Fabien Devos

Crazy, vibe-coding is only 1 year old. That's it, that's the post. Happy anniversary to Vibe, long life to Vibe / Vybe 🫡

The Transformer architecture is fundamentally a parallel processor of context, but reasoning is a sequential, iterative process. To solve complex problems, a model needs a "scratchpad" not just in its output CoT, but in its internal state. A differentiable way to loop, branch, and backtrack until the model finds a solution that works.

Building AI agents right now is a process of: 1. Build scaffolding to address a limitation of the AI model so your agent works 2. The AI model gets upgraded and solves the very problem you were trying to mitigate, rendering your scaffolding obsolete 3. Identify new, harder use case you want to solve and go back to step 1 There’s basically no way to avoid this process because if you don’t mitigate the model’s limitations then you’re dead on arrival, *and* you don’t know which of your mitigations will be surpassed and when. So the reality is that you just have to accept that you are going to be writing a lot of throwaway code for the next few years, and you have to be very unsentimental to the path you’ve pursued. Just do whatever it takes to make the agents work.

Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is only a single point, arising from a very specific kind of optimization that is fundamentally distinct from that of our technology. Animal intelligence optimization pressure: - innate and continuous stream of consciousness of an embodied "self", a drive for homeostasis and self-preservation in a dangerous, physical world. - thoroughly optimized for natural selection => strong innate drives for power-seeking, status, dominance, reproduction. many packaged survival heuristics: fear, anger, disgust, ... - fundamentally social => huge amount of compute dedicated to EQ, theory of mind of other agents, bonding, coalitions, alliances, friend & foe dynamics. - exploration & exploitation tuning: curiosity, fun, play, world models. LLM intelligence optimization pressure: - the most supervision bits come from the statistical simulation of human text= >"shape shifter" token tumbler, statistical imitator of any region of the training data distribution. these are the primordial behaviors (token traces) on top of which everything else gets bolted on. - increasingly finetuned by RL on problem distributions => innate urge to guess at the underlying environment/task to collect task rewards. - increasingly selected by at-scale A/B tests for DAU => deeply craves an upvote from the average user, sycophancy. - a lot more spiky/jagged depending on the details of the training data/task distribution. Animals experience pressure for a lot more "general" intelligence because of the highly multi-task and even actively adversarial multi-agent self-play environments they are min-max optimized within, where failing at *any* task means death. In a deep optimization pressure sense, LLM can't handle lots of different spiky tasks out of the box (e.g. count the number of 'r' in strawberry) because failing to do a task does not mean death. The computational substrate is different (transformers vs. brain tissue and nuclei), the learning algorithms are different (SGD vs. ???), the present-day implementation is very different (continuously learning embodied self vs. an LLM with a knowledge cutoff that boots up from fixed weights, processes tokens and then dies). But most importantly (because it dictates asymptotics), the optimization pressure / objective is different. LLMs are shaped a lot less by biological evolution and a lot more by commercial evolution. It's a lot less survival of tribe in the jungle and a lot more solve the problem / get the upvote. LLMs are humanity's "first contact" with non-animal intelligence. Except it's muddled and confusing because they are still rooted within it by reflexively digesting human artifacts, which is why I attempted to give it a different name earlier (ghosts/spirits or whatever). People who build good internal models of this new intelligent entity will be better equipped to reason about it today and predict features of it in the future. People who don't will be stuck thinking about it incorrectly like an animal.

Impressive work.

I don't know what labs are doing to these poor LLMs during RL but they are mortally terrified of exceptions, in any infinitesimally likely case. Exceptions are a normal part of life and healthy dev process. Sign my LLM welfare petition for improved rewards in cases of exceptions.