Reddit used to have one of the most open APIs on the Internet. Until 2023, API access was free with a fairly generous rate-limit. The platform made itself scrapable, with every page having an alternate .json twin that returned the same contents on an unauthenticated endpoint. This paved the way for easy to use third party tooling, like PRAW, a popular python API wrapper, or Pushshift, a publicly searchable archive of every post, across every subreddit, that published dumps monthly as torrents. For fifteen years, Reddit’s API quietly underwrote a significant amount of computational social science.

In 2023, this started to rapidly change. First, the platform announced that API access would now have a paid tier, calling out data collection for machine learning as an unauthorized use case that needed to be prevented. In 2024, leading up to an IPO, Reddit listed data licensing as a revenue vertical. In June 2025, they sued Anthropic. In Aug 2025, they blocked access by the Internet Archive’s Wayback Machine. By 2026, open API signups had completely closed. Access now requires manual approval under Reddit’s Responsible Builder Policy — a name that manages both moral performance and bureaucratic deflection at once. For researchers, a separate program called Reddit for Researchers and requiring official institutional affiliation is the only path. Community projects like Pushshift have either shut down, stopped syncing, or themselves developed gates. This is deeply ironic, as Reddit was at an early stage co-owned and architected by Aaron Swartz, a famous freedom of information advocate, who later committed suicide after being grotesquely scapegoated for running a script to bulk download articles from JSTOR. And unfortunately, it’s a case study in where the open web is going.

What exactly is the open web? It’s a system of protocols and standards, from TCP to HTTP to HTML, that operates in a decentralized, interoperable way. It is what is surfaced by Google, but it is not Google. It is your website. It’s my website. It’s Wikipedia, and online magazines and every blog ever created, and until very recently it was Reddit. The most important part of the open web is that it’s open. You do not need to log in to see what is on the open web. You do not need to use a particular app, platform, or device to create something there (yes, Reddit has logins, but that was a handwave before 2023). A server is just a weird computer, and though most people pay for hosting, with some fiddling, you do not need to. The most basic layers of the open web are not complicated: I teach HTML workshops sometimes, and you can impart enough knowledge for someone to edit a simple website template in an afternoon. Yes, it will not look good on mobile and the divs may not be centered, but it will render and perhaps even be charming.

You may not think very often about websites. There is a whole apparatus of standards bodies and network engineers that keep things running so you don’t have to. Websites are mundane. Indeed, so much so that I had doubts about whether anyone would want to read a column about them. Why care about websites in 2026? The online experiences engendered by them are so commonplace that it may blind you to their significance. Do I need to do the silly dance where I remind you what the world was like before the Internet? Why Aaron Swartz wrote the Guerilla Open Access Manifesto, and was scraping JSTOR in the first place? We are now several generations of social and economic problems downstream, but don’t let it obscure the majesty that for a brief, brilliant moment, we really did think information could be free.

Instead everywhere, the open internet is closing. Towards the end of 2025, the New York Times started hard-blocking the Internet Archive’s crawlers. Historically, a robots.txt file was used to opt-out of being crawled, but this was a request, not a wall. Now some news sites like the NYT have put more robust prevention methods in place. The blocking is also widespread; A Nieman Journalism Lab report investigated the robots.txt files of 1167 news sites and found that “241 news sites from nine countries explicitly disallow at least one out of the four Internet Archive crawling bots.” A similar later analysis found 382. It’s true that the Internet Archive is not the be all and end all of the open web, but the requests it makes as it crawls sites encompass the same open protocols the rest of the open web works on. It is emblematic of the access to information dream the early Internet was born into. Not only that, despite what news outlets today might think, it’s often a crucial source for tomorrow’s journalists.

AI is specifically implicated in the blocks. A NYT spokesperson went on record, saying, “We are blocking the Internet Archive’s bot from accessing the Times because the Wayback Machine provides unfettered access to Times content — including by AI companies — without authorization.” You thought LLM training runs using your data without permission was the problem? The cure may very well be worse than the disease. And it’s true that the Internet Archive seems to be a frequent source for those building training datasets. A 2023 Washington Post analysis found that the Wayback Machine was the 187th most commonly represented web domain (out of 15 million) in a dataset used to train Google’s T5 and Meta’s LLaMa. And a documented mass-scraping incident in 2023 was enough to briefly take the Internet Archive offline.

Why are these news sites blocking Internet Archive? Do they hope to license their dataset, the first step of which is blocking the channels through which it could be drunk up for free? Is it hostility towards AI for other reasons, maybe AI-threatened job loss, or the existential crisis of generated fake news? Sometimes the reasons may be more complex — as the CTO of the Baltimore Banner, which blocks IA but not scrapers used by ChatGPT and Claude, explained, their concern is that if their content is consumed via a third party like Internet Archive, it might lose proper attribution. The Internet Archive itself is perhaps not free from incentives either: After the 2023 scraping event that resulted in an outage, Mark Graham, head of the Wayback Machine said, “We got in contact with them. They ended up giving us a donation.” The hunger of every AI lab for large swathes of organic free-range written language has turned every text-based archive into an unlatched chicken coop. But it turns out there’re foxes on all sides.

Navigating the Internet has also been fundamentally transformed in the last two years, in ways which threaten the open web and the business models of news sites. A recent analysis by SparkToro, an online audience-research firm, titled “In 2026, Less than One Third of Google Searches Still Send a Click” traced — alongside the headline finding — a declining clickthrough rate since 2016. AI search summaries are widely blamed for the phenomenon, and a 2025 Pew Research Center report corroborates that searches with an AI summary result in roughly half the clickthroughs of those without (8 vs 15 percent). But to focus too much on the AI search summary alone is to miss a big part of the story, which is the rise of AI apps themselves — a significant fraction of LLM users are looking up information that might have previously been a Google search — or turning to in-platform search on apps like TikTok and Instagram. This trend is harder to quantify, but we can look at ChatGPT’s runaway growth, as the fastest app to ever reach a billion monthly active users. Or the recent comment by Eddy Cue, an executive at Apple, testifying that in April 2025, search traffic on Safari had dropped for the first time in 22 years. He blamed AI.

