now · what I’m on lately
A working set of obsessions.
Updated whenever something new sticks. Inspired by nownownow.com.
- The Bitter Lesson — Sutton, again
- Sparks of AGI — Bubeck et al.
- The Pragmatic Programmer — Hunt & Thomas
- Three-Body — Liu Cixin · finally
“Don’t let anyone tell you, you can’t do something.”
MOPI-HFRS A2C extension
· writing the paper
Bug-report benchmark
· eval harness
- · for model alignment is it more ideal to address it during training instead of post-training behavior alignment?
- · will AI safety always be a point of debate as the world holds different definitions of morality and risk?