Catalog — In Print
Two tiers. Field guides at $9 — curated essay compilations, ~5,500 words each, one sitting. Books at $19 — code-first deep-dives, 4,500+ words, one weekend. Both serve the same readers: practitioners who'd rather have an opinionated map than another exhaustive survey.
Field Guides — $9 each
Four essays on why retail algo trading fails — and the architecture that survives.
Compiled from the trading vertical of the daily brief. Why most retail systems blow up at month four, the six lies backtests cannot simulate, how institutional desks actually use ICT, and an honest comparison of every real platform in the category.
Four essays on why most recovery apps fail — and the relapse-aware architecture that works.
Compiled from the behavior vertical. Why habit apps don't survive month three, the discipline-app paradox (more apps → worse outcomes), what every recovery app gets wrong about relapse, and an honest comparison of the apps actually built on Marlatt's research.
Four essays on what shipped in AI this month — and what the production patterns mean.
Compiled from the AI builder vertical. Mozilla using Claude Mythos to harden Firefox, OpenAI gating Spud to defenders, what actually shipped at Code w/ Claude 2026, and an honest review of the AI engineering books worth reading.
Each book is one weekend of reading, one Monday of shipping.
Books — $19 each
The decision tree, the code, and the failure modes nobody warns you about.
Most AI agent demos fail at message four because the agent forgets the user. Five chapters covering the four axes of agent memory, the five production patterns with copy-paste-ready code, and the five failure modes nobody warns you about.
The whole system, not just the components.
Builders understand one piece — memory, or tools, or planning — but not how they fit together. This is the systems-level architecture for moving an agent from demo to production.
How agents connect to real tools, safely.
Model Context Protocol gets thrown around as an acronym. Almost nobody is implementing it cleanly. This is the implementation playbook — server design, auth, sandboxing, the patterns that scale.
The reliability discipline most teams skip.
Most agents ship without real evaluation, then break in production with no signal of why. This book is the eval framework — the metrics, the benchmarks, the test infrastructure.
Which models actually matter and how to deploy them.
Llama, Qwen, Mistral, DeepSeek, Phi — too many models, no clarity on which to use when. This book ranks them by use case, with deployment recipes for each.
The architecture call most teams get wrong.
Multi-agent is exciting and almost always premature. This book is the decision tree — when single agents work, when multi-agent earns its complexity, and the patterns that actually ship.
“If you could ask Claude this and get the same answer, the book gets rewritten until you can't.”
Editorial Standard
Not buying yet?
Free monthly long-form on the AI ecosystem. Ten editorial pieces — the releases, skills, and arguments that moved this month. Refreshed every month, sent to your inbox. Cancel with one click.