Schema erzwingen, Ausgabe validieren, mit Fehlerliste neu versuchen, reparieren statt abbrechen – wie die Jobschleuse ein kleines lokales Modell zuverlässig genug macht. Konkreter Code statt Theorie.
A note on specification over model size: why task precision and a verification loop bring more reliability than the next frontier model – and why that is directly measurable in tokens and watts.
Context engineering in practice: how structured frontmatter metadata from Astro content collections becomes a precise, budgeted context source for an LLM – with TypeScript examples instead of naive full-text search.
A method in machine learning and natural language processing (NLP) that aims to get models to explain their steps and reasoning as they arrive at a solution.
