Prompted LLLM output only involves brains on the input side, the source training data. Nothing is communicated to the end user brain because they don’t try to understand the output and never had the knowledge or experience necessary to do so if they wanted to.

Now that it has, I'm overjoyed that my work has had some tiny part in training the models that are helping millions of people implement their tech ideas or even just write some text.

"The machine superintelligence learned from your ideas" is awesome sci-fi, not a reason for pearl-clutching. To stop publishing purely because you don't want AI to learn from you is misguided at best and misanthropic at worst.

But to me there's a strong difference between programming and non-programming things re: AI. I don't feel the same way about any other subject to write about.

It's not about wanting to keep things for myself in an ownership sense.

It's great that there are people that enjoy being a training bot for a corporate product, but some people still want to talk to other people. And if a system ever becomes an AGI, I am fairly sure it will no longer need any human input, so our days as training bots are likely numbered. Enjoy it while you can I guess.

FWIW I feel more or less the same _and_ I think this gives society a solid ground for demanding some sort of collective compensation from AI labs, just like mining or oil companies pay for natural resources https://distantprovince.substack.com/p/the-past-is-a-natural...

You do you, of course, but don't call it pearl-clutching if people have different worldviews.

I can also editorialize.

In my view, Trump is leveraging the shadow docket to get decisions over turned by members of his own party who don't need to provide detailed (or any) explanations as to why or information about who voted for what.

http://www.incompleteideas.net/IncIdeas/BitterLesson.html

Recent comment touching on this in relation to LLM's in more depth: https://news.ycombinator.com/item?id=49322695#49323341

Commenter below gets it absolutely correct: stockfish, which runs on your 5 year old phone, is dramatically better at chess than Fable. Like, so much better that it’s not even remotely comparable. The theory of the Bitter Lesson, and it’s only a theory, is that LLMs could eventually outperform stockfish. It’s not true today and it remains to be seen whether it will ever be true. For now, specialized models are absolutely better at specialized tasks.

I can't find the comment you're referring to, but the latest versions of stockfish are based on neural networks trained on millions of games, so if anything the Bitter Lesson turned out true here.

Stockfish is the best chess search engine we've got, and you can learn some good heuristics for chess search policy that will make time-limited chess search a lot more powerful. That's perfectly in line with the Bitter Lesson.

In contrast, LLMs playing chess are relying solely on learned behavior. The inference harnesses surrounding them aren't designed to do chess things, they're designed to do autoregressive token decoding, which isn't a search process. Reasoning traces can resemble a search process, but they're far less efficient - the LLM would have to work out each legal move, test each one, calculate a score, and simulate minimax over all of that. Assuming the LLM is smart enough to even do all that.

A hand-crafted approach can absolutely beat data if your approach unlocks more search and/or learning than the general solution.

Now let's look at the bitter lesson again. It says that general methods that leverage computation are ultimately the most effective, and by a large margin.

That's different from just saying to leverage computation (which is how I would interpret "unlocks more search/learning"). If the lesson is "more computation wins, when sufficiently channeled" you're basically looking at a truism. Of course more computation beats less when it's used right. The bitter lesson is about abandoning specialization in order to get more computation, and while there's a couple ways where that helps with chess, there's a lot more ways where it's counterproductive. It looks like it's more true for Go than it is for chess, and that it's not universally true. It probably correlates with the state space.

Not really, if anything it's closer to the opposite. The Bitter Lesson essay literally has this as an example:

> These researchers wanted methods based on human input to win and were disappointed when they did not.[1]

and

> Enormous initial efforts went into avoiding search by taking advantage of human knowledge, or of the special features of the game, but all those efforts proved irrelevant, or worse, once search was applied effectively at scale[1]

The actual bitter lesson is this:

> breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning. The eventual success is tinged with bitterness, and often incompletely digested, because it is success over a favored, human-centric approach.[1]

Applying to the "LLMs-for-chess" example the bitter lesson approach would be to put many, many more games into the LLM.

Does this work? People have trained fairly small LLMs that are competitive Stockfish at the ELO 1500-2000 level, eg: https://github.com/kinggongzilla/chess-bot-3000

This seems to be evidence that large LLMs probably don't have as much chess training data as Stockfish does.

