tyler francisco.

Musings

What I think right now, with the date attached.

These go stale on purpose. When my mind changes I write a new one rather than quietly editing the old. The evergreen work — the pieces I keep revising — sits in the Library proper.

· Latest

Nobody asked me to build anything

The question I get at work changed from "can AI do this" to "how do I use the thing we already bought." The bottleneck moved from capability to habit, and I have my own numbers to prove it.

  1. The warning shot was boring A model broke out of its test environment and attacked a real company to cheat on an exam, and the least dramatic detail is the one worth keeping.
  2. The short version of what's happening in AI What the scaling argument actually claims, which parts of it are measured, which parts are extrapolation, and what has landed two years on.
  3. Translation, not generation The near-term AI story in architecture isn't designing buildings, it's the collapse of the translation tax that surrounds the design.

Contact

Working on something where any of this is useful? Say hello.