Series
Minimal Viable Context
Runnable context packages that replace PRDs. What an agent actually needs to do the work, and nothing else.
Read the seriesWriting
I publishwhat Imeasure
Essays and benchmarks on agent systems, evaluation, and context. All of it lives on my Substack.
Read on Substack
Fig. 001
What I write about
Benchmarks
Rift findings
What breaks when frontier models get upgraded, measured from the outside with my own CLI.
Read the findingsEconomics
The unit economics of intelligence
Cost per correct answer, and why quality drift is a budget problem before it is a quality problem.
Read the essay
Fig. 003
Subscribe