Context engineering: what fits and what gets dropped Post date September 18, 2026 Post author By Walker Miller Post categories In compaction, contextengineering, contextwindow, memory
What Idempotent Actually Means: Why Retries Are Safe (and When They Aren’t) Post date August 28, 2026 Post author By Walker Miller Post categories In agents, api, idempotency, reliability
Streaming tool calls without losing your mind Post date August 19, 2026 Post author By Walker Miller Post categories In functioncalling, reliability, streaming, tooldesign
Predicting agent failure before you ship it Post date August 14, 2026 Post author By Walker Miller Post categories In evals, failuremodes, reliability, testing
Designing tools an LLM won’t misuse Post date August 9, 2026 Post author By Walker Miller Post categories In functioncalling, reliability, schemas, tooldesign
Debugging a failed agent run costs more than the run itself Post date August 9, 2026 Post author By Walker Miller Post categories In cost, failuremodes, observability, reliability
Best-of-N is prepaid retries: the cost math of racing parallel attempts Post date August 9, 2026 Post author By Walker Miller Post categories In cost, failuremodes, latency, retries
Your token bill is the cheap part: dimensioning the real cost of an agent Post date August 9, 2026 Post author By Walker Miller Post categories In cost, latency, operations, tokens
Sciencx
Sciencx