Notes on agentic development.

Journal Guardrails for regulated AI should be mechanical, not philosophical

Oct 5, 2026

Guardrails for regulated AI should be mechanical, not philosophical

We once shipped an agent that looked fine in staging—then audit logs were missing, and the kill-switch was a UI button.

Read more

Journal Don’t add microservices until you can produce receipts for the tax

Oct 4, 2026

Don’t add microservices until you can produce receipts for the tax

We’ve shipped multi-tenant SaaS and headless WordPress—complexity only sticks when you can measure, operate, and prove it.

Read more

Oct 3, 2026

I Run My Company From My Phone. A Mac mini Does the Work, and Every Action Is Audited.

How I built an AI employee with approvals, pull requests, gated deploys, audit logs and a kill switch, and why I can set it up for your business.

Read more

Journal Auditability isn’t a log feature; it’s a control-system design

Oct 3, 2026

Auditability isn’t a log feature; it’s a control-system design

We stopped an agent before it hallucinated a payout—because audit trails and kill-switches beat “trust the model” every time.

Read more

Journal Guardrails aren’t policy text—ship them as engineering controls

Oct 2, 2026

Guardrails aren’t policy text—ship them as engineering controls

We had a policy-perfect LLM fail production due to missing circuit breakers; controls beat compliance prose every time.

Read more

Journal WooCommerce caches fail “like random users”—here’s what we guardrailed

Oct 1, 2026

WooCommerce caches fail “like random users”—here’s what we guardrailed

A checkout race + cache invalidation + ACF quirks produced “random user behavior.” Our fix was boring guardrails with sharp edges.

Read more