Dead code was always a nuisance, but AI coding agents have made it dangerous. LLMs don't distinguish between live and dead code — they see every line as an existing implementation to follow, maintain, or stay compatible with. This changes the economics of code cleanup entirely.

Struktur is a fresh attempt at building a reusable tool for LLM-based structured data extraction, informed by everything learned while building Data Wizard at Leuphana University. This post traces what changed, what stayed, and why.

Struktur is an all-in-one tool for extracting structured data from documents using LLMs. It handles chunking, validation, retries, and merging through configurable extraction strategies, available as both a shell-composable CLI and a TypeScript SDK.

The Remote Labor Index claims to measure AI's capacity to automate economically valuable work, but its methodology – zero-shot performance with limited tools and time – measures something far narrower than that claim suggests.

An article describing a WebSocket-based pub/sub approach designed for real-time coordination of distributed AI agents as an alternative to traditional polling mechanisms of orchestrators like OpenClaw.

A thesis exploring the technical feasibility of developing a flexible tool for provider-independent structured data extraction using Large Language Models.

Laravel Introspect is a tool that allows querying for structured information about Laravel codebases either using a fluent API or natural language queries. Existing static analysis focus on catching errors and are less flexible for modern LLM-based development workflows.