time. Went to the dentist (lol), French 79, Spacca, and Yellow Dog.

Thoughts

• You can just buy a textbook, read and study it, and you’ll know more about that subject than 99% of people. If you are concerned about not knowing something, read a textbook and see how you feel. ChatGPT can help. Partially inspired by how Greg Brockman became a machine learning practitioner (Elon did this for rockets too).

• Remind yourself that you will cross bridges when you get to them. There is little point in worrying about bridges you can't get to. Don’t end your journey because of some hypothetical situation.

• Every futuristic photo is filled with tons of skyscrapers, but is building them something we are improving at? We can coordinate better and build better, but can we build faster and more efficiently? Where are the prefabbed skyscrapers?

• One of the most important ways to get a lot done is getting started quickly. Avoid procrastination and just start. Ship an MVP or write a draft and iterate.

• The Iliad feels modern because the characters have complex relationships with each other, their enemies, and the gods. Each has backstories and good reasons for why they act the way they do.

• Advice on evaluating art from Tyler Cowen: Ask which ones you would steal. Go to auction houses and ask if they're worth it. Think of art as a puzzle: it requires context, history, psychology, and theology to understand; figuring it out makes you smarter.

• People become more risk-averse as they get more senior because they deal with smarter people more able to argue against something. It requires more energy to convince someone, which means it happens less.

• A lot of people treat startups like extracurricular activities rather than being serious about them and treating them like business. It is an extension of the obsession with extracurriculars in high school and college. Part of the index mindset.

Recommendations

• AI alignment has complex, theoretical reasons for their worry about the end of the world, but no real numbers on why.

• Is society all under various levels of hypnosis? How do we break free?

• This post on planning for unplanned work is basically the perfect blog post. Well-written, practical, and has great graphics and structure.

• Two new entries for one of my favourite genres, people give a lot of random life advice: 100 Tips for a Better Life and Thirty Observations at Thirty.

• Hundred Rabbits writes some of the most interesting posts about computing. They care a lot about sustainability and maintenance of software and hardware, which is much different than most of the rest of the tech world.

• The podcast Signals and Threads by Jane Street about ultra-optimized computing is always interesting.

• Sarah C. M. Paine has a deep knowledge of history of the US military and its rivals. Also some great meta points about understanding other areas’ history and its importance.

• Nick Bare with his marathon coach about creating the conditions for success. Fun fact: Olympic marathons are often on 5+ year running programs.

• 63% of direct labor time on mega-construction projects is spent waiting for materials and equipment. Microsoft made more revenue from gaming than LinkedIn. You can do everything in Excel.

• 90% of the millionaire lifestyle is already available to you. Caring is a source of alpha. The best health advice might be get more sun and eat more beef. Simple formula for doing things. Personally writing app invites. Does meditation allow you to sleep less? Try talking to ChatGPT.

• Dupont chart room. IRL Skyrim shelf. Powerful skateboarding.

Upcoming

• Visiting SF (just got back).

• Spending time with family and friends. Writing, coworking, practicing Spanish, reading, and coding.

What I Did This Month

• Went to London, specifically Croydon, for a PostHog offsite. Speedran London in 1.5 days while there (mostly art museums, V&A was the best). Haven’t figured out London’s food yet, everything I ate was mid. The English countryside is out of a movie.

• Published The hidden reason why BC’s startup scene struggles which blew up on Twitter and was republished in Vancouver Tech Journal. For PostHog, wrote a bunch of comparisons and more.

• Made some upgrades to Iliad Translations. Added Emily Wilson, Rodney Merrill, Michael Reck, Herbert Jordan, and W. H. D. Rouse.

• Also went to see Guns N’ Roses, BC Lions (congrats Matt for winning the pie-eating content), and Whitecaps. Celebrated Thanksgiving and multiple birthdays (Luisa!). Ate pink pineapple.

Thoughts

• My advice for working well remotely: work in person with friends when you can, change your scenery at least once per day, get the most important thing done first, and show your work more than you think you need to.

