Correspondence
Precise notes only.
Notes on inference economics and verification economics are welcome. So are engineering reviews of AI runtimes, distributed systems, and financial systems infrastructure where the work has a chance of becoming a durable artifact, review memo, benchmark, or operating plan.
Archive summary: Correspondence is for selected work where the output can become a durable artifact: a system, benchmark, protocol, paper, implementation reference, operating layer, or paid technical review.
Good fit
01AI runtime, evaluation, reliability, or agent-system reviews.
02Inference-cost, verification, or technical-diligence questions.
03Architecture calls where the output should become a memo, benchmark, or operating plan.
A useful note includes
01What system or question exists?
02What constraint makes it hard?
03What artifact should survive?
04Why should the work matter after the project ends?
Selected Work
Selected work is considered when the output can become a durable artifact or a concrete decision. The best inquiries include context, constraints, timing, and the artifact that should exist afterward.