[1] http://www.incompleteideas.net/IncIdeas/BitterLesson.html

This was/is basically a strawman though. Like maybe "human input winning" was desirable for chess masters but for computer science wonks? Not the point or the disappoint. It's always neats and scruffies fighting about using some kind of recognizable method (logic) instead of magic (ML).

> breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning.

More to OP's point I think: nowadays when someone wants to beat you over the head with the bitter lesson, they aren't as careful to include learning and search. They want to say learning leads to intuition (magic) whereby we can avoid work (logic/search), and maybe argue or assume from there that neats and scruffies is settled. TBF, something like reasoning in latent space does resemble intuition!

But the real lesson is confirmed every time we bother to check, and not very bitter for anyone. Search/learning/logic are ALL always necessary on any sufficiently difficult problems, and hybrids that interleave always outperform everything else. Stockfish being the example in this thread that different camps of absolutists would like to claim, but also all the MCTS examples, evolving examples, and new hybrids all the time. My favorite lately: https://arxiv.org/pdf/2511.08983

But as it relates to more/less magic, maybe the more modern lens on this is e.g. https://arxiv.org/html/2505.11581v1 . Is manual feature-engineering more like what you'd evolve, or more like what you'd get from SGD ? Feasibility and performance is always a question, there are others like what is robust, stable, adaptable, predictable, explainable. Maybe the manual-features people were interested in something besides the manual part? Maybe the story isn't so simple, and maybe it's not finished yet.

The common mistake is to think “maybe if we use a blend of raw data and hand-crafted heuristics, we’ll get the best of both worlds!” But the bitter lesson says no, beyond a certain point it’s better just to use the data.

Thinking that an LLM might be able to improve on purely “big data” machine learning seems to me to be the same incorrect idea. Its “intelligence” is no more useful than human intelligence. The LLM is based on a massive data corpus, sure, but the amount of data specifically about chess in there pales in comparison to just playing billions of games of chess.

The conclusion of the bitter lesson would be that a large language model trained on chess commentary as well as being trained on millions of chess games would outperform stockfish which is only trained on millions of chess games.

If you can draw any lessons from chess commentary, I think it’s very reasonable to call it “hand-crafted heuristics.”

And raw games plus raw commentary is all the data you have. You can make more games but those can be fed to both stockfish and the LLM competitor. So it seems like a valid interpretation of the bitter lesson to me.

The key point I was trying to get at is that the human insights don't contain anything that can't be mined from vast amounts of gameplay. Every human insight can eventually be rediscovered and made rigorous by data (in chess, at least!) In the short term, those insights are useful, but in the longer term, they add nothing at all.

Note also that "raw gameplay" here can mean new games -- you can generate as much data as you need, you don't need to rely on real recorded games.

But isn't that the bulk of what we're shoving into LLMs, and it makes them much smarter? If it's useful there but not in a chess AI then that seems like a significant crack in the bitter lesson.

> Note also that "raw gameplay" here can mean new games -- you can generate as much data as you need, you don't need to rely on real recorded games.

Yeah I mentioned that, generated games are useful. But if we're being fair and letting both AIs use generated game data, does the more general LLM ever actually overtake the specialized stockfish like the bitter lesson suggests?

Another way to look at this is that giving the LLM the commentary is a way to avoid complaints of hiding data from the LLM, since it'll have strictly more info than stockfish. But if we cut that from the training data and only give it a basic description of chess and lots of raw game data then it's going to get even worse than it already is at chess. Meanwhile stockfish never had that commentary, just actual hand-crafted heuristics and training on game data, and it's very strong.

The bitter lesson is that simply scaling training on more games—including self-play—trumps any hand-crafted human input, whether that's fine-tuning on human commentary or clever engineering tricks.

Current models are just high-dimensional interpolation engines. The denser the data sampling, the more accurate the interpolation gets. Given a choice between denser sampling and anything else, denser sampling always wins. That is the bitter lesson.

Computer chess is the canonical example of this.