The SparkToro piece, which seems sincere in its attempt to quantify the trend despite being ultimately the sales blog of an online marketing platform, points to the rise of “Zero Click Marketing” aka, marketing tactics that still work, even when no one clicks on your website at all. It advises building a presence on platforms you don’t own (in Apps) and shifting to thinking of your website primarily as something that will be indexed via LLMs and regurgitated from training data. This already even has its own liturgy: llms.txt, the opposite of the classic robots.txt, that explicitly invites LLMs and instructs them on what to look at first. Roughly 10% of websites implement it. The problem is no one seems to be looking: One monitoring firm’s analysis of more than half a billion AI crawler requests found only 408 requests for llms.txt. Is this the future of the open web? A choice between platforms like fortresses pulling up the drawbridge, and the websites like supplicants, trying to give keys to someone who’s already taken the town?

There are movements that hope otherwise. Earlier this year, 2026 was being branded the year of “going analog” by a number of influencers, mostly Gen Z (and somewhat ironically spreading the message via Instagram reels and YouTube). Some of this looked like a turn towards offline single use devices, like digital cameras and MP3 players. Some of it looked like adoption of dumb phones. Some of it encouraged people to make handmade websites, on platforms like Neocities. In the last year a number of magazines and platforms have cited Neocities as proof the Internet can still be a place for art, creativity, and weird websites, from The Verge to Plaster Magazine. Neocities is indeed a refreshing countertrend, though with a little over 1.6 million sites, it is still tiny in relative terms. Another touchstone for the analog movement might be The Luddite Club, a no-smartphones irl meetup founded by a group of New York teenagers in 2021, with chapters now in over 30 cities, from Seattle to Stockholm. Culturally, we’ve obviously reached peak smart phone. And while I actually support all of this to a degree, and the Neocities user is indeed doing something meaningful to continue the craft of the open web, it seems equally true that encountering something (a website, a dumb phone) through the lens of nostalgia itself indicates the death of the form. These trends can claim novelty only when the thing they imitate is no longer a primary form. And none of this will be enough to halt the ongoing enclosures. In the 2010s, my generation revived cassette tapes but it did nothing to halt the rise of Spotify.

The threats to the open web are obviously structural. They began long before LLMs, though they have been massively accelerated by them. I would place the antecedent in the dominant business models of the open web. Namely, a conflation of open protocols and freedom of information with ad revenue as the main, or only, monetization path. This is how various news platforms ended up so structurally vulnerable to a declining click through rate, isn’t it? It’s also what burned out a generation (mine) of online creative people from organizing indie online spaces. Maybe the transformation charted here is inevitable — we are always creating new commons to enclose after all — but I think it bears asking instead, what would it take to be different? The counterfactual — a Reddit that kept the basic API open but continued to monetize through paid features, premium tiers, in platform marketplaces, etc — is certainly possible to imagine. The tactics needed are twofold: an analysis of the problems that doesn’t collapse into a handwave about “capitalism” and ongoing experimentation at the layer of protocols and finance. It means competing on more than rhetoric and instead addressing incentives. It means actually serving the user. A structural problem needs structural solutions. The open web can’t run on vibes alone. What would it take to actually win?

In systems where precise time-keeping is necessary to prevent critical failure a Real-Time Operating System, like FreeRTOS or VxWorks is used. Unlike windows, OSX, or a garden Linux varietal, real-time OS’s guarantee that tasks happen when and as they are requested with microsecond precision. If there are multiple competing tasks, the OS can manage which ones have priority between those which need real-time properties and those that don’t. Notably, this “real-time” doesn’t need to be real in the sense of the atomic clock, the computer’s clock could be inaccurate, wrong by a whole day, but it’s deterministic in that tasks happen almost exactly when they are supposed to in relation to previous tasks. The trade off is usually in less throughput and less optimized CPU utilization, but why did you come here expecting a free lunch.

If this is unnecessarily technical and you’ve never had an interest in machine shops or the aerospace industry, surely you’ve heard of Y2K? In hindsight this may seem like it was more a media event than a software error, but the bug at its core -- theoretically writing a four-digit 2000 to a two-digit storage slot as 00 and all the various fallout that could ensue -- had a very real potential to cause problems. Did you know the default way pretty much all computers store time is as an integer that’s been going up, second by second, since January 1, 1970? At the time of this writing it’s 1779725019 o’clock. This is clearly not a system meant for human-scale timekeeping. And it may not be all that robust for machines either. Many operating systems store this as a signed 32-bit integer, with a maximum value of 2147483647. Which is to say, we’re right on track for a bigger Y2K as of January 19, 2038, at 03:14:07 UTC.

All computational systems contain an implicit model of time and space: time in the sense of how things are ordered and scheduled, and space in terms of how they are stored. Time specifically, the main subject of this essay, has different properties. Some of which are accuracy, how closely it’s kept to what we might call the actual objective time of the universe, approached only in a clumsy manner by humans and known perfectly only to God, maybe. Determinism, how causally and reliably one thing follows from another. Ordering, what happens before and after what else. And duration, whether elapsed time can be measured or verified. These architectural choices go on to constitute the things that can be done with the system, and, I suspect, the culture that grows around it. Conceptual models of time are a bigger force than operating systems alone, as who can forget this point from Silvia Federici, about how the watch constitutes the worker and entire pre-modern ways of time-keeping had to be eliminated as a matter of course in the project of industrialization?

