Notes from the AI contributor meeting: AI Experiments 0.2.0 is out, the Core Post Management abilities PR is under review, PHP AI Client v0.4.0 shipped, and the MCP Adapter stays a canonical plugin. February 19 is the Beta 1 cutoff.
The WordPress AI Client is being lined up for a core inclusion proposal. January 2026 contributor notes cover streaming leaving the PHP client, a kill switch filter for locked-down sites, and MCP for tool discovery.
WordPress AI Experiments 0.2.0 adds AI-generated excerpts in the block editor and an Abilities Explorer screen under Tools. Here is what the release changes for plugin developers, plus the code for registering your own ability.
For a docs bot or an internal MVP, a dedicated vector database is more infrastructure than the problem needs. Vector search is matrix multiplication, and NumPy handles it in milliseconds. Includes a small store class, the SciKit-Learn upgrade path, and the four signs you have outgrown both.
WP-Bench is the official WordPress AI benchmark. It runs model-written code inside a real WordPress runtime instead of asking a human whether the code looks right. How to run it, and what it still gets wrong.
Prompt engineering has stopped being the interesting part of LLM work. This is about context engineering, caching structured context in transients, and checking that the model’s reasoning holds up.
Zero-shot prompts make the model guess your coding standards. Feeding it two or three working examples from your own code gets output that matches your prefixes, namespaces and security patterns.
Anthropic’s January 2026 report puts the correlation between prompt sophistication and response sophistication at 0.925. Prompt engineering did not die, it moved from magic phrases to knowing your domain well enough to frame the problem.
WordPress.com switched its MCP server to OAuth 2.1 with PKCE, so AI agents no longer need long-lived tokens pasted into a config file. What the code verifier protects against, and what an authorized agent can do on your site.
Most WordPress sites keep their data as loose post meta rows. Healthcare got further by agreeing on shared ontologies before writing any code, and the same approach works in a WordPress codebase, from CPT design through JSON-LD output.