In the case of stockfish, the harness is a tree search around the neural network evaluations.

Just kidding of course

It would be better to compare models at how well they can write the code for chess engines, otherwise it's just saying that Fable is not a good CPU emulator, which is obvious.

The Bitter Lesson is about general-purpose algorithms vs. specialized algorithms. Historically, chess engines were programmed to look at a chess position and use positional understanding (imparted by the human programmers) to decide what the best move is. But eventually, the chess engines that actually became stronger than humans were instead programmed to just check every possible move and countermove and see which ones lead to a win. (I'm oversimplifying, but you get the point.) So even before Stockfish contained a neural network, it was considered an example of the success of the Bitter Lesson.

As it applies to AI agents, the Bitter Lesson would predict that the best possible agent would simply possess A) a way to do anything it wants, B) a way to evaluate whether what it did was correct, and C) a ton of compute. Then just turn it loose on your task. (The fact that the "brain" of the agent is an LLM is kind of irrelevant - you could also imagine the brain just being a program that generates random syntactically-correct code. What the LLM achieves is that, the random generator would take millions of years whereas the LLM is much more efficient at creating plausibly-working code. This is analogous to a chess engine's pruning heuristics.)

The hard part here is B. We've seen some great agentic successes when rewriting an existing project in a new language, since the agent can just use the project's prior test suite as its evaluator. But when developing a new project, you're still figuring out the finer details of how everything is supposed to work. As the old saying goes - writing a spec that perfectly describes how a program should work, is equivalent effort to just writing the program.

And perhaps at the end it all gets a single pass by a god-tier model for overall sanity and congruence, but the actual work, planning, coordination, and even user interaction was done by cheaper and faster agents of much more limited capability.

It really is absurd to ask programming questions to a model also trained about the lifecycle of a fruit fly.

Instead of building small models from scratch, we train an enormous model and use ridiculous amounts of GPU memory. In the end, the whole thing is shoved into RAM because we don’t know where the useful parts are…

We certainly would know where they were if they were just in smaller models in the first place!

The correct analogy here is Fable as a second tier player assisting a SuperGM in running stockfish, then assessing its output to identify promising variations.

There might be a limit somewhere that prevents the bitter lesson being axiomatic - for instance where simulations for anything can be exhaustive - so that judgement isn't needed any more as an arbiter. But while there are problems sufficiently complex or large to require a breadth models don't currently have, greater scale and compute will continue to convert to better decision making, and the bitter lesson will remain true (true enough).

https://mcpmarket.com/server/stockfish

https://til.simonwillison.net/llms/mcp-in-claude-and-chatgpt...

Dumb AIs are needed for customer service. Most of that industry is still at "press 1 for sales, 2 for billing..." and needs something that will run locally on a 1U server.

Needs? Customers want something that immediately answers their question/solves their problem, but that’s far away, even ignoring the “run locally on a 1U server” and that that may not be in the company’s interest.

For many companies, that support line is a cost center, not a PR mechanism.

Also “Press 1 for sales, 2 for billing...” has the big advantage that it handles all accents, speech impediments, etc.

Long term I think a solution where a user’s agent trained on their voice, running on their phone communicates with the support agent of a company is where we will end up, and support phone lines will cease to exist.

GP isn’t suggesting that focused narrow model(s) will be more capable than large model, but that many small focused models can have sufficient capability while being more optimal.

Also, the bitter lesson is just wrong. The bitter lesson is about hand tuned AI vs computational general methods. However in truth today’s AI uses both. We have general compute heavy models which require narrow expert instructions (eg tools internet docs).

LLMs would not be as good without expertly written context, and expert context without LLMs aren’t as good either.

The models are not even really trained bitter lesson-style anymore. That concept peaked during the era of pre-train scaling, back when it was thought that making a bigger and bigger GPT-3 would automatically solve all problems through prompting. In 2026, the most important part of training is post-training, which uses vast quantities of niche, hand-curated data to fit the models for specific tasks in domains like tax law.

