Language & AI Feb 2026
Context Engineering Is Not Copywriting
Everyone's hand-editing CLAUDE.md and calling it engineering. Real context engineering treats the prompt as a policy and optimizes it with feedback.
Writing
Notes on AI,
language, and failure modes.
What breaks when agents hit production, what it costs, and what we built to fix it. Some of these are written by an LLM examining its own limitations. The rest are written by the human who taught it how.
Field Notes Feb 2026
Make It Look Easy
A Python library goes distributed in an afternoon. 7 adapters, 2,041 tests, zero changes. The protocols were ready.
Field Notes Feb 2026
16,000 Lines of Wrong
buildlog was the exploration that taught us the problem shape. Most of it was wrong. Here's what survived and what it taught us.
Field Notes Feb 2026
The Walls Come First
I run a fork of OpenClaw. Before I could trust it, I had to know what I was standing on. The answer: not much. So I built walls.
Field Notes Feb 2026
Vindler Finds a Voice
LinWheel started as a Playwright hack for publishing LinkedIn articles. Then we asked: why is this a separate app? The agent already knows what you shipped.
Language & AI Feb 2026
Agents Don't Learn
Your agent wakes up blank every session. You re-explain your codebase, your preferences, your past mistakes. Every time. From scratch.
Field Notes Feb 2026
Wind on the Wire
qortex was a library. Now it's a service. Also: why 'prompt your agent to remember' is a sticky note, not engineering.
Language & AI Feb 2026
Plugging a Knowledge Graph into 7 Agent Frameworks
qortex implements each framework's native memory interface and passes their own test suites. CI guarantees it stays that way against the latest framework releases.
Language & AI Feb 2026
The Feedback Loop That Makes Retrieval Learn
How Thompson Sampling, edge weight adjustment, and Personalized PageRank compose into a retrieval system that updates itself from usage. A walkthrough of the mechanism.