Gemini 4 Argon: what to use it for—and what not to
Where long-horizon frontier intelligence belongs, how to operate it safely, and when a smaller model is the better engineering choice.
Read the field note// Field notes
Writing from
the work.
Practical analysis of emerging AI models, system architecture, local intelligence, automation, digital products, and the systems behind growth.
When a decision model beats a chatbot
A practical guide to TypeSafe Jev and Cloudflare Clef, with production patterns for routing, scoring, gating, and review.
Read the field noteOne workflow. More than one model.
A model-routing framework for choosing code, decision models, fast generation, frontier reasoning, and human authority.
Read the field noteWhat the AIDA Innovation Hub taught me about building local AI
From more than 150 submissions to a ten-minute finalist pitch: the lessons behind Rostock.ai’s biggest milestone so far.
Read the field noteLocal AI needs trusted data before it needs a bigger model
Why source quality, local ownership, and review workflows matter more than another layer of generic intelligence.
Read the field noteAutomate the workflow—not the confusion
A practical sequence for finding the repeated work worth automating before choosing tools or adding AI.
Read the field note