• Many of the world's top resorts didn't exist 50, like Punta Cana. They sprung out of the ground with relatively small amounts of capital. The rise of remote work, nomadism, and network states could spring many more over the next 10-20 years.

• What you think about is what you should do. Many people suppress their true passions to follow the passions of others. Having desires and not pursuing them creates dissonance Work on the projects you are excited to work on. Follow obsession. Doing this creates momentum. It enables you to do more and be better.

• It is clear when you are eating is junk food. It isn’t clear when your spending your time and energy on junk. There is infinite fake dopamine in the world now, it’s bad for your spirit.

• Action produces information. The best path forward is just to do something. After doing something, you are in a better position to make a decision. Repeating endlessly is a path to success.

• Renaissance Florence day labourers earned enough to live on working 3 days a week.

• How many people are playing the wrong game? They could be playing a game they would enjoy and have a good chance at winning, but instead play a game other people say are important. Deciding to change games is hard, and convincing someone else to change games is even harder.

• It is shocking how many people can attribute a significant amount of their success to cold outreach. Two examples I heard this month: Jesse Iztler's Marquis Jet marketing strategy was 10 hand-written letters a day, and Investor Michael Dempsey sends 10-15 cold emails per week to interesting people. Cold outreach rules the world.

Recommendations

• David Perell disappeared for a bit but he has come back stronger than ever. His “How I Write” podcast is excellent. Two highlights: Riva Tez and Marc Andreessen. Unironically makes me want to do Write of Passage even though I’m already writing online for a living (his marketing is working).

• The writing of Dryden Brown makes me a lot more bullish on Praxis succeeding. Good writing is one of the best signs, it shows strong clarity of thought and creativity. Also, appreciates the value of aesthetics. Two great pieces from him this month on how opinion leaders have outsized and centralized effects in crypto and how to implement techno-optimism.

• Steward Brand writes to most exciting pieces on the seemingly most boring subject: maintenance. This time how maintenance wins wars.

• Why isn’t there more ambitious people in the world? Rote learning, linear progression, lack of sense of historical belonging, lack of goodwill, and more.

• Like him or think he’s a normie, Malcolm Gladwell manages to be one of the best popularizers of ideas. You can figure out how he does it by looking on a paragraph by paragraph basis, look how he writes.

• What has become a bit of a traditional now, Matt Lakeman’s notes on Ghana was excellent plane reading.

• Designing Data-Intensive Applications. Hundreds of thousands of the smartest people in the world spend their time making sure your apps don’t break or aren’t slow. This is the intro textbook to learning what they do. For someone unfamiliar and interested, it’s an amazing overview. There are so many situations you need to plan for.

• Some cool AI tools I’ve discovered recently: phind (search engine and pair programmer), RecordOnce (AI video demo and tutorial creator), and Metaphor (another search engine).

• Master getting started. Post lifting PRs. Measure presence. What happened to the serious public intellectuals? One year to rewrite the web to be multiplayer. New ancient text just dropped. New dev advice. Cold weather gets people locked in. Infrastructure is the blocker, not takes. Life is writing ordered lists. AI isn’t even close to making it to the real world. Stop overengineering. Do as much as you can. Avoid the index mindset. Download the code and struggle. Naan in butter chicken. Bronze II Jane Street blackpill.

Upcoming

• I love Fall because it’s a great time to work, especially in Vancouver. Eventually, you embrace the rain and it becomes a partner in good work.

• Writing, coding, reading, meeting cool people, lifting weights.

What I Did This Month

• For PostHog, published about feature flags as a service, SQL, our Temporal batch exports system, decoupling deploy from release, and a newsletter on uncovering users’ real problems.

• Upgraded Iliad Translations with some new and improved styling in Tailwind.

• Got my art hung in my apartment (after 2+ years of it sitting on the floor). Saw the Challenger map once again. Ate Liquids and Solids as well as BBQ many times. Ran more, changed my leg day to fix my knee and squat form. Went camping for the first time.

