§ Writing
Long-form analysis on architecture patterns, observability, orchestration, and production operation of analytics agents. Grounded in what practitioners actually ship.
§ Posts
How to coordinate multiple analytics agents: orchestration topologies, wave-based execution, and the trade-offs between disk-based coordination, Temporal workflows, and Redis-backed queues.
Read →How to know what your agents are doing: structured trace files, query-level cost attribution with BigQuery job labels, INFORMATION_SCHEMA monitoring, and alerting on the failure modes that matter most.
Read →How to build agents that remember what they learn: cross-session memory with file-based persistence, handling schema drift, and context budgeting so long investigations don't degrade.
Read →How to move from a generic LLM with database access to purpose-built analytics agents. Covers agent specialization, disk-based I/O, domain context, and prompt tuning for analytical rigor.
Read →A step-by-step guide to giving Claude Code and OpenAI Codex direct access to your BigQuery warehouse, so they can query your data, explore schemas, and run multi-step investigations autonomously.
Read →