Charlie writes about AI and capital. Find him on LinkedIn.
Recursive Advantage is where I write down what I'm thinking about. I work in the AI industry but I'm not an AI researcher — closer to a curious, informed observer trying to make sense of something that's moving faster than most people around me seem to think it is.
A lot of what's here is inspired by Leopold Aschenbrenner's Situational Awareness — one person's effort to think through where this is all going, in public. I read it and felt like more people should do the version of it that's theirs. This is mine.
The name comes from a pattern I keep noticing: the durable edges in this cycle aren't the ones you can buy or build once. They're the ones that feed themselves. Data into models, models into distribution, distribution into more data. Capital into compute, compute into capability, capability into pricing power. Recursion seems to be the actual mechanism, not just a metaphor — which is the part I find interesting enough to keep writing about.
Some of what's here is analysis. Some is speculation. I try to be clear about which is which, and where I can't, that ambiguity is usually the interesting part.
If something here is useful, wrong, or both — I'd like to hear about it.
Things I'm thinking about
- Where frontier model capabilities actually plateau, and where they don't
- The economics of inference at scale
- What 'agentic' means once it gets past the pitch decks
- How public markets misprice AI exposure (in both directions)
- The gap between benchmark performance and real deployment