Thoughts

• The Visual Display of Quantitative Information. As much as computer tools have automated it, there is still alpha on doing it well. Learned it is surprisingly similar to good writing. It requires ruthless focus and editing. The book is also filled with beautiful and inspiring graphics.

• There is a shocking gap between the quality of art and code LLMs create versus the quality of writing. This is because of the quality of the sources (training data). Few people publish bad art online. Few people publish code that doesn’t at least work online. Basically everyone publishes bad writing online.

• Nearly everyone is too risk-averse. Doing MVPs reveals how risk-averse you are. Having an MVP fail isn’t as bad as you think it is. It makes you realize that often motivation is the blocker, not anything else.

• For everything you are trying to learn, working with a coach or a tutor can help you be more successful. You can even do this for things you don’t think of as “coachable” skills.

• When talking with someone, how interesting they find you is a function of how interested you are in them. Every person you meet knows a lot about something you know virtually nothing about. It's your job to discover what it is.

• Startups have a big advantage over big companies in marketing because as a company grows, they care more about attribution and take fewer risks of illegible channels. A lot of marketing simply doesn’t scale either, but startups often benefit from doing things that don’t scale.

• ARIA is an example of an ambitious government agency trying to transform the UK economy and create a radically better future. Doing this requires risk and government agencies will not take risks unless they are forced to. BC can learn a lot from this, specifically InBC and Innovate BC. Matt Clifford, the Chair of Aria, explains this well.

Recommendations

• Discovered this month. Although mostly focused on value investing, there are some standout gems worth reading including “The Reading Obsession” (Warren Buffett’s network is more important than you realize), “Focus: The Last Superpower?” and "Three Years Of Writing Online.”

• At least 2-3x a year, Elon sets an insane deadline which requires all-hands-on-deck and around-the-clock activity to meet a goal. This helps make his grand missions a reality.

• Great ideas require the cultivation of a state of mind that isn’t found by accident (via ).

• We are entering an era of new technologists who focus on increasing the “capability landscape” (via )

• Three excellent podcasts I’ve been listening to lately: Audience of One, The Danny Miranda Podcast, and Moth Minds (via ).

• A highlight from Justin Mares’ episode with Danny Miranda: Joe Lonsdale has 3 minute calls with 100 of the USA's top (entrepreneurial) graduates every year where he asks what he can do for them and then gets his team to do it. This is what 10/10 networked looks like.

• An app for visualizing weather year-round. A JavaScript library for interacting with PDFs. An app for testing and developing with webhooks, wildly functional design. A map of pinball tables.

• There’s still a lot of alpha left in newsletters. Lenny’s newsletter hit 500k subscribers, seems to be a successful business, and is focused on a niche profession (product managers). Similar newsletters are waiting to be created.

• Praxis reading list. Jarred Sumner shipping Bun. Don’t be afraid to read the code, most of it is just for loops, variables, and if/else statements. A real, successful, walkable development project. Elon Musk’s algorithm. Foundational books. SF Bay area hosted 106 in-person AI events in August 2023.

• Do you ever yearn? Digging a hole. Believe children can do great things. Write something online every day forever. Believe unbelievable things more. Walk around, get a coffee.

Upcoming

• Celebrating birthdays. Going to the UK.

• Writing, coding, working out, reading. Thinking about AI, programming fundamentals, and coaches.

What I Did This Month

• Wrote a lot. Published History Is Written by the Translator, Reclaiming Your Junk Time, and LLMs are mid (but that’s ok).

• For PostHog, published newsletters on GitHub and GitLab feature flag usage and A/B testing, a blog on writing great product survey questions, and more.

• Improved styling for Iliad Translations and added Peter Green, Anthony Verity, and Ian Johnston. Learned a lot of SQL.

• Did the Grouse Grind twice, watched The Bear and Oppenheimer, went to a Whitecaps game, and ate multiple formats of BBQ.

