WRITING
OCCASIONAL LONG-FORM ON INFRASTRUCTURE, CRYPTO, AND THINGS I BUILD. FULL POSTS LIVE HERE; SOME FIRST APPEARED ELSEWHERE.
002 / LATEST22 JUL 2026
0 → 200 TOK/S
MY LOCAL INFERENCE STACK
MY LOCAL INFERENCE STACK
Bare Windows gaming PC to production local inference in under a day — 200 tok/s at 131K context, served over WireGuard, $0 per token for every routine agent role.
READ →