contextwindow.blog is a technical deep-dive blog on AI systems in production: LLMs, context windows, agents, prompt engineering, evals, and WordPress AI tooling.
Who runs this
Lucas Radke. Product Manager at Automattic, on the WordPress VIP team. Based in Karlsruhe, Germany.
I ship AI-adjacent product work for enterprise WordPress customers. This blog is where I write about the parts of that work that generalize: what held up, what broke, what I would do differently.
What you will find here
Posts on context engineering and prompt design for production systems, agent patterns (Claude Code, MCP, tool orchestration), evaluation discipline, LLM infrastructure and vendor comparisons (versioned and dated), WordPress integrations (Abilities API, Content Guidelines, MCP adapters, Flavor), and honest failure write-ups.
What you will not find: “AI 101” content, listicles, roadmap speculation, or hot takes on company drama.
Who it is for
Working engineers and product technologists with around five or more years of experience. You read API docs, run things from the command line, and prefer a linked paper to a paragraph summary.
How we write here
Every factual claim on this blog has a linked source: a paper, a benchmark, a changelog, a repo, or a measured number. Benchmarks include their date and model version. “GPT-4 does X” without a version and a date is wrong here.
When a post is produced or adapted with AI assistance, the post ends with a disclosure block listing the tools used and their role. Human review is required before publish; disclosure does not substitute for review.