Thoughts

• Blood Meridian by Cormac McCarthy. In novels, there is a balance that needs to be struck between randomness and complexity. The first time I read Blood Meridian, I didn’t fully understand what was going on (it takes time to get into McCarthy’s writing style), thought it was random, senseless violence, and gave up 1/3 of the way. This time, I made it through and realized its deep complexity. Justin Murphy and crew’s analysis helped too. There are seemingly endless interesting ways to analyze it. Modernity, Nietzschean, American, Colonial, Christian, and magical reading are all real and possible.

• The grass is greener where your friends are. For some people unhappy with where they live now, they see moving as the solution. Happiness might come in another place, not because that place is better, but just because their social circle is reset. You get a chance to make new friends. Ones you might fit better with. Travelling as a young person is a short-term version of this.

• Validating a startup is about validating a worldview. Your opportunity is the difference between what you believe and what the world perceives. You can shortcut by writing a blog post that validates that worldview as well. A startup can be seen as a worldview wrapper.

• Getting more people to use AI will reveal humanity's preferences for its alignment. This means one of the best potential AI alignment tasks is teaching people about using AI, helping them understand what it's good at and its limitations. Give them a Fingerspitzengefühl of AI.

• How much of the population of the world lives on land filled in from the sea and lakes? Boston, Netherlands, Singapore, New York, and Dubai all have done it. Why don’t cities (Vancouver) see this as an option?

• Maybe you’ve heard of how the world completely changed in 1971, but have you heard about the massive amounts of protests happening worldwide in 1968?

• Does bad weather make the world more productive? If there is bad weather, you’re forced to stay inside and work. Industrialization enabled this to happen, and the information age accelerated it.

• Why are LLMs good at SEO? Because “SEO” isn't about writing good article, but articles lots of people can find. By being based on writing for "lots of people," it can do a good job copying it.

Recommendations

• Stephen Wolfram on The Joe Walker Podcast. Wolfram is inspiring. Does a 4.5-hour interview because he has the freedom to and it’s a good conversation. Runs a sizeable tech company for 30+ years and doesn’t plan to go public. Continues to do groundbreaking research in many fields. Supports the ambitions of young people. Repeatedly works on projects at the top of humanity’s ambition levels with large time scales. Helps anyone curious by creating massive amounts of informative content online, doing streams and AMAs all the time.

• When coding, I usually use a combination of Copilot and ChatGPT. Often this requires copying and pasting context for them. Cursor is a VS Code based editor with generation, context, and chat built in. It’s working well so far, see a demo video.

• Textbelt is an open source API to send text messages, extremely simple, and literally works by sending texts as emails to every service provider, which gets rendered for receivers as texts.

• What the humans like is responsiveness and 100 things I know. A lot of being a good person is being responsive, attentive, and thoughtful over the long term. Give people your focus and you will get rewarded.

• How do you go from 0 to 1.5M followers on Twitter in a year? Spend ~8 hours per day writing threads like The Cultural Tutor did.

• You Should Be Working On Hardware. Simply put, hardware (compared to software) is an inefficient market. The most admired entrepreneur on the planet isn’t a SaaS founder, but a multi-time hardware one (Elon).

• Organizations are powered by spirit, and there are more ways to invigorate the spirit than just making money. Brunello Cucinelli is an example of this, and he channels history for many more examples. Founders Podcast episode on him is also excellent.

• There are a massive number of software engineering blogs out there, many support by the top software engineering organizations. Most of them are not well written, but they are extremely informative if you’re willing to find the right information and do the work. diff.blog is a good way to find them.

• Reading list for a new city. Search university professors on YouTube for better content. Hemingway reading list. Mike Mentzer high intensity training philosophy. Don’t get distracted by good ideas. Walk around. Iterate, cultivate taste, solve for distribution, and get good at managing your psychology. You don’t need a government project for AI. The world belongs to the energetic.

