Clever WordPress code tends to fall over in production. What changes when you optimize for reliability instead: a chunked WP-CLI batch pattern, debugging a failing WooCommerce checkout methodically, and using LLMs for edge cases rather than features.
The WP AI Client is up for a core merge with Beta 1 ten days out and no PR to test. Notes on that proposal, the admin reskin in ticket #64308, the Release Squad now in place, and what Gutenberg 22.5 changes for 7.0.
How I use Claude Code with WordPress Studio to scaffold plugins: setting up a local site and terminal, installing the CLI, the context a prompt needs, and why the naive admin notice it hands you is an XSS bug waiting to happen.
The contrast-color() function only works in Safari and Firefox so far, so dynamic text colors still need a workaround. Here is the OKLCH lightness trick I use in block themes and WooCommerce stores: one declaration, no JavaScript.
Gutenberg is experimenting with content guidelines: a machine-readable home in the admin for voice, tone and structural rules, so authors and AI assistants work from the same standards instead of a PDF nobody opens.
Generic hosting and WordPress hosting differ in the parts nobody advertises: PHP worker counts, server-level and object caching, and a firewall that knows WordPress attack patterns. Here is what to check before you sign up.
Pydantic v2 runs on a Rust core, but only if your models let it. Four changes that keep validation on the fast path: Annotated constraints instead of field validators, model_validate_json, TypeAdapter for batches, and leaving from_attributes off.
Mechanistic interpretability is the Xdebug of AI research: hook a forward pass, zero out a neuron, and see what the model was actually doing. A walk through the residual stream, TransformerLens, circuit tracing and steering vectors.
Agentic analytics moves the consultant’s job from writing queries to architecting the systems agents read from. What that looks like in practice, including a clean WordPress REST endpoint built for agent consumption.
How I set up Claude Code context: local docs and meeting transcripts in a /knowledge folder, a .claudignore that keeps the agent away from secrets, and a bash script that bundles only the files a given feature needs.