Writings
Semi-Automated 24/7 Cursor + GitHub Learn Loop — What I Run Day-to-Day
field note ·
After go-live: feed agent:ready issues, approve PRs from GitHub Assigned, promote learn:candidate into rules — semi-auto 24/7, not a merge factory.
Read
Company OS Learnings — What I Kept After Retiring the Control Plane
field note ·
Company OS was a TUR control-plane experiment. It is no longer day-to-day ops — but the six-layer model, adapter boundaries, and approval ladder still transfer.
Read
Cursor Cloud Automations on This Site — Issue to Draft PR
field note ·
How this Astro repo ships from GitHub issues: agent:ready labels, a cron Execute automation, Node 24 Cloud Agents, draft PRs, and CI before a human merges.
Read
Six Months with Cursor on Databricks — Why Did I Wait on Skills?
field note · · Updated
After half a year of AI-assisted Databricks work in Cursor, installing official Databricks agent skills changed more than any model swap. Here is what was missing — and why I should have done it on day one.
Read
Databricks Cost Optimization: 5 Commands You Run Directly from Cursor
tutorial · · Updated
Five practical tips for optimizing Databricks costs that you can run directly from Cursor. From finding unused clusters to analyzing Spark configurations and reducing data spill.
Read
MCP Gateway Lessons from Production
field note ·
What breaks when you put a Model Context Protocol gateway in front of real tools: permissions, blast radius, timeouts, and silent half-successes.
Read
Routing Tables for Content and Data Agents
field note ·
Copy-ready routing tables for content and data agents — task lanes, model tiers, and approval gates using the same A–D discipline as coding agents.
Read
Eval Fixture Pack for Model Routing (Copy-Paste)
tutorial ·
A small, stable eval battery you can copy — fixtures for triage, JSON, edits, critique, and irreversible-action refusal — to promote models into agent lanes safely.
Read