• Fishpond mosaic. Historical Chipotle. Chimp plays Minecraft. Milk is good. America vs Radiohead. How to screw over the oligarchs. The sun will continue to shine. Cringe is in. Business is a wrapper of aesthetic taste.

Upcoming

• Working on my coding skills. Learning TypeScript, reading Designing Data-Intensive Applications. Upgrading Iliad Translations style. Writing small scripts to automate stuff and work with LLMs.

• Writing. Working on articles about AI alignment, validating startups with blogs, and learning to code.

What I Did This Month

• For PostHog, published “A/B testing examples from Airbnb and YC's top companies,” “Why we test in production (and you should to),” “Adventures in null handling: Null money, null problems,” and more.

• Travelled to Europe for two weeks. Started in London to spend time with coworkers. Drank the worst “juice,” ate at Wagamama, saw The Comedy of Errors at The Globe (highly recommend), and went to the Tate Modern (which got rid of the Rothko room I liked so much when I was there last).

• Next, went to Cluj in Romania for a wedding. Drank a lot of Borsec (GOATed Romanian sparkling water), visited many churches (with better art than most art galleries), ate at KFC twice (it’s better there), and hung out with friends.

• Finally, went to Bologna (via Rome because of a cancelled flight). Ate a lot of pasta, gelato, and mortadella (the best), visited an anatomical theatre, and went to a dinosaur and rock museum.

• Launched a new project dedicated to summarizing and comparing different translations of Homer’s Iliad called Iliad Translations. Let me know what you think and if you have any recommendations.

Thoughts

• London

  • Restaurants are able to be weirder (higher variance) when they can consistently sell high-margin alcohol and can rent smaller spaces. Alcohol subsidizes much of what is cool in London in a way Vancouver cannot possibly match.

  • People in London dress well because it is consistently cool all year. When it is too hot or too cold, it limits someone's fashion choices.

  • Something is wrong with my mental model of London. It is expensive and people don’t make that much money, but restaurants and stores are packed and people are well-dressed. Where is the money coming from?

  • London is definitely 10x cooler than Vancouver and there is 10x as much to do. This is with ~3x the population (depending on how you slice it).

• Romania

  • The mosaic is a beautiful art form. The combination of architecture and art. Long lasting and made for the common people. Bringing culture to the masses.

  • On the highway, you fairly regularly pass horses and buggies and their version of moonshine (Țuică) is sold in jugs on the roadside.

  • Surprisingly, the prices for groceries and coffee in Cluj were relatively similar to prices in Vancouver. The alcohol and housing were significantly cheaper.

  • My favourite Romanian food is plăcintă, a fried pastry filled with cheese.

  • Churches progress as funds are raised for them. It seemed common for churches to be in various states of completeness. One church we went had been in progress for 30+ years. This didn’t stop the basement from being used for Sunday service.

  • The most popular energy drink in very Christian Romania is named "Hell.”

• Bologna

  • Cluj had the most anti-communism graffiti I've ever seen, Bologna had the most pro-communism graffiti I've ever seen.

  • Bologna has the nicknames of being the wise, fat, and red city based on their university, cuisine, red buildings, and left-wing politics. All of it feels extremely true and connected. Being a student, having delicious meals, not wanting to work, and being a communist seem correlated.

  • Cold cuts (mortadella) and cheese were significantly better than any of the pasta meals I had.

  • Breakfast in Italy is a waste of a meal. Lunch is 1000x better.

• Read the texts the experts are embarrassed they haven't read.

• If you want to have good conversations, and you’re not naturally gifted (you’ll know), you actively have to put effort into those conversations. You need to be actively listening, asking questions, trying to make the conversation a better one.

• People will take your semi-garbage if you post it online for the right price (which might be free). Transactions on Facebook Marketplace and Craigslist likely don’t show in official GDP numbers but I wouldn’t be surprised if they accounted for a significant percentage.

• Usually, advice for someone else is just advice for yourself in disguise. You basically repeat what works for you (in your eyes) which likely isn’t what’s best for someone else. People take advice seriously, so be wary.

