AI industry builder / agent workflows
Brad ZhangLongform AI writing, forum-grade X thinking, and open-source proof for founders, operators, and early AI teams.
Open to
Remote / Sprints
Topic dossier
Latency, serving, evaluation, and the real operating cost of useful model behavior.
What becomes visible, how quickly, and under what confidence threshold when models leave the lab and enter service?
Inference is framed here as a publishing problem. Latency, evaluation, and serving all become clearer once visibility and responsibility replace abstract infra vanity.
Lead essay
Every production model is also a publishing decision about what deserves to become visible fast.
Serving, latency, and evaluation become clearer when treated as editorial constraints instead of abstract infra metrics.
Project evidence
Diagramming as editorial infrastructure
A project that treats architecture diagrams not as cosmetic add-ons, but as durable explanatory surfaces for AI systems work.
Session continuity without continuity theater
A checkpointing layer for coding sessions, built to survive interruption, context loss, and handoff without pretending memory never breaks.