AI Philosophy

Why current AI is powerful but incomplete, what's actually missing, and how to build the machinery for artificial intelligence that gets better by living — not by being redesigned.

Featured Essay·10 min read

AI Isn't Dumb. It Just Hasn't Lived Yet.

Why the size of the model isn't the problem — and what biology figured out 3.8 billion years ago.

Every few months a new model drops and people argue about whether AI is 'really' intelligent. They're asking the wrong question. The first living organisms weren't smart either — but they had the machinery for intelligence to emerge.

Read essay

Core Arguments

The Krebs Cycle of AI

Next token prediction is essential. It's also not the point.

No biologist points at the Krebs cycle and says 'that's life.' It's the substrate, not the intelligence. Next token prediction is AI's Krebs cycle — essential, solved, and not the point.

4 min

750 Megabytes

The most devastating argument against the scaling hypothesis, hiding in plain sight.

Human DNA encodes the most sophisticated intelligence on Earth in 750 megabytes. AI models use terabytes and can't remember yesterday. The difference isn't size — it's that every byte of DNA was forged against reality over 3.8 billion years.

4 min

Heuristic, Not Deterministic

Why evolution beats engineering for open-ended problems — and what that means for AI.

Engineering builds deterministic systems that handle known scenarios. Evolution builds heuristic systems that handle scenarios that have never existed. The real world is non-deterministic. Only heuristics scale to infinity.

5 min

Write DNA, Not Code

The most important shift in how we use AI to build software.

DNA doesn't encode behavior. It encodes the machinery for generating behavior. The most important shift in AI isn't writing better code — it's writing better specs. Generative blueprints that build adaptive systems through lived experience.

4 min

Origin

These ideas crystallized while building Reeves, a personal AI operating system. The conversation started with engineering — “how do we make this a better app?” — and ended somewhere deeper: “how does a computer learn to know a person?” The answer turned out to be the same answer biology found 3.8 billion years ago.