If you look at value as purely the LLM output, then there's a valid argument that the best frontier models will always be better than fine tuned specialists. (I'm not convinced personally, but it's a defensible claim)

But that misses two dimensions: 1. The cost of acquiring that output 2. What is actually "good enough" for that specialist domain

Not every output needs to be the best to produce value.

And as specialist models increase in cost, their cost/value proposition goes down.

At some point, there's a threshold where cheaper, fine tuned models are "good enough" at the task and also substantially cheaper than the expert models.

That's where fine tuning helps.

Personally, I became a believer in fine tuning after fine tuning a 1B Qwen model as a second pass over my local voice transcription app, achieving excellent accuracy at ~zero token cost and waaaay lower latency than if I'd invoked my Claude subscription under the hood.

Secondly, the bitter lesson is predicated on compute being cheap. There was a period where a hand-tuned algorithm informed by human expertise would outperform a raw alpha-beta search at Chess. Then compute got cheaper, and DeepBlue ascended to the top. Compute is now expensive again relative to the tasks being performed. We are absolutely still in a period where human expertise in training LLMs will outperform a naive approach with more raw compute.

In the latter case, the chess example would tend to support the Bitter Lesson, rather than refute it.

I would also be VERY slow to claim that general-purpose models will never be competitive at chess. It wasn't so long ago that transformers couldn't add two-digit numbers reliably without resorting to tool use. They are now as good at "mental arithmetic" as any human savant. It wouldn't surprise me at all to see someone come up with a model that just happens to be really, really good at leveraging the portions of its general training data having to do with chess.

In fact you could argue that AGI demands such a model, if we are to assume that LLMs are a guidepost in that direction.

> I would also be VERY slow to claim that general-purpose models will never be competitive at chess.

This is not the claim. The claim is that for the same amount of compute, a general-purpose language model will never beat a Chess model. I'm dubious, but allow for the possibility that a language model could eventually compete at a top level against humans with enough compute. However, it will never compete with a dedicated Chess model with similar resources. Training a model for a specific task with the same amount of compute will outperform training a general-purpose model with the same amount of compute. This should be common sense, right? The bitter lesson was only about compute over human algorithms, not at throwing compute at a generalised domain over throwing compute at a specific domain.

You made arguments against two claims that I did not make (that I was trying to refute the bitter lesson or that I claimed that LLMs could never be competitive against humans at Chess), so I'd like to ask you read my statements a little more carefully this time.

I see arguments like this all of the time in ML. There is a large branch of statistical theory that can cast some doubt on this.

Stein's paradox: when three or more parameters are estimated simultaneously, there exist combined estimators more accurate on average (that is, having lower expected mean squared error) than any method that handles the parameters separately.

This goes deeply against "common sense", but is mathematically proven, empirically proven, etc.

If you are just taking a chess model vs. an LLM, you are right.

But if you want a chess model, a medical diagnosis model, and customer service chat model, at a high level the intuition of Stein's paradox is that one combined model will get you better performance than three task-specific models.

Most practitioners don't get this because what amounts to graduate-level stats theory is often brushed aside, and we keep relearning it the hard way.

https://en.wikipedia.org/wiki/Stein%27s_example

And in 2020 Stockfish 12 adding some NN evaluation. And then in 2023 Stockfish 16 entirely removing the classical position evaluation code.

https://stockfishchess.org/blog/2023/stockfish-16/

Bitter lesson #1: don't waste time optimizing code when a faster processor is around the corner.

What countered it: Moore's law stopped working.

Bitter lesson #2 similarly relies on scaling laws that might have diminishing returns wrt model runtime vs intelligence. Runtime matters for turnaround on the problem you're solving.

However, what most people think of as Moore's Law--CPU speed doubles every 18 months--broke somewhere between 90nm and 22nm.

And even the actual Moore's Law--2x the transistors every 18 months--doesn't hold for all types of chips anymore. Memory only gained 2x density over 10 years.

The actual argument of the Bitter Lesson essay is pretty limited but people's interpretation of it has gradually drifted until it's seen as prediction that current LLM will reach AGI at a large enough scale.

When you're doing something "at scale", you want smaller models, weaker heuristics and rules written in code, as they're much cheaper, much faster, much more predictable and much easier to reason about.

