WooCommerce for Claude came out of Automattic’s Radical Speed Month. It uses MCP to give the model analytics lookup tables, a store knowledge layer and an AI-readiness score, all on one endpoint. Plus the provider pattern and how to run it locally.
A “5x improvement” on its own tells you nothing. Three questions pull it apart: which dimension moved, from what baseline, and against which period. Examples from Core Web Vitals, model accuracy, and a transient cache bug that faked a 10x speed boost.
Claude Code gets more reliable when it can run its own output: unit tests through WP-CLI, parallel prompts compared against the goal, and Chrome screenshots over MCP for front-end work. Also why an agent looping with no check-in is worse than no loop.
AI coding tools copy whatever patterns already exist in a repo and know nothing about battery budgets or packet sizes. In IoT that combination turns into technical debt faster than a team can refactor it away.
A store you can launch in an hour is fine for a prototype. For a real business, the shipping zones, tax classes, asset loading and payment webhooks you set up on day one decide how much refactoring you do later.
WordPress 7.0 benchmarks show TTFB about 10% worse than 6.9 while LCP holds steady, which puts the cost on the PHP side. Notes on the script module dependency ticket, oversized Link headers that make Nginx return 502, and polling in real-time collaboration.
Binary churn flags tell you that a user left but say nothing about the timing. This post covers discrete time-to-event modeling: why fixed intervals fit WordPress data, how right censoring skews a model, and a small PHP function for hazard rates.
An LLM can retrieve the right document and still quote the wrong price. This is the detection and healing layer I built to catch that: five failure patterns, a faithfulness score, deterministic patches, and a SQLite drift monitor that caught a real bug in staging.
LangChain is fine for a prototype. Once there is real traffic, several agents coordinating and a pager attached, the abstraction costs more than it saves. This is the case for writing the orchestration layer yourself.
Utility software gets designed to be invisible, which is why people resent it. Linear, Vercel and MacPaw treat the maintenance layer as something a person actually sits through, and it shows.