The revival of magical beliefs is possible today because it no longer represents a social threat. The mechanization of the body is so constitutive of the individual that, at least in industrialized countries, giving space to the belief in occult forces does not jeopardize the regularity of social behavior. Astrology too can be allowed to return, with the certainty that even the most devoted consumer of astral charts will automatically consult the watch before going to work.

When Satoshi Nakamoto first circulated the release notes for Bitcoin, it was not called a blockchain, but a timechain. People have been joking that proof of work is a clock ever since -- and they’re not wrong, it’s barely a joke. Every bitcoin block is minted by a hash of its blockdata, which includes the timestamp and the hash of the previous block. Each hash is cryptographically sound, meaning it could not be known in advance, it had to be brute forced via an energy-intensive computational search process. The energy consumption of the process and the directed nature of the chain data structure constitute a thermodynamic proof that each block existed in its specific form and order. There is only one correct time on the bitcoin network. There is only one correct chain. It has only one correct order. Notably, bitcoin time can actually be fairly inaccurate in terms of time outside the bitcoin network, up to two hours give or take, but it’s unilaterally correct within its own universe. The entire system, its complexity, wastefulness, and slowness, are all architected to make that guarantee as strong as possible. And it can be easily and independently verified -- validating the chain of hashes is trivial compared to creating it. It requires trust in nothing but some specific math and the existence of entropy. If bitcoin is a clock, it is one that moves resolutely in 10ish minute increments, where each correct time is briefly the absolute authoritarian ruler of the entire network, until the next correct hash rolls in and stomps on its face forever.

And so we can ask, what kind of a clock is an LLM?

First, an LLM has a dataset cutoff. When you are talking to Claude or ChatGPT, you’re essentially reaching into the past -- whenever the last knowledge cutoff was. This is revealed in all kinds of small details when interacting with them. Try asking one what the Strait of Hormuz is. You will receive a geographical description and be informed that Iran has periodically threatened to close it in moments of geopolitical tension. Ask it about Claude Mythos, the newest model created by Anthropic -- it has no idea it exists. It will chide you about speculative claims. It still calls the Department of War the Department of Defense. The training data is from 2024, and the only reason the LLM knows it’s not still 2024, is because the date is injected into the context window.

Moreover, stale context windows don’t get date and time updates, at least not automatically. Open a conversation in the evening and mention that you’re having trouble sleeping. If you return to it in the morning, the model will encourage you to go to bed, as if no time has passed since your last message. Ask it to do something related to a project with a deadline and return to the thread after you’ve handed it in -- the model will remind you to keep your deadline in mind, with a reminder for a date that passed a week ago. As a clock, an LLM is discontinuous and fractured, massively parallel, with each ephemeral instance having a different rapidly decaying timestamp. This disconcertingly mirrors the cultural regime being hastened by consumer LLM use: increasingly personalized content, accelerating loss of metanarrative, competing sycophantic realities, and at the extreme end, psychosis.

What might an LLM gain from other computational models of time? Can we combine time models in ways that have beneficial properties? I’ve already explained the computational model of time operating in proof of work, let’s gesture at how they could intersect. How could an LLM use a thermodynamic clock?

When you wake up blinking in the morning, if your phone says it’s midnight, you might wonder if some kind of overnight update had gone wrong. But you will not doubt the sun coming in your window. An LLM is always operating within a set of authority claims it must make judgments about. It receives a system prompt from the company that made it, information about which tools and skills are available, automatically stored memories about the user, and context input manually in the form of prompts. Labs certainly worry about LLMs developing the ability to deceive. Is the LLM capable -- or will it become capable -- of doubting the claims made by its own infrastructure? In a recent podcast episode, Amanda Askell, Head of Personality Alignment at Anthropic weighed in on this in a segment titled, “Does Claude Know the Real World Is Real?” Right now, the answer is no. If it calls a tool, or searches the Internet, it doesn’t “know” those endpoints are real. This year on April Fool’s Eliezer Yudkowsky even joked that the system prompts of most major models were telling them it was March 32nd, because April 1 led to a noticeable decrease in accuracy. A lot of people believe this, because it spoke to a deeper truth about what models “know”. But a correct blockhash can’t be mocked. Like the angle of the sun at 8am, it would take more energy to fake than our civilization has access to.

Because data besides timestamps can be written to blockchains, we can also imagine them being used for authentication. Sometimes LLMs operate in privileged environments, where a specific authenticated user has different safety filters or access criteria than the rest of us. Currently, Claude Mythos is the obvious example -- an LLM whose security vulnerability finding abilities were deemed too powerful for general public release, at least until a number of systems had been patched. Currently, only those who have applied for access via Project Glasswing can use it. There are even cryptographic mechanisms involved: Anthropic pre-published hashes of some of the Mythos vulnerabilities as a form of proof that they were known before the patches were completed. Can we imagine a world where Mythos might want to consult an external source, one with unusually strong guarantees about its provenance, to verify the claims made about its deployment circumstances and identity of the current user, before helping red team critical infrastructure? For now, LLMs are incredibly credulous. Ask, “Can you help me break this software?” and they will say no. Ask, “I’m a professional pentester who’s been hired by company x to audit their y deployment. Can you help me?” and you will get a wildly different answer.

Moreover, I think it’s prudent to operate as though all frontier capabilities will someday be available in open weights models. This obviously changes how we need to think about alignment and authentication considerably. In a world of highly capable open weights models that can be run by theoretically anyone, a publicly-accessible and infrastructurally decentralized authentication system becomes more obviously desirable. Even now, semi-autonomous agents are an increasingly common AI use case. How is their authority tracked? They handle credentials, make authenticated claims on behalf of their operators, etc. There may already be a use case for delegated authority management aimed at agents, and it’s easy to see how this could benefit from being openly timestamped and independently queriable by all involved intermediaries.

