Engineering blog
Notes from building Remediate Labs — agent architecture, retrieval design, debugging stories, and the unglamorous decisions that make production agents actually work.
RAG Finds the Candidate. The Live Store Confirms the Truth.
A search index can't be a source of truth — that's a category error, not just a bug. The general rule for any vector or search index sitting in front of a live data store.
Why the Same Bug Kept Creating New Incidents (And What That Taught Me About RAG)
Three layers of dedup. Four independent failure modes that all had to fire simultaneously. The compound bug that exposed them, and the principle that makes it not happen again.
Why My AI Agent Kept Adding Null Checks Instead of Fixing the Bug
Five PRs to teach a fix-generation pipeline that the crash site is almost never the fix site. The producer/consumer distinction, RAG's structural blind spot, and what it took to find the actual bug.
The Three Camps of Retrieval Architecture: A Practical Guide
Improved RAG, GraphRAG, Ragless. Each is right about something. Each is wrong applied to problems it wasn't designed for. A decision framework for picking the right default and extending deliberately.
My Agent Ran for 58 Seconds and Made Up Every Number
How a CICDAgent fabricated a confident CI failure report without calling a single tool — and the four trace signals that would have caught it sooner.
Four Patterns for Structuring Agents (Part 2)
Once you accept ReAct isn't the default, what do you use? Four structural patterns that emerged from building a 5-agent PR automation system.
Not Every Agent Needs a ReAct Loop
ReAct is one pattern, not the default. Three patterns for what each agent actually needs from the LLM — and when the answer is no LLM at all.
The Fix That Broke My Agent in a Different Way
MANDATORY CONSTRAINTS stopped the hallucination. Then the agent looped on the same tool 10 times. Three words on the constraint fixed it. Why precision in prompts is everything.
Typed Boundaries Make Multi-Agent Systems Readable
Where you put parsing logic decides whether your orchestrator stays readable as agents multiply. Parse at the boundary, not in the caller.