François Chollet

The tweet below got enormous pushback at the time I posted it, more than 5 years ago. But this prediction is looking obvious by the day now.

I don't read/write code these days, I only instruct an LRM. But it's not because I think LRM code quality is perfect, or even good. And it's not because I think my instructions are always being perfectly acted upon -- I know they aren't. The thing is: LRMs don't just write your code, they unlock new ways to control what you're doing and maintain understanding of your system, which are about as effective as reading code. You can better test your codebase, you can audit specific components, you can spin up visualizations, you can ask the LRM to red team, etc. You can do all of these things at high speed -- you couldn't before. The benefits of reading/writing code can now be achieved in new ways. All in all, the ROI of writing code by hand is no longer looking good, not because LRMs got perfect but because they're so fast at doing things that you can develop new workflows around them, workflows that end up being more productive than the old workflows.

The purpose of the AI industry should be to produce tools that, in the human hand, will improve human prosperity and welfare. It should not be to create a "successor species" to the human race. Entertaining such a thought makes you, de facto, the enemy of all present and future humans.

I think the "difficulty" of software engineering is essentially constant no matter what abstraction level you move to, because human cognition adapts to new tools until it can fully utilize itself. Tools are only affordances, not a magic wand that makes work disappear. Great software engineering was immensely challenging before. It is still immensely challenging now, despite very different workflows.

Evolution doesn't benchmaxx. It optimizes for generality and robustness, because overspecialization makes a species fragile when the environment inevitably shifts.

At some point in evolutionary history, matter learned to feel. We don't know how. It wouldn't be scientific to dismiss the possibility that it could happen again. (I don't believe it happens in current AI systems, because they don't have the charactistics I would expect to see in a sentient agent)

What is a way to test human-level generality in artificial intelligence? "The meta-benchmark of being able to pass ARC-AGI-(n+1) immediately upon release"

A lot of messaging around AI (by its proponents) enthusiastically frames it as a form of destruction, obliteration even: "AI just killed XYZ" - "It's over for XYZ" (in reality, this is practically never accurate) The way regular folks perceive AI is very much shaped by such takes. The backlash is inevitable.

A GNN library built natively on Keras 3 -- with models running on JAX, torch, TF with full hardware acceleration (Apple Silicon, TPU, etc.). "K3-Node achieves 100% public API parity with PyG and incorporates state-of-the-art foundation models and architectures from Spektral and StellarGraph" https://github.com/anas-rz/k3-node

Do I know anyone who knows anything about cyanotype printmaking?

The critical distinction between base LLMs (2024 and earlier) and modern LRMs is not symbolic tool use. It's the switch from a transductive paradigm (intuit the answer to the query) to an inductive paradigm (intuit the program/instructions that produce the answer to the query). They're trained to be inductive, and they perform test-time induction, i.e. test-time prediction of a NL program / reasoning chain. This unlocks entirely new capabilities -- in particular fluid intelligence. Base LLMs, to this day, have ~0 fluid intelligence. LRMs have substantial levels of fluid intelligence. The performance of LLMs on ARC 1 (a benchmark from 2019) remains ~10-15% today. Scaling them up by a factor ~100,000x got them from 0% to 10%. Meanwhile LRMs the same size or smaller saturated ARC 1 in 2025.

String theory is what happens when an intelligent system is starved of sensory input and starts hallucinating. Intelligence needs grounding. For this reason, intelligence is practically never the bottleneck to solving impactful problems. The bottleneck is iterated contact with reality.