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: ai-agent
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 readControlling Bedrock Agent Behavior: Prompt Stages, Parsers and Inference
How to override Bedrock's pre-processing, orchestration and post-processing prompts in Terraform, with a parser Lambda, JSON contract and per-stage inference settings.
22 min readWhat a Schema Can't Tell Your Agent: Open Knowledge Format on BigQuery
A BigQuery and Gemini ADK experiment: Open Knowledge Format bundles in Git gave an agent what the schema couldn't, stabilized answers and fixed a 29% GMV error.
13 min readA Local AI Agent with Ollama: Narrow Tools, Validation and Hard Step Limits
A local AI agent on Ollama that behaves: narrow tool allowlists, a capped loop, Pydantic structured output, explicit state and a test set of failure cases.
9 min readFrom RAG Agent Dreams to a VBA Prototype: Separating Raw and Derived Data
A freelancer's project tracker in Excel and VBA doubles as a lab for decision logic: independent status fields, no tertiary data, logical deletes and views to observe.
6 min readMulti-Agent Support Ticket Triage With LangGraph and Ollama
Build a three-agent ticket triage graph with LangGraph: a classifier, a responder and a human gate, running on local Ollama or an API with one changed line.
16 min readBuilding Agent Memory That Survives the Conversation Without Hoarding
A three-tier memory design for agents: let session state die, keep long-term notes small and editable, and fetch live data fresh. Plus a multi-agent handoff failure.
9 min readFour Rules for Coding Agent Memory That Stays Honest
Storing more notes won't fix coding agent memory. Four rules keep it honest: visible staleness, fail-closed redaction, loud truncation and kept constraints.
9 min readWhat the MCP Roadmap Means for Trading Research Servers
What the new MCP roadmap means for trading research servers: identity, long-running tasks, progressive discovery and result contracts beyond a long tool list.
9 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.