disturbed by the narrative of technological inevitability — of unstoppable advancement toward a world of human obsolescence — that removes the human responsibility to think critically about how technologies are built and introduced into society

There’s a kind of death that’s necessary for life. We see it in no-self meditation experiences; in social change that only happens with the next generation; in relationship endings that change us. In our world of great control & permanence, we often forget to let things die.

We’re all starved of touch and connection these days but for some reason in the Bay Area it manifests as authentic relating groups… autism or innovation?

For those who want AIs to be our children: Frankenstein is a good read

wow, just discovered that substack notes is far more interesting than twitter!

This is very cool — a true open science initiative for building AI models.

It's amazing how much algorithms shape my behavior — on Twitter I want to post edgy takes, on Substack I want to form ideas I suspect the effect compounds, and I'd become much more interesting if I spent as much time on Substack as I do here

I think AI progress feels roughly even across industry labs because: 1. The main ideas are conceptually straightforward and thus diffuse easily (i.e. RL over long reasoning chains) 2. Higher quality data makes a difference, but it's not noticeable for many use cases, so even

GPT-4.5 is quite a good writer. But we experience it as a bad writer because when writing, we’re trying to express an internal experience, and it can’t capture that. To me, this says “good writing” is more about capturing our human experience than about the words produced.

How do we move beyond prototypes and build robust software systems with AI? This Thursday at the AI Engineer World's Fair, our CTO Josh Albrecht will share lessons and insights from building Sculptor, our coding agent environment. 🗓️ June 5, 2:40 PM 📍 Yerba Buena Ballroom