WordPress 7.0 is close, and the WordPress Core AI leadership is going through its first big handover. Felix Arntz and James have announced they’re stepping back, and Jason Adams is taking over as Team Rep. In my years of dealing with core updates, a leadership change has usually meant one of two things: burnout, or a real change in technical direction. This one looks like the second.
It’s more than new titles. The handover comes as the Abilities API (introduced in WordPress 6.9) matures and the “WP AI Client” gets ready to ship in core with WordPress 7.0. This work changes how WordPress exposes functionality in a machine-readable way, and I’d take it seriously rather than filing it under “shiny object” development.
Why the WordPress Core AI leadership shift matters for devs
For most of us, “AI” in WordPress has meant a mess of plugins all calling the OpenAI API in their own messy ways. Under James and Felix, the WordPress Core AI leadership spent the last year building the “Abilities API” to fix that. It’s a standard registry where any plugin can declare what it can do, in a form an AI agent can actually understand.
If you haven’t read the Abilities API source yet, do. It’s a discovery layer, not a thin wrapper around REST endpoints. That said, I’ve watched enough “Beta” features land in core to know implementation is where things break. The spot I’d watch is race conditions when several “Abilities” fire in the same AI orchestration loop.
You can read more about the earlier phases of this work in my breakdown of WordPress Core AI Development: Leadership & Roadmap Insights.
How I look at the Abilities API
The “Naive Approach” to AI integration was to send a prompt and hope. The “Architect’s Approach” is to register an ability so WordPress takes care of validation and permissions. Here’s a simplified example of registering a custom ability with the new framework:
<?php
/**
* Example of registering a custom ability for the Core AI framework.
*/
function bbioon_register_inventory_ability() {
if ( ! function_exists( 'wp_register_ability' ) ) {
return;
}
wp_register_ability( 'bbioon/get-stock-levels', [
'label' => __( 'Check Stock Levels', 'bbioon' ),
'description' => __( 'Returns current stock for a specific SKU.', 'bbioon' ),
'category' => 'ecommerce',
'schema' => [
'input' => [
'type' => 'object',
'properties' => [
'sku' => [ 'type' => 'string' ],
],
'required' => [ 'sku' ],
],
],
'callback' => 'bbioon_handle_stock_query',
'permission_callback' => function() {
return current_user_can( 'manage_options' );
},
] );
}
add_action( 'init', 'bbioon_register_inventory_ability' );
This is where the new WordPress Core AI leadership will spend most of its time: making sure these callbacks don’t turn into performance bottlenecks. Once an AI client starts querying twenty registered abilities at once, your server’s transient cache will either save you or sink you.
Looking toward WordPress 7.0 and beyond
With Jason Adams leading, the focus is on getting the WP AI Client into core. If you’ve followed the WordPress 7.0 AI Integration plans, you know security and logging are the biggest hurdles. We need to know exactly what an AI agent did to a site, especially when it uses a “Capability” to refactor content or change settings.
I’ve seen plenty of “Canonical” plugins die on the vine, but I expect the MCP Adapter and AI Plugin to fare better, because they’re where future stable core code gets tested first. If you’re building AI features, keep up with the official Make WordPress Core notes on this transition.
If keeping up with the WordPress Core AI leadership changes is eating your dev hours, I can handle it for you. I’ve been working with WordPress since the 4.x days.
Final takeaway: stability over shiny
Changes in the WordPress Core AI leadership are healthy. New people look at hard problems like the Model Context Protocol (MCP) and the Abilities registry with fresh eyes. You don’t need to chase every update, but your sites do need to be ready for the WP 7.0 baseline. Start moving your custom logic into registered “Abilities” now, or you’ll be catching up when the AI Client lands.
