Field notes · Vahue research
Insights from the AI-Native frontier
Engineering playbooks, enterprise AI patterns, and lessons from shipping production LLM systems — written by the Vahue team.
What it feels like to work with Fable
Claude Fable 5 is a new Mythos-class model for long-horizon agentic work. I tested the shift on the recent Nvidia's "Hack for Impact" hackathon in London — PowerRanger, an EV fleet flexibility agent. While getting a win in the Economics category, got a few quirks.
Latest articles
Trump's AI order, decoded
A plain-English read of the June 2026 Executive Order on advanced AI innovation and security: the 30-day cyber sprint, voluntary frontier-model framework, and explicit bar on AI model licensing.
Deploying LLMs Securely in Enterprise Environments
A practical guide to integrating large language models with sensitive business data while staying compliant and secure.
Evaluating Code Data Sources for Training Large Language Models
A practical comparison of the major code dataset sources — from open-source repos to dedicated coding teams — and how to choose the right one.
The Case for Human-Written Code in LLM Training
Why human-authored code remains essential for building reliable coding assistants — and where synthetic data falls short.