• There are many ways to learn to code, but everybody trying to teach you wants to make you think there is only one. You can do 1s and 0s upwards, edit one variable, read the manual, build a project, or copy the masters. They all work, but what is really key is writing code and managing motivation.

Recommendations

• The French Intifada. If you’ve heard about the riots in France at the moment, this will help you understand why they are happening. The history of French North Africa is nearly incomprehensibly brutal and not that long ago.

• The Art of Not Being Governed. Dan Wang recommended this book on the tribes in the highlands of Southeast Asia. Reading this made me realize how inconceivable it would be to live outside the state in the modern world. It requires generations and generations of work and the development of strategies for living outside of the state you would never consider as important like growing roots instead of grain and not writing down any history.

• Alice Maz has written two of the best posts I’ve read about important subjects for me. First, about dynamics in internet communities and economies (in Minecraft). Second, about learning to code (and what to think about).

• Been recommended many times on Twitter but Jeremy Giffon on Invest Like The Best was excellent. Listened to it twice in a row. Takes an opposite approach to me, instead of many small swings regularly, goes slow and takes big swings less often. Tweet length summary.

• Tyler Cowen is an Information Monster, but you probably aren’t. That means the good of AI (creating information) outweighs the bad for him, but that calculation might not be true for you.

• Before listening to this excellent podcast on Horatio Nelson (part 1, part 2), I had no idea who he was. Now I think he’s a fascinating and deserving national hero.

• The best writing advice is simple. For example: you can write anytime, anywhere; and actually write, don’t pretend.

• Heraldic map of Europe. “The Devil Went Down to Georgia” is the most American song. Nobody looks out plane windows. Alan Kay as based Urbiter. Ice cream good. Serious reading is an individual pursuit. Ad hominem arguments are valid. Respect the ability to hang.

Upcoming

• Writing, potentially a personal piece on “junk time.”

• Working on Iliad Translations, styling needs an upgrade and adding more translations.

What I Did This Month

• Went to Calgary, Banff, Waterloo, and Bellingham. Doing this involved waking up at 3 am 5 times in 2 weeks for various reasons. Banff has optimized the tourist experience very well, with so many outdoor activities to do. Waterloo continues to expand every time I go there (and seems to have a lot of space to continue). Also how such a strong University ended up there is still a mystery to me.

• Had dinner with friends and family, made pasta (twice). Ate lots of burgers, feijoada, and pastel. Drank raw milk (fun fact: illegal to produce in Canada, but not to consume).

• Wrote lots, mostly for PostHog. Published real user monitoring, feature flags vs configs, a bunch of tutorials on HogQL, and more. Wrote more code for personal projects which I am very happy about.

Thoughts

• To be or to do. Once you get over the hump of having agency in life, the next hangup many people get caught on is wanting to be someone rather than wanting to do something. The world has more money than ideas and more ideas than agency.

• A lot of people say “we need to build more housing” without any practical way to do that. Just saying it isn’t going to make it happen. The alpha in housing development is in regulation (I think), so you need to figure out practical tactics to leverage this. Dezoning, public-private partnerships, expanded role in government actually developing housing, and higher levels of government providing lower levels more leniency are all potential solutions. Labour, materials, and processes are pretty optimized otherwise (no alpha). Big projects are done because the fixed(ish) cost of dealing with the government is spread over more investment.

• For any skill you want to improve at, pretend you have a coach (for accountability and programming), and follow their lead. Push yourself, do things every day, and look up what experts say online (YouTube, papers, books).

• Intensity is the strategy, work harder, you can figure things out and make them happen.

• So much of work is constrained for random reasons that people believe those constraints are real, and it warps their thinking about their whole lives.

• Weightlifting teaches you strategies to defeat local maxima.

• Coding requires downloading and holding a massive amount of context. The more effectively you hold the context, the more successful you'll be.

• You need to live somewhere you can walk. Walking reclaims the world from hostile, unseen forces in your mind.