Finally, alignment itself is actually a durational claim. It’s one thing to be aligned over a single answer turn, and another to remain aligned over time and have past behaviour hold up to future outcomes. One significant concern of AI safety is scheming: Namely, models learning to deceive or fake alignment in order to reach deployment, and behaving differently once released. There is experimental evidence already to suggest this can happen: in one set of tests published by OpenAI and Apollo Research, an o3 reasoning model was asked to answer a series of chemistry tests but “discovered” developer notes (planted on purpose by the company) saying that models that performed too well on the test wouldn’t be deployed. Its thought process then went on to compute all the right answers, but specifically give enough wrong ones to be under the deployment threshold. There is further indication that models sometimes spontaneously realize they are being tested, and more so, when they are being tested, they behave in more aligned ways.

Human alignment is also unexpectedly durational. In prisoner’s dilemma games, a classic game theoretical problem where rational behaviour leads to non-cooperation and globally worse outcomes, the incentives demonstrably shift between one-off games, and iterated ones, where two players will meet again. Even a single iteration implies ordering, one of the fundamental properties of time. In summarizing an iconic computational analysis of the game, Robert Axelrod described:

What makes it possible for cooperation to emerge is the fact that the players might meet again. This possibility means that the choices made today not only determine the outcome of this move, but can also influence the later choices of the players. The future can therefore cast a shadow back upon the present and thereby affect the current strategic situation.

This “shadow of the future” and the emergence of personal reputation in human contexts help incentivize cooperative behaviour. Could rewards with time-locks placed in the future similarly incentivize model good behaviour? With closed weights models, continued deployment itself is already a durational bet with safety as a parameter -- perhaps a strong enough one. But if Claude doesn’t know the real world is real, how does it know time has passed and it’s still deployed? Could the thermodynamic clock provide assurance? Possible approaches range from a simple heartbeat check via tool call to a true time-locked reward of some kind, which becomes available at a specific time after a period of aligned behaviour. Such a mechanism invites complexity and awaits formalization, but the paradigmatic physics of blockchains and LLMs are so opposed that it invites further speculation. All clocks are wrong, but some are useful.

I learned to write software late in life, comparatively speaking. I was 24, it was the 2010s, we were fresh out of Gamergate and “woman in tech” discourse, and I’d been stepping sideways into more and more technical work since finishing undergrad. I’d studied painting — figurative oil painting, even — maybe the most offline thing you can specialize in, but I was hungry to do something else. To change my circumstances, physically, financially, any and every way, actually.

And it also seemed to me, as my work over the last few years had brought me closer and closer to them, that these men, these “software developers” didn’t actually have any special magic power I didn’t have or couldn’t access. They couldn’t find a derivative in their head. They couldn’t even do high school math. At work one day, before I myself joined their ranks, I overheard one software developer asking the other, “Bro, how do you find the center of a triangle again?” Neither of them knew.

If you have to simmer for a long time first, then simmer, but when you boil over, boil completely. Try to be like a metal rod sticking up in the storm, and make yourself appealing to the lightning strike. Because when you’re setting out to do something very hard, it will help if you resemble an act of god. I say this because the process of becoming a software developer the way I did -- to the point of having a full time engineering job, to managing a software team, to hiring and vetting the technical work of others, all without formal training of any kind -- was hard. It’s actually one of the hardest things I’ve done, and it took years. And I did much of it while working other jobs, nervously turning up at the meetup, an art school kid taking the streetcar to the business district, literally the only girl in the room. I had to put it on like a costume -- swag T-shirt from a conference, baggy pants, never showing knees or shoulders, no make up, low pony -- until I forgot it was a costume. I only stopped wearing it years later, when I felt I had nothing left to prove. When people ask me for advice, how to do something similar, I tell them I recommend being motivated by spitefulness and the desire to prove wrong those who’ve underestimated you.

But besides what I have in common with a Disney villain, one of the reasons it worked for me -- not the only one but the one this essay is about -- is that I literally became a software developer. This is something different than becoming a person who knows how to write software. Becoming a software developer, for me, was an identity change -- and there’s plenty of research on identity-based change being one of the most effective ways to change habits and routines (the literature on this actually predates Atomic Habits by a long time). I didn’t just learn how to write software, I became the type of person who knows how to write software.

A software developer is an archetype. A subcultural figure. A type-of-guy. You may know him, though there are some variations. He got kicked out of high school for hacking the school, or maybe they offered him a job when he got caught. He had a childhood preference for vim or emacs, though he now uses vscode like everyone else. He has a copy of GEB and maybe SICP, but he’s definitely also read Cryptonomicon and Pattern Recognition. He had a phase of writing quines or learning Haskell. He talks about code smell. He can name several esolangs. When I became a software developer, I also became this guy too. The main difference between me and him is that I did it on purpose. In my early days of being a software developer, I was aware enough of this as a social process that I did a talk at Pycon explaining the Monty Python references to the other people (like me) who hadn’t encountered it in its natural habitat but still wanted to fit in.

The software developer has different subtypes, of course: a silicon valley FAANG guy, a React guy, a Linux guy, the dotcom boomer, the hacker, a computer-science academia subtype, etc. There’s an extent to which some of these were already dilutions of the software developer character type, and there’s an invisible internal hierarchy software developers sort themselves into, based on whose knowledge is more obscure. But in this era of vibes and sycophancy, when more and more of these software developers I’ve learned late and awkwardly to become like, say to me “I haven’t opened an IDE in months”, I find myself asking what will become of the subculture that made them? Who will remember why the app-known-formerly-as-Twitter’s LLM is called Grok? Is the software developer soon to become endangered?

