A client updated to WordPress 6.9 the day it shipped and broke their landing page. Here is why I hold major core updates until the first maintenance release, and the filter I use to keep them in staging.
I wired a client’s store straight into one AI provider and regretted it two weeks later. The Abilities API and the WP AI Client give you a way to register tools and send prompts in WordPress without hardcoding a model.
A marketing team could not find a campaign page in Appearance > Menus. WordPress 6.9 narrowed the menu search to post titles, and one filter on wp_ajax_menu_quick_search_args brings the old behavior back.
A client’s launch caught on with a subreddit and their $5 shared plan started killing PHP workers, four hours of downtime at the worst moment. No caching code fixes a server with no workers left.
Notes puts editorial comments on the editor canvas, the Abilities API replaces scattered capability checks for automated workflows, and classic themes finally load block styles on demand.
A Make WordPress Core proposal would allow the PHP short echo tag for single statements. Here is why the old objection stopped applying at PHP 5.4, and how to keep the rule strict in your own phpcs.xml.
A 60-site multisite network taught me that bring-your-own-API-key doesn’t scale. Managed hosts should register AI providers through the WP AI Client, the same way they already hand you MySQL and an object cache.
A client’s WooCommerce store was slow because five plugins were each calling their own LLM API on every save. That is the problem the Ai_Client in WordPress 7.0 and the new Abilities layer are meant to solve.
A client wanted a sparkle button by Friday. What WordPress 6.9 actually shipped matters more: the Abilities API and the MCP adapter, which give AI agents a defined way to talk to a site.
The November 2025 core committers check-in set the direction for WordPress 7.0: PHP 7.4 as the minimum, AI models running in the browser instead of on your server, and a reworked WP Admin.