• WWE has the 10th most subscribed YouTube channel.

• “I once heard that 90% of culture is just ‘winning’ - when a company is winning, everyone’s happy, rich, being promoted, and they see their work as contributing to something bigger than themselves” or in the wise words of Al Davis: “Just win baby.”

• Instead of building physical monuments, rich people now just build monuments to themselves. Homes, experiences, health, “charitable foundations.”

• Creativity is literally magic. You are combining ideas pulled from the ether to shape something real.

• Most good opinions are copied from someone else. Good AND original opinions are the product of a lot of work.

Recommendations

• Have enjoyed discovering ’s writing this month. My favourites are “How to build a worldview,” "I am asking you to want something,” and “You should consider yourself exceptional.” A lot of the ideas of simple in theory, but difficult in practice, which is why they are important to constantly repeat. Pick a thing and believe in your ability to bend the world for that thing.

• Your monthly reminder that the world is a malleable place.

• I have been thoroughly George Hotzpilled for a while now. Truly a man bringing power to the people (from iPhone jailbreaks to self-driving cars to now AI computing). One and two good podcasts with him this month.

• A lot of America is chronically sick because they don’t eat well. Justin Mares talks about this health crisis and Simon Sarris writes about how he eats (which is a potential solution).

• There are a lot of places in the world that are shockingly different from the place you live now, Benin is one of them and Nigeria is another. You can learn a lot about the world by exploring these differences.

• Michael Crichton wrote a lot. During medical school, he wrote “seven novels, a movie, several more manuscripts and screenplays and a fast‐paced, provocative book of non‐fiction on the state of the American hospital.” He could write “16 hours a day for a week or two, often turning out 10,000 words a day.”

• In my travels this month, the Calgary Central Library was one of the highlights. Definitely an acceptable monument to the wealth of the region. Bonus: they have a cover outdoor basketball court across the street.

• Jena had scenius in the late 18th century, a group of people in one location with interesting and impactful ideas. Not talked or researched at all compared to Scotland, Vienna, Silicon Valley.

• I have enjoyed Cindy Sridharan’s writing about software development, favourite ones include: “A decade in review in tech” and “What is DevOps?"

• Large technical projects are difficult, but critical to advancing. Especially with the power of computers, you can take them on. Mitchell Hashimoto provides advice, highlights include “I try to think what a realistic project is where I can see results as soon as possible” and “No matter what I'm working on, I try to build one or two demos per week intermixed with automated test feedback as explained in the previous section.” Similarly, Ryan Kulp on new products (wireframe, data model, frontend, backend).

• A mainstream narrative of history is progress is the ability to “expanding circle” of moral concern. It turns out this ignores a lot of narrowing that happens too. For example, we care much more about animals, but much less about religion or ancestors.

• MrBeast is at the top of the YouTube game and isn’t stopping. Says repeatedly he is optimizing for making the best videos possible (and doesn’t mind spending tens of millions to do so). Can’t help but wonder where are the MrBeast’s of other fields.

• Future Canada plan. 31% of Canadian new home cost is government taxes and fees. Canadian startups should look global rather than local. 40% of men and women are still in school in their 20s (up from 8% in 1940). The car robbed us of our friend, the horse. Incomes are up, net worths are down.

• There are more efficient ways to turn money into status for relevant stakeholders. Recreate the masters, rewrite the classics. Just go do the most important thing right now. You are not writing the state of the union, you are writing to your friend. Don’t try to build an audience, read some books.

• Fire joke in the team chat. VCs are the army and navy. Jacked candidates. Rotisserie chicken. The Social Network soundtrack has created a lot of enterprise value. Cormac McCarthy was the last real writer. Every social network contains a bad implementation of half of Urbit.

Upcoming

• Writing, writing, writing.

• An Iliad-related website that my coding time has been going towards.

• Traveling to Europe (specifically London, Romania, and Bologna). If you know someone I should meet or have any recommendations, let me know.