RAG Failure Maps
A retrieval-first exploration of where RAG systems break, from weak chunk boundaries and noisy evidence to grounding failures and misleading answer confidence.
Read design noteLab
Exploratory design notes on retrieval, extraction, evaluation, and tooling.
These notes describe questions and candidate workflow patterns. They do not provide linked implementations or reproduced evaluation results, and should be read as conceptual material rather than demonstrated systems.
Grounded Extraction Agents
A conceptual note on multi-step extraction, structured outputs, and evidence checks. No working agent implementation or comparative reliability result is established here.
Read design noteDSPy Retrieval Loops
A conceptual note on DSPy-style retrieval and prompting loops, emphasizing evaluation objectives and the evidence needed to establish an improvement.
Read design noteLocal AI CLI
A conceptual note on composing local AI-assisted document tasks through a command line, with explicit inputs, outputs, and workflow boundaries.
Read design note