WordPress 7.0 shipped an AI Client, a Connectors API, and a Connectors screen, and left the actual features to plugins. Here is what that means for self-hosted sites, where provider keys should live, and what to test before enabling anything on production.
Agents write hooks as anonymous functions, which works until someone needs to remove one later. Here is why remove_filter fails on closures, the workarounds I avoid, and the naming rule I enforce before AI code ships to a client site.
OCR failures don’t throw errors on a WooCommerce site; the engine stores garbage as confidently as clean text. This is my read on a fourteen-engine benchmark, why swapping engines moves the failure instead of fixing it, and the small PHP validation layer that catches silent failures.
When a coding agent turns an afternoon of WooCommerce maintenance into an hour, the saved time doesn’t stay with you. Your contract decides who captures it, and hourly billing hands the gain to the client. This is about pricing the outcome instead.
Agents are useful on WordPress work until the day one runs wp plugin update –all on a live store or rewrites the database without a dump. This is the WordPress version of the blocked-commands list: what to gate, the WP-CLI flags that help, and the AGENTS.md rules I write for client repos.
AI keeps drifting from your design system because it is guessing. The fix is written spec files, a real token layer, and an audit. For WordPress that means theme.json presets, template hooks you refuse to regenerate, and a grep for hard-coded colors.
Padding every short product description up to the longest one wastes real GPU time. A C++ backend that packs sequences into one stream fixes it, and the same budget-first instinct already runs inside WooCommerce’s Action Scheduler. Here’s how the two compare.
A founder gave ChatGPT an intern’s month-long data problem and got 80 percent of a prototype in 48 hours. I look at what that means for WordPress research spikes: the demo you get, the shortcuts agents take, and the understanding you quietly skip.
Combining two coding agents works when the second one reviews what the first wrote. I explain how I split Claude Code and Codex across WordPress client work, which repo rules the reviewer needs, and where staging still beats both.
Qdrant TurboQuant is a one-line config change that cuts vector memory, but it keeps float32 copies on disk for rescoring and runs calibration on every segment. Before patching a production collection, here is what changes and when a WordPress search build actually needs it.