Engineering notes for what changes after the demo.

Long-form guides on agent reliability, context and tool design, structured generation and model infrastructure.

Your Agent Memory Needs Two Clocks

Mutable facts need one interval for when they applied and another for when the system knew them.

Featured guide26 Jul 20264 min read

Tie every prompt workaround to a failure case, then test its removal when the model changes.

26 Jul 2026AI Agents4 min read

Ship the model, reasoning mode, and history adapter as one versioned protocol.

25 Jul 2026AI Agents4 min read

Hiding a tool from an agent changes its options, not the authority behind them.

25 Jul 2026AI Agents4 min read

Milestone reach rates reveal the transition where agent reliability collapses.

25 Jul 2026AI Agents4 min read

Hybrid attention turns prefix caching into a memory-consistency problem across KV pools with different lifetimes.

11 Jul 2026LLM Infrastructure4 min read

56 LLM and agent frameworks ranked by GitHub stars, with language, age, size and funding.

59 design, writing and code patterns that leave the reader to do the real work.

“Every thought — however small — gets an em dash.”

The em dashFirst place, writing