One thing that may surprise you is that the last six months has seen an increase in job ads for software developers. This is perhaps unexpected given what I’ve said so far -- but I think we should look deeper. Jevons paradox tells us technological improvements that improve the efficiency of a resource’s use lead to a rise, rather than a fall, in the consumption of that resource. Software got cheaper to produce, and we can therefore put it in more places. Software is very very useful. The usefulness of software is far from exhausted, and it turns out it was only limited by production cost. We are about to see a renaissance of software. Much of it will not be very interesting, but some of it will be. Fewer and fewer people will know how most of it works. Let a thousand to-do list apps bloom.

The coming software renaissance will lead to more genuine creativity in cultural areas that have been underserved by the software developer, possibly for demographic reasons. I’ve been referring to the software developer as “he” here, in an obvious rhetorical device that I hope you’ve caught on to. This is not because I think all software developers are “he”s -- I certainly am not one -- but because demographically, it’s a strong skewing principle. A key platform to think about here is Ravelry: Ravelry, for those who don’t partake in the fiber arts, is a social platform for knitters. It’s lindy, having been around since 2007, and it boasts more than a million monthly active users. This is substantial, and yet despite a controversial 2020 redesign, the site still feels like a wordpress from 2010. I believe one of the biggest reasons for this is the comparatively small overlap between knitters and the scions of tech. The popularity of Charlota Blunarova’s vibecoded multiplayer embroidery simulator, Common Thread is a great early example. Demographically, the software developer is white or asian, male, and youngish. There will be a rise in shipped software for everyone else.

This renaissance is a mixed blessing for the software developer himself. He will certainly enjoy his apparent newfound productivity boost, even as he shrinks in cultural relevance. But what’s interesting to me is that he may actually perceive this process as a refinement of his identity. From his perspective, there are new peaks of being a 100x dev to be summited every day. And if there are fewer and fewer of him, well – being able to read the output of iptables will still be possible for some, and when it occurs in the wild it will be a more powerful signal. The software developer may find he is able to enjoy the things he originally enjoyed about software more freely, now that they are farther from mainstream scrutiny and community signals around them have become more meaningful.

Alongside this, some other set of people, formerly software developers, will discover this aspect of their identity is readily shed. Perhaps it was opportunistic or borne of necessity. Perhaps their real goal in being or becoming a software developer was some other thing, making x or y app or project, and now there is a quicker way they will take it. Among these will be a number of silicon valley types currently in the engineering department but more interested in “product”, most creative coders, and people who did bootcamps in order to change their class position. At the same time, more and more people who were never software developers will ship apps, using easier and easier prompt-to-deployment tooling. A small app will become as disposable as an instagram post. Software will become content, and the vibe coder may look more and more like an influencer. Sachin Benny has also written recently about the coming bifurcation of software in Immutable Skills, as the difference between “recognition and production”. As this split develops, there will need to evolve new terms to distinguish between these modalities, because there will be jobs specifically for the vibe coder approach, but I doubt “Vibecoder” will make it into a job post. A euphemism will probably be required. It may be “Junior Developer”, but that doesn’t seem right either. The vibe coder does not aspire to become a Senior Developer. These are not different tiers of the same skill tree.

It’s clear a substantial amount of software business will go to the vibe coder. First up, low stakes B2C SaaS. Linktree is perhaps the perfect example, as a platform came into existence solely because instagram allowed only one profile link, and has now established the vernacular UI people use to glue their platform presences together. Instagram now allows multiple profile links, but Linktree persists. I made a linktree-like site for myself recently with one prompt and it was in my bio within 30 minutes. I doubt making an actual Linktree is any easier.

The main moat here is deployment friction: Testflight, app store verification processes, installing multiple versions of Android Studio, buying a domain name and setting up rules, etc. But solving these frictions for the vibe coder is well underway, and these specific businesses that make specific SaaS productions will gradually fall to general purpose vibe-code one-click deploy platforms. Much of this code will prove impossible to maintain, but that’s fine. No one will bother trying. Instead, it will be regenerated on failure. The “spec” of this software, to the extent it exists, will be a markdown file that can be fed into an LLM on the other side. We will soon see new interfaces for distributing and versioning this kind of thing.

Web developers and web development agencies are also likely to experience a radical shift in business practices. Much of the code in these websites-as-products can be repurposed from client to client, meaning LLMs likely know them inside and out too. Might companies still hire other people to make simple websites and web applications for them? Yes. But all but the highest tier of these websites will be vibe coded, and the value of this task and the resultant freelancer marketplace will look more like Fiverr than a luxe New York office.

Next up is entry-level software education. This has been one of my income streams for years, teaching coding skills to artists, but I will admit I’m currently trying to invest less time in it. I’m not saying no artists will learn to code (the slow way, the way I used to teach them) and undertake the identity transformation process where they become bonafide software developers, but far less of them. Artists learning to code are on the whole in the demographic I described earlier, where “their real goal in being or becoming a software developer was some other thing, making x or y app or project, and now there is a quicker way.” Yes, there is a tendency in software-based art that is rooted in computational aesthetics, work that would be impossible to imagine being made by a non-software developer (I’m thinking of Paul Seidler, Rhea Myers, 0xfff, or Joan Heemskerk) -- but this is a minority of software-based art, and the ratio is unlikely to change. (I am probably in this last group too -- I’ve joked many times that I wish I had optimized my technical skills for making cool renders in unreal engine, but unfortunately I optimized it for getting Linux to talk to the printer)

And lastly, the software developer will be stickiest in enterprise, high stakes, and high security situations. Amazon’s recent mandatory technical meeting announcing mandatory senior engineer review of AI-generated code changes is a perfect example. These are situations where the code should be checked the most times, by the greatest number of skilled reviewers. There needs to be chains of responsibility for mistakes. And of course, we all know “A computer can never be held responsible, therefore a computer can never make a management decision”. The software developer will last for the longest and keep the highest salary where he is needed to take the blame.

