Observability for Agentic Engineering Systems
If an autonomous coding run fails, "the agent failed" is not a useful diagnosis.
Jul 2, 20269 min read11
Command Palette
Search for a command to run...
Latest articles
Making Autonomous Coding Agents Safe Enough to Ship
The useful question is not whether an AI agent can write code. It is what the system does when the agent is wrong.
Jul 2, 20268 min read6
Building AutoPR: From GitHub Issue to Pull Request
How I structure an AI coding workflow around queues, durable state, focused agents, QA gates, and human review.
Jul 2, 20269 min read
I Almost Made the Wrong RAG Pipeline the Default
The most complex RAG pipeline felt smarter. The evaluation report made me change the product decision.
Jul 2, 20269 min read23
Catching Broken RAG Pipelines Before They Run
The best time to discover a missing `prompt` is before retrieval, reranking, and the LLM call.
Jul 1, 202611 min read19
From Simple RAG to GraphRAG Without Rewriting the Pipeline
I made "RAG type" a YAML file, not another Python branch.
Jul 1, 202612 min read30