Agentic coding: how I debug WordPress sites now
A three-step way to point an agent at a broken WordPress site: let it read debug.log, make it write a hook-level plan before it edits anything, then iterate on what it ships.
Tag: AI Assistants
Run Claude Code for free with Ollama and a local model
Ollama 0.14 speaks the Anthropic Messages API, so Claude Code can run against a local model instead of a $200 plan. Here is the setup, the three environment variables it needs, and the git init step that trips people up.
The WordPress AI Client heading into core 7.0
Notes from the February 4 contributor meeting on the WordPress AI Client merge into 7.0: why OpenAI, Google and Anthropic are moving out to separate plugins, why the off switch is a constant rather than a filter, and what MCP adapters mean for the Abilities API.
WordPress 7.0 AI integration: the road to Beta 1
Notes on the WordPress 7.0 AI integration roadmap: the WP AI Client merge proposal, the GatherPress demo built on the Abilities API, how the native client compares to the MCP adapter, and what has to land before the February 19 Beta 1 deadline.
What the WordPress.com Claude Connector actually does
WordPress.com now has an official Claude Connector built on MCP and OAuth 2.1. It gives Claude read-only access to site data like traffic, comments and content, with no custom REST bridge to maintain. Here is what it can do, how to switch it on, and where it stops.
Syncing frontend and backend code in one repo
Separate repos for the React frontend and the WordPress backend are why payloads drift out of sync. Notes on moving to a monorepo with Roots Bedrock, keeping an AGENTS.md map of endpoints for coding agents, and the stale REST cache that broke a checkout.
Why I let my agentic architecture write its own tools
Pre-built agent toolboxes break the moment a client asks something you did not plan for. Here is the plan-code-execute setup I use instead, where the agent writes its own Python, plus the hallucinated CSV that taught me why the analyst step is mandatory.
Build a persistent LLM memory layer with DSPy and QDrant
Chat history arrays are a log, not memory. This walks through the LLM memory layer I build instead: DSPy signatures for extracting atomic facts, QDrant for per-user vector storage, and a maintenance agent that decides when an old fact gets updated or deleted.
How Recursive Language Models get around context rot
A million-token context window does not mean a model can reason across a million tokens. Recursive Language Models hand the context to a sandboxed Python REPL and let the model write code to partition and query it, which is how DSPy now implements the strategy.
LLM optimization techniques worth reading from January 2026
The January 2026 Towards Data Science pieces I actually used: Ryan Pegoud cutting LLM memory 84% with Triton fused kernels, ACE context engineering, and Hugo Lu on the ceilings at Databricks and Snowflake.