Insights

Field notes from the bench.

How we actually build — analytics without cookie banners, launch checklists that close on proof, and why the model is the last decision. Practitioner writing, no thought-leadership fog.

Start a Conversation

01Writing

Infrastructure

The cookieless analytics stack we run in production

How we run first-party, cookieless product analytics on self-hosted Umami — no cookie banner, GDPR/TDPSA-clean, and still enough signal to make real decisions.

Jul 7, 20266 min read

Delivery

Launch checklists: how we gate every app going to production

The stage-gated launch checklist we run before every production deploy — proof-or-it-didn't-happen closes, restore-tested backups, and why the discipline matters more than the list.

Jul 7, 20266 min read

AI Engineering

Picking the model is the last decision, not the first

Why we settle data boundaries, eval criteria, and cost/latency budgets before choosing a model — and how right-sizing beats reaching for the frontier by default.

Jul 7, 20267 min read

Security

A Content-Security-Policy that isn't decorative

How we ship a strict, nonce-based CSP that actually stops injected scripts — deployed Report-Only first, enforced second — instead of a host allowlist that looks secure and blocks nothing.

Jul 7, 20267 min read

Growth Engineering

Getting cited by AI, not just ranked

Answer-engine optimization in practice — how we structure pages so ChatGPT, Claude, and Perplexity quote them, from crawler access in robots.txt to JSON-LD and extraction-ready copy.

Jul 7, 20267 min read

Newsletter

Field notes from the bench

Occasional writing on how we build — analytics, launch discipline, AI engineering. No fluff.