About
I research internal bottlenecks for people helping superintelligence go well:
I write at chrislakin.blog, and tweet @chrislakin.
Updated October 2026 from San Francisco.
Things I believe:
Longstanding bottlenecks tend to be locally optimal (see also: efficient markets ~ Chesterton’s Fence ~ safety)
Outcomes must be tracked years later
There’s not much time, so we must move fast
High-performers benefit more from unlearning than more learning
Getting what you want requires aiming at what you want (not what you don’t want!)
Use of willpower cuts against globally optimal performance
Our best decisions are not “made” but revealed through intuition
People progress in stages. There is little use in targeting internal bottlenecks when external bottlenecks (health, sleep, financial runway) are present
When people in AI try to resolve longstanding bottlenecks on their own, usually either nothing happens, they spiral in emotions, or it works at first but it’s mysteriously difficult to keep doing
Outside of work, I’ve…
lured my nyc gf to sf
Previously, I’ve worked on
physics at Carnegie Mellon
