Featured
Recent Posts
- Published:
All Agents Are Useless at Estimating Their Own Work
I asked four coding agents how long their own work would take, then timed them doing it. Every one was wrong, by 12× to 177×.
- Published:
Why the gradient is a list of partial derivatives
Building the gradient formula from scratch using a ski-slope picture, with minimal calculus assumed.
- Published:
Investing When Anyone Can Build
If building software is nearly free, what are VCs actually investing in?
- Published:
Agents as Bounty Hunters
I built a benchmark that pits coding agents against each other in a bug-finding treasure hunt.