Mary Olowu

Most table renames are database-easy and deploy-hard. The SQL looked trivial: ALTER TABLE...

Let the model interpret. Let software validate. Let workflows own state, approvals, and side effects.

Not all the context. Just the small set of durable facts that keeps an AI-assisted project coherent across sessions.

AI can write good code in isolation, but across sessions it forgets the decisions and tradeoffs that keep a real project coherent.

How I structure Claude Code for long-running monorepo work using guardrails, maintainers docs, lightweight memory, and reusable workflows.

Store GitHub webhook events with signature verification in five minutes. The same pattern works for any provider.

One API call to subscribe a customer endpoint. HMAC-signed deliveries, 5 retries over 40 minutes, full delivery log, one-click replay. No queue, no retry logic, no Svix subscription.

Cross-posted from the Centrali blog. The canonical version with code highlighting and updates lives...

A practical local testing workflow for Stripe webhooks using Stripe CLI forwarding, replay, and run-level logs. Stop guessing whether your handler actually ran.

A config-driven CLI that indexes MkDocs, markdown, JSON, and HTML into Meilisearch with atomic zero-downtime re-indexing. You define the document shape; content-mill handles extraction and swap.

If you've ever tried to make your docs, blog posts, or changelogs searchable with Meilisearch, you...

A simple mental model for distinguishing recursive trigger loops from legitimate high-volume traffic — and why this distinction makes rate limiting actually work.