TechPrufer
Get started
Cole Medin
Cole Medin logoCole Medin · Feature trail

Decision-making AI models

3 videos · Sep 2026 – Oct 2026

What this trail covers

This post details practical use cases for Jev, a class of AI models designed for fast and cost-effective decision-making. It highlights Jev's application in AI coding workflows, including acting as a security hook for coding agents, playtesting games to find bugs missed by unit tests, driving browser testing on real applications, and classifying GitHub issues and pull requests for workflow routing. Jev operates by making decisions between LLM calls, enhancing token efficiency and reducing costs. This post further elaborates on Jev's role as a filter for an AI second brain, handling tasks like email triage, prompt injection guarding, news filtering, and duplicate detection. It emphasizes Jev's ability to provide confidence scores quickly and its integration with cheaper LLMs via OpenRouter for cost optimization.

All Cole Medin trails
  1. Jev Can SERIOUSLY Supercharge Your AI Second Brain (Here's How)This post introduces Jev as a mechanism for supercharging an AI second brain by offloading decision-making tasks from expensive LLMs. It details four specific use cases: filtering emails (draft or skip), acting as a prompt injection guard, filtering a news firehose, and catching duplicates. The author demonstrates the prompts used for Jev and highlights its integration with OpenRouter, allowing for the use of cheaper LLMs and providing confidence scores in under a second. The post also includes a live demo of Jev rating real news articles.
  2. The BEST Jev Use Cases for AI Coding (Absolute Game Changers)This post introduces four specific, practical use cases for Jev in AI coding: 1. Jev as a security hook for coding agents, guarding against unauthorized actions. 2. Jev for gameplay testing, identifying bugs that traditional unit tests miss. 3. Jev acting as an intermediary between LLMs for decision-making, improving token efficiency