Few-shot prompting for WordPress code you can ship
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.
Category: AI
Prompt engineering sophistication sets the ceiling
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 7.0 AI features: what lands in Beta 1
Beta 1 is four weeks out. Notes from this week’s AI Contributor meeting: how the WP AI Client gets merged into core, the permission filter coming for AI prompts, and why WP Bench matters once you test models against real sites.
Air quality data repositories, and the formats that break PHP
Where air quality data actually lives, from OpenAQ and the EPA feeds to Copernicus and NASA Earthdata, what NetCDF, COG and Parquet mean for a PHP stack, and a transient-based cache that keeps the fetch off every page load.
WordPress.com MCP integrations now run on OAuth 2.1
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.
What the WordPress Abilities API actually shipped in 6.9
The WordPress Abilities API landed in 6.9 Beta 2, minus the JavaScript registration, which moved to a standalone NPM package in the Gutenberg repo. Notes on that split, the AI Experiments plugin, the WP AI Client 0.1.0 plan and the MCP Adapter refactor.
AI workflow automation in a warehouse that runs on clipboards
Low-tech logistics companies do not need AGI, they need to stop counting stock with a pen. Voice-driven inventory counts in n8n, damage reports generated from a photo, and a WordPress REST endpoint that takes the data in without breaking WooCommerce stock.
Why Google Trends data breaks machine learning models
Google Trends hands you a 0-100 score, not search volume, and that score moves when the date window moves. Notes on the normalization and rounding problems, plus the overlapping-window stitching I use to rebuild a daily series a model can actually learn from.
AI customer journeys require structural metrics, not tone
Perplexity and BLEU tell you nothing about whether an automated customer journey actually goes anywhere. The CDP framework scores continuity, deepening and progression against a taxonomy of journey stages instead.
Detect LLM hallucinations with vector geometry, not a judge model
LLM-as-judge asks a model that hallucinates to catch hallucinations. Displacement Consistency scores the vector between a question and its answer against grounded neighbors instead, with no source document and no second model call at inference time.