An AI chat over your WooCommerce orders will hand you confident numbers that don’t match Analytics. Let the model pick the question and let your code answer it. Here’s how I’d wire that with one REST endpoint and a strict args schema.
Naive RAG wrappers that just chunk PDFs into a vector store break on real contracts and reports. Here is the four-brick approach I use instead: structured parsing, question parsing with an expert dictionary, filtered retrieval, and typed generation.
WordPress.com’s Posts to Podcast feature turns blog posts into AI-generated audio from the dashboard. I look at how it stores the files, the media bloat risk, and how to manage the assets with WP-CLI and a theme hook.
Tightening the prompt won’t make an LLM integration reliable in production. Here’s the control layer I put between WordPress and the model: input guards, a circuit breaker, response validation and retries.
WooCommerce.com now runs its production store on Nightly WC Core. An AI-assisted pipeline handles patch replays and diff reviews, and the setup has already caught a dropped 15GB index before it could ship in 10.8.
Felix Arntz and James are stepping back as Core AI Team Reps, and Jason Adams takes over. The handover lands as the Abilities API matures and the WP AI Client heads into core for WordPress 7.0.
Using Claude for data science now goes well past chat prompts. Three skills I’d pick up: throwaway interactive dashboards for EDA, Claude Cowork for Jira and reporting chores, and Claude Code for tracing broken dbt pipelines.
The WordPress AI plugin 1.0.0 adds Request Logging and Connector Approvals, which give site owners real control over which plugins use AI and what they spend. Here is what changed and what developers should do about it.
The latest WordPress.com sprint brought Jetpack Podcast, the WordPress Workspace Mac app with an AI agent, and a Blueprints Gallery in Studio. The common thread is local-first development and less repeated setup work.
A production recommender on Amazon EKS lives or dies on infrastructure, not the model. Here’s how I set up a multistage system with NVIDIA Triton, cut feature lookup latency with an in-memory cache, and scale GPU nodes with HPA and Karpenter.