Notes on agentic development.

Journal Treat WordPress AI like payments: contracts, audits, kill-switches

Nov 13, 2026

Treat WordPress AI like payments: contracts, audits, kill-switches

We shipped an AI-assisted cart updater on WordPress and learned the hard way: no kill-switch = corrupted carts.

Read more

Journal Guardrails beat clever prompts when compliance can page you at 2am

Nov 12, 2026

Guardrails beat clever prompts when compliance can page you at 2am

We once had an AI agent hallucinate a policy step; the fix wasn’t better prompts—it was kill-switches, audit trails, and deterministic fallbacks.

Read more

Journal Headless, agents, microservices, multisite: don’t start there

Nov 11, 2026

Headless, agents, microservices, multisite: don’t start there

I’ve seen the same expensive failure: teams build distributed complexity before they prove the workflow needs it.

Read more

Journal Answer-engine safety starts with llms.txt, not more SEO

Nov 10, 2026

Answer-engine safety starts with llms.txt, not more SEO

We had a model confidently cite the wrong WooCommerce policy—LLMs.txt and boundaries beat keyword rankings every time.

Read more

Journal Headless, microservices, multisite, agents: don’t pick complexity by default

Nov 9, 2026

Headless, microservices, multisite, agents: don’t pick complexity by default

We’ve seen each “modern default” fail under latency, permissions, and ownership—here are the decision tests we actually use.

Read more

Journal Audit trails and kill-switches are the real LLM prerequisites

Nov 8, 2026

Audit trails and kill-switches are the real LLM prerequisites

We didn’t need “better answers” first—we needed an audit log, redaction, and a kill-switch when compliance asked for proof.

Read more