If your task is "tell me which pages on the internet are most relevant to this search query", running Claude Fable to evaluate every single page would be ridiculous. You may run something like it at the end, to synthesize information from a small result set for example, as that's a "human-sized" task.

If your task is "write a function to sort a list of 1mn entries alphabetically", a small coding model will exactly match the performance of a frontier model (it'll 1-shot it, with 100% accuracy). So the only differentiator is cost, and the frontier model will cost orders of magnitude more than the specialised model.

For most intelligence based tasks, you don't (and never have) needed the tool which "performs best at all tasks". You need the cheapest one which performs adequately for your immediate task.

This doesn't mean the bitter lesson is incorrect. At the frontier, it's still correct. It means that it's not applicable at all to lots of tasks.

VibeThinker 3B constitutes extraordinary evidence, IMO. The first such evidence I've seen myself. Very small model, very low literacy, almost no world knowledge, but it is as good at math and logical reasoning as models a hundred times larger.

The Bitter Lesson is a valid and trenchant observation about how about we got here, but I think it's a mistake to assume it tells us very much about where we're going. Too much has changed recently and is still doing so.

Is this whole thing than maybe a read vs write optimisation again? Spent more time and effort training more knowledge into the model upfront and get it out in a single question instead of training a small model and needing more steps to answer the same question?

Any similar model aimed at coding?

A >10B model for mass spawning/swarming and reporting back to a larger model

I wouldn't use it for anything important without heavy supervision, as it's very weak outside its specialty. Not ideal for instruction-following tasks.

    document.body.style.cssText='margin:auto;width:600px;'

Absolutely false. At least when it comes to multimodal inputs, even a simple classifier will outperform the largest LLMs who still hallucinate details or don’t describe audio and images accurately.

And there’s also the issue of cost/inference speed. Running a trillion parameter model for all tasks will be incredibly costly, require a cloud API, while a tiny CNN can be run locally or at a cost multiple orders of magnitude lower.

So there's a "fibred side".. the most likely candidate seems to be "fibred categories" which I hadn't heard of before, and it's talking about one side of some mapping between two sets such that if f is the primary function and f(x)=y then there exists an inverse function g(y)=x? Was it something that converted some data bidirectionally with a different algorithm on both sides?

The capstone of the inverse function would be the most important thing about it maybe?

My best guess is "In the process of working on the inverse function, the existing name (of the inverse function itself maybe?) was made to reflect the operation of the inverse function, so now the name does not follow the same naming convention as the name of the primary function (which does not contain its 'capstone')."

Its original wording is certainly dense and harder to follow for us, but it's fascinating how the model finds this the best fit for what it's trying to express IMO. Like it arrives at its own ways of overloading words/concepts, and things we would refer to in different ways in different contexts all get compressed to the same more-useful/complete idea.

Codex has never said anything nearly so alien as the Claude examples I've seen floating around, interestingly. I wonder if it just has a better training on choosing its words to present to the user or if it inherently arrived at a somewhat different mapping that favors 'plain language' more.

As far as I can tell "the capstone" is what Claude usually calls my current goal if it thinks it is a satisfying result.

I have several similar folders with variants of a construction, but taking differently structured input. They are named “plain”, “fibred” and “indexed”. So the fibred variant is clear enough.

The Claude speak I struggle with is “the capstone” and what name it could be talking about. And what folding means here. I think it just means:

“I changed an important result of the construction in the fibred variant, but kept the name. so the fibred variant is now different from the others.”

What argument? I don't know what to take away other than "Claude likes certain words". Some of them are kind of amusing, but I'm not convinced the vocabulary is bad or that this is a problem, just from looking at this.

LLMs were not taught to say the phrase "load-bearing seam" from humans saying it, because humans have never said it. It's almost definitely an artifact of post-training and nothing more.

There will also be models that make better use of tools, like static analysis and fuzzing, and they will find different defects. Social engineering is going to be a big skill to learn too, as it is a much softer skill.

They do often enough that it's not a surprising event, depending on prompt quality, context available, ability for the result to be objectively judged and iterate on by the agent, etc. For frontiers on very high settings at least.