Articles for people who
ship the stack
Rewrites on React, Node.js, TypeScript and AI. A translated body is shown when it exists. If it does not, you see the English text and a note that says so.
Tagged: llmops
Building an Agent Improvement Loop with LangGraph and LangSmith Evals
A step-by-step LangGraph and LangSmith loop: trace collection, three scoring layers, failure clustering, offline regression datasets and pre-ship evals.
24 min readAn AI Engineering Roadmap for 2026: From Python to Agents and LLMOps
A phased 2026 learning path for AI engineering, from Python and math through classical ML and deep learning to RAG, MCP, agents, evals and deployment.
14 min readWhen the Dashboard Is Green and the Agent Is Wrong: Tracing State
The dashboard is green and the agent is wrong: failures hide in state passed between healthy steps. What tracing, routing records and evals must capture.
12 min readWho Grades the Grader? Validating Agent Eval Suites Before You Trust Them
Who validates the eval suite? Judge drift, outcome-only scoring, private failure modes and hidden cost can make LLM evals lie. Checks for the suite itself.
7 min readEvaluating RAG Retrieval Beyond ANN Recall, with Turbopuffer as the Example
A fast vector index can still miss the answer. Learn Turbopuffer's storage tradeoffs, hybrid BM25 plus vector search, and how a golden set measures real recall.
7 min readFrom Evals to Self-Improving Agents: Five Eras of Agent Evaluation
Agent evaluation now means auditing traces. See how deterministic checks, LLM and agent judges, diagnosis and a meta-harness build a safe improvement loop.
7 min readTracing a LangGraph Blog Pipeline with Langfuse: Spans, Scores, Warnings
Instrument a multi-agent LangGraph pipeline with Langfuse: one trace per article, nested spans, warnings, scores, and a debugging routine for weak drafts.
8 min readReplay-Safe Agents: Keeping LLM Calls Out of the Workflow With Temporal
Why an agent that crashes mid-run must not re-fire its LLM: activities as a non-determinism boundary, replay mechanics, heartbeats, retries and stateless graphs.
16 min readEngineering an Evaluation Control Plane for Production AI Agents
How to build continuous agent evaluation: separate execution and control planes, trace-based assertions, golden datasets, CI gates, lifecycle testing and a launch checklist.
10 min read
About these articles
Request a 24h estimate
Need the same stack in a production operator layer? Send the brief — estimate within 24 hours.