Notes on agentic development.

Journal AI in WooCommerce must act like a transactional service

Sep 17, 2026

AI in WooCommerce must act like a transactional service

We learned the hard way: when the model is wrong, checkout still has to be right—every time.

Read more

Journal Regulated AI is guardrails, not smarter models

Sep 16, 2026

Regulated AI is guardrails, not smarter models

The model isn’t the risk boundary—your workflow is. Audit trails and kill-switches decide whether failures are safe.

Read more

Journal Answer-engine SEO starts with llms.txt you can actually version

Sep 15, 2026

Answer-engine SEO starts with llms.txt you can actually version

We cut agent failures 38% by treating llms.txt like code: retrieval rules, staged rollouts, and auditable versions.

Read more

Journal Audit-proof AI isn’t “better prompts”—it’s kill-switches and traceability

Sep 14, 2026

Audit-proof AI isn’t “better prompts”—it’s kill-switches and traceability

We once watched a model drift break a finance workflow. The fix wasn’t tuning—it was deterministic logs, traceable versions, and instant rollback.

Read more

Journal We killed WooCommerce checkout race conditions with boring, proven telemetry

Sep 13, 2026

We killed WooCommerce checkout race conditions with boring, proven telemetry

A Fortune 500 apparel checkout started double-charging—race conditions and permission drift hid until we instrumented every state change.

Read more

Sep 12, 2026

llms.txt is API design: stop trusting crawlers to guess your meaning

The wrong excerpt isn’t a “SEO issue”—it’s a contract breach with crawlers. llms.txt fixes the contract by design.

Read more