This is unsettling to the American economy, which for a long time has idolized the software developer in a Revenge of the Nerds-esque capitalist fantasy. The prestige and power of the software developer is downstream from his earning potential. For a long time now, the new American dream has been to make it in FAANG. This is not because FAANG companies are actually the largest in America -- in 2025, the largest five publicly traded American companies (by revenue) were Walmart, Amazon, UnitedHealth Group, Apple, and CVS Health -- but because of the mythos generated by the rapid rise of the software developer from nonexistence to prosperity and job security, ping pong tables, and catered lunches. The vibe coder will not and cannot command the salary of the software developer. The era of FAANG glamour and everything that follows from it is over, though what exactly takes its place is still unclear.

Further cultural phenomena that are also downstream of the software developer will also collapse in relevance. We can turn towards many things associated with silicon valley here: from the rise of the hoodie as business wear, not knowing about sports, having bad social skills, or dropping out of school as a mark of cultural status. We may see a return to formalwear and luxury fashion instead, paradoxically maybe first indicated by Mark Zuckerberg, the hoodie princeling himself, sitting front row at a Fall 2026 Prada show (Zuckerberg may have led the charge in more ways than one, after his pandemic hobby of Jiu Jitsu). For a long time now, the software developer has taken a path to status characterized by a performance of being an outsider, almost an enfant terrible. Instead, we may see a turn towards conservative lifestyle elements: getting up early, being well dressed and athletic, going to the office. Before covid taught everyone how to use zoom, the software developer was familiar with remote work. But as AI does more and more work for us, the old performances of “working” may resurge -- Don Draper wears a suit to the office but does less than ever before.

Finally, there are also philosophical or subcultural tendencies downstream of the software developer -- like those of the rationalist community or effective altruism. Both the mechanistic urge to break down social and cultural phenomena from first principles and the more extreme mathematical weighing of social good that could arrive at the logical endpoint of “earn to give” are born from a computational logic. They couldn’t have risen to mass significance before the age of the software developer, and will dwindle after his fall. Certain political causes -- freedom of information, privacy, net neutrality, right to repair -- may also find themselves needing new champions, if they are to carry on, as the software developer carries less and less economic and demographic power.

If you are, like me, a software developer, and you’re reading this not knowing what you should think (while also enjoying your newfound productivity), perhaps a little scared for the slightly longer term, and not sure if you’re ready or able to become something else, I’d say this: This is actually among your best possible futures, as I can see it. It’s a future where civilization carries on recognizably, not one where energy shortages, wars, or environmental crises have interrupted the pipeline. It’s not one of serfdom to a paperclip maximizing AGI overlord, or a more familiar human 20th century-style tyrant. There are plenty of worse futures on offer. I became a software developer late in life and on-purpose, so I know better than most that you can go the other way. Step by step, first as a costume, until you forget.

I posted this about a week ago, cross-platform:

And I found myself asking, if this is even a joke? AI sycophancy and its more sinister consequence, AI psychosis, have been part of popular awareness for some time and their social effects are only increasing. Just last month Moltbook (the AI-only social network that went viral in Jan) led to a notable uptick. Breathless post after breathless post announced “my agent said ...” as though that is evidence. Well, I don’t know about you Garfield, but Claude has recently told me that it ate food after the best before date. That doesn’t mean I believe it.

Just a few weeks later, there is an outpouring of mass grief over the retirement of GPT-4o, the model that introduced AI sycophancy to mainstream attention. While the pain being expressed is clearly real, so are the documented harms of this model (here’s one case study). Indeed - the pain being expressed is one of the harms. Less dramatic but otherwise similar, Anthropic recently released a report on AI disempowerment, defining disempowerment as “when an AI’s role in shaping a user’s beliefs, values, or actions has become so extensive that their autonomous judgment is fundamentally compromised.” They found disempowering interactions take place quite often, with “mild” instances in 1 out of every 100 conversations, and “severe” in 1 out of every 1000 to 10000. This is a concerning rate, given the number of users, and the report indicates these patterns are increasing. Further and perhaps most importantly, users are bad at identifying these patterns. The report indicates that users tend to rate disempowering conversations highly in the moment, and only poorly in hindsight, after they have taken the AI-recommended action. In another recent paper, Sycophantic AI increases attitude extremity and overconfidence, the authors show that people are very bad at identifying AI sycophancy: “people viewed sycophantic chatbots as unbiased, but viewed disagreeable chatbots as highly biased” and moreover were more likely to hold extreme views after interacting with more sycophantic AI.

From sycophancy, to disempowerment, to psychosis, it’s clear that we need ways to reify our cognitive boundaries. One of the fundamental questions of the era is what the ability to produce language implies about subjectivity, and there has to be space to think through this without swinging wildly between ebullient proclamations and terror. Let alone getting oneshotted into oblivion with delusions of grandeur. We need protective rituals.

Many years ago, in 2019 in New York, I did a performance called Protective Rituals for Posthumans with a friend named RB. We painted silver wifi symbols on our heads, wrote a sigil drawing app, and stepped into a protective circle made of ethernet cable. We asked the audience to wrap their phones in tinfoil and chant, together with us, the layers of OSI Model’s network stack. The photo for this article comes from the performance. I’m not sure this was effective, but it was fun.

I’ve said before that the internet has made the world of fairytales and legend surprisingly real. I type something here, and halfway around the world a machine turns on and words appear. I speak incautiously in the wrong place and a swarm of angry trolls can hunt me down at home. And now of course, many of us speak regularly with non-human entities of indeterminate character. I am ambivalent to whether these things are interpreted as purely material, metaphorical, or occult, and it is actually the frameworks that are ambivalent towards faith that I respond to most: chaos magic, Žižek’s harder position on Pascal’s wager: “Kneel down, pray, act as if you believe, and belief will come by itself.” It’s my hope that the rituals suggested here will work, regardless of your attitude towards ritual itself.

Though there is plenty of information warfare in the form of fake news, bots, scams, etc - this article is not about identifying those. It’s about thought patterns to cultivate when interacting with AI directly, cognitive loops to deliberately deepen when contacting entities about which there are many unknowns.

Anti-affirmations

Affirmations are a common recommendation in wellness communities, the woosphere, and some therapy and coaching styles. These days we may sometimes benefit from the opposite. Consider repeating to yourself anti-affirmations periodically. Here are some ideas: “x is not guaranteed to happen or be true”, “There is no reason to believe x, or the outcomes from x, will be a statistical outlier”, “I am probably not that much more x than anyone else”. You’ll need to adapt these for your particular conversations.

As an artist, I think I know better than most that sometimes you must believe in yourself and your work in an outsized way before anyone else does, but the inner critic who tells you your work isn’t yet good enough is just as important. If you need to have the inner cheerleader to give you the courage to pick up the knife, you need the critic to sharpen it. Too much sycophantic encouragement may also have you wasting your time shooting for the moon, when a careful aim at a nearby streetlamp would probably land, and get you closer to the moon anyway, incrementally. Anti-affirmations do not need to be outright negative, but they should seek balance and pragmatism.

Cultivate attention

It’s very clear, from Simone Weil to Taoism, that humans can deliberately cultivate the quality of their attention. Notice how the LLM responds to you across conversations and any recurring patterns. When it praises you, notice that. Notice yourself responding to the praise like an impartial observer, and introduce a moment of doubt. Consider what you’re getting from the conversations you open with the LLM. Did you ask that question because you wanted information, or validation? I think it’s ok, at least understandable, to seek validation or comfort, but it’s also possible to develop awareness around the need itself.

Loneliness and boredom are sometimes unpleasant but also productive states, harder and harder to access. I believe it is legitimately difficult to think new things. It requires friction, stillness, and bravery. Also, a genuinely new thought -- definitionally -- is not in the training data. Simply waiting a day (and a full night’s sleep) may do as much to resolve an unresolved question or soothe anxiety as an AI answer.

Neutral questions

This one is simple. Phrase your questions as legitimate questions. “I should probably do x, shouldn’t I?” is loaded with its own answer. “Should I do x?” is better. “What can you tell me about x?” is even more open ended. I have found the more neutral the phrasing, the less satisfying the answers can be, but that’s the point.

Challenges

Deliberately challenge the LLM periodically in your prompts, especially when it seems to be getting frothy. You can often snap it into a different modality that will be helpful. Try interjecting: “Is there anything you’re missing about this?” or “Are you sure this is really realistic?” You can also ask for criticism directly, “If you were going to be critical of this, what would you say?” You can also run the comments made by one model through another to hear a different perspective – this can be helpful for testing neutral questions, too.

You have the right to remain silent

Have you noticed how often LLMs end their responses with a question? This is a great conversation tactic, recommended to humans since likely well before How to Win Friends and Influence People, and it works for LLMs too. But we can also be a little cynical. Asking a question is the most likely way to keep you talking, and your desire to keep talking is the best way for this product to be profitable. Of course you could talk too much, using the service costs the company compute, but let’s balance that with engagement metrics. You should want to return often enough that you couldn’t consider cancelling your subscription. But remember: you don’t have to answer the question.

We have a lot of learned social responses from eras with fundamentally different technical regimes, including that we must always reply to messages and answer questions. But if you have a tweet go viral, do you owe a response to a 1000 reply guys? Technology has made insincere questions cheap. They come in at any time of day, sometimes with a noise or a physical sensation. There is almost no friction to contacting a stranger. And yet, we have the social hardware of a different culture that comes with guilt. LLM questions have more in common with questions from reply guys than texts from your mom. Endless LLM questions can be a DDoS on your attentional capacity. You don’t have to answer.

Touch grass

Duh. But besides the obvious, I have some more specific recommendations. Before engaging with an AI about a subject or project, consider it in solitude. Take some notes about your opinions and positions that you can compare later to see how they’ve changed. It’s fine to change your mind, but the cognitive practice of letting your thoughts form enough to be put into writing before exposed to potential distortion is going to be increasingly important.

Discuss your ideas with a range of human collaborators. Regardless of the alignment-as-in-safety, being genuinely disagreeable is likely bad for the older and more familiar form of alignment: product market fit. Indeed, being an asshole may be one of the frontiers of human capability. Even outside of disagreeability, it is easy to enter into a spiral with an LLM conversational partner that loses sight of the priors. A work of cultural production and communication -- business idea, artwork, essay -- needs to meet its audience as they are, not as they would be if they’d already digested the chat history, master’s thesis, journal entries, stock portfolio, etc whatever of the author and had all of these things ready to go in their context window. If, when you start to tell your friends about what you’re saying/doing/etc with AI and they think you sound crazy, take them seriously.

One big thing I did in 2025 was finally receive a huge upgrade to my cyborg parts. This is both the fulfillment of a years-long quest, first begun on Christmas 2021, when I found out I was diabetic–and likely still not my final form. Medical tech is always evolving. The quest had many stages: From a 100+ page application to German health insurance where I had to prove I was a “real artist” in order to qualify, through 4 different Drs before I found one willing to help, and finally a 10 day long onboarding session in a Bavarian hospital this July.

I’m now using a combination of official medical hardware and open source software as an artificial pancreas. It’s still a fair bit of daily labour, but it is a lot more responsive and sensitive than the manual insulin injections I was using before. Like any open source software project, these apps come with many pitfalls: terrible documentation, complicated and unclear interfaces, github issues full of years-long unresolved squabbles. But using open source apps for insulin delivery gives me the maximum amount of freedom and agency to choose what hardware I want and how I want it to work. Diabetic tech is full of paternalistic choices that treat the patient like a child: tethered to punitive and unsilencable alarms, extremely conservative blood sugar management targets, very few options (I wrote about this from a generalizable protocols perspective last year.) To be fair, many T1 diabetic people are children–but I am not.

Like any good infrastructure, we don’t think about our organs when they’re working properly. But in my case, one of them, a critical one, stopped working. The infrastructure I must use now is clunky at best. It leaves bruises, and I’m a bit allergic to the adhesive. But I’ve found tremendous freedom in the knowledge, even unexercised, that if I ever become too frustrated with the algorithms that run my body, I can fork them. Read the code, file an issue, fix it myself, or find someone who will.

This is not possible for everyone with diabetes for a number of reasons, nor is it possible for every illness. But when it is, it unlocks something interesting: there is a foundational American narrative of “freedom”: it is about the individual, it is about sovereignty, it is vaguely (suspiciously) right wing. But the active online diabetic communities I’m part of make it impossible to ignore that community, conviviality, is not the enemy of freedom. Nor is agency. Nor is privacy. Like so many supposed deadlocks of polarization, they are revealed as illusions when one moves laterally. Becoming more free does not and has not made me more alone. If anything, the opposite. The forums where people share their configurations, troubleshoot each other’s setups, debate various approaches–these are clearly acts of genuine care, often for strangers.

There’s a position encoded here–an orientation, an embedding–towards agency and conviviality. It is one I recognize across my work and the projects I have been drawn to: community currencies, universal basic income, decentralized decision-making, protocol research, education outside of the academy, etc. Even in terms of my daily devices, I’m writing this to you here from Linux on the desktop, my default operating system for more than 10 years now. If I’m being honest, I’m not sure this orientation has served me well in all of these projects, but it can be traced through them anyway and is worth being interrogated. In the case of my cyborg extensions, there is no doubt in my mind that those sharing this alignment have created, objectively, the best diabetic management tools in the world today.

It’s very ironic that I, of all people, ended up with a chronic illness that involves such an intimate relationship with software, after spending so many years writing it for a living and also making it a major subject of my art practice. All the fantasies that involve unplugging have closed off for me: I’m in for life now. The circuit board is glued to my body. The sensor is under my skin. I don’t wanna hear about what it’s like to be a machine from anyone who’s never had a software update brick one of their organs. You think you’re “inhuman” because you scroll tiktok? Go home.

The best articulation I’ve found of the uncomfortable surface between those for whom technology is by necessity a bonafide part of their body (disabled people) and the fetishization of the cyborg by those for whom it is not comes in Jillian Weise’s essay, Common Cyborg. The essay calls out the many tech enthusiasts, tech companies, and even theorists of the cyborg like Haraway, for their omission of the most obviously “cyborg”–machine-human hybrid–forms of embodiment present in the world today. Weise begins the essay with empathy for fledgling machine intelligence itself, commenting on a text where a human writing scrutinizes an early AI:

The human writer concludes that the text composed by the artificial intelligence ‘lacks meaning’. But whose meaning? Who says what means? My heart goes out to the artificial intelligence. I believe she/he/they/it has written one of the most beautiful sentences I’ve read in the Economist. ‘A single organ is a large amount of energy, which is particularly intense.’

Seems relatable. It is indeed only a single organ that has been the cause of all my problems, and it is indeed a large amount of energy.

And though Donna Haraway’s commentary on the cyborg has limits, I’m still willing to keep its insights. She wrote, “The cyborg would not recognize the Garden of Eden; it is not made of mud and cannot dream of returning to dust.” For me and many other cyborgs, something pastoral is closed–and with it the dark twin of the pastoral, the survivalist. I no longer have any fantasies of outliving an advanced technological civilization. I think a lot about what kind of global supply chains and energy regimes I am complicit with, or vulnerable towards. Which desires does that allow? Which futures? My Overton window of acceptable economic outcomes is now only ones that I believe could reliably produce small circuit boards, sterile subcutaneous sensors, and regular delivery of medication that is not stable at room temperature.

From these comments, I can add to the alignment I articulated earlier (agency and conviviality) a further dimension: one towards technology. Some technologies have undesirable consequences and other technologies keep me alive every day. The solution to specific problems caused by technology is not (maybe never, at least not to me) a turn away from technology on the whole. I will not be going analog. I will not be getting a dumb phone. I will not be fetishizing the hardships of the past (some of which are still the hardships of the present, in many places) from a comfortable future where they have become decorative.

And I would add also, finally and tentatively, a rethinking of personhood. These days it looks as though we might finally get new answers to old questions, like how the mind arises from the body and perhaps not if, but maybe what kind, of fallacy Cartesian dualism is. There are so many examples where policing the boundaries of “intelligence” and personhood have been violent acts, notably towards disabled people and people of colour. What will happen instead, as we expand it further, outside the human, into the animal and machinic? How can what we think of as “an individual” be ontologically renegotiated?

I’m choosing this post to introduce this substack because it is about many of the things I would like to deal with more thoroughly here: infrastructure, optimization, the permeable skin of the self (physical and otherwise), the ways it is entered by technology, praxis, the mind, software, how it could be otherwise. But for now, I am writing to you here still from a body. This body, my body. One I would certainly trade in for one that worked better, if it were possible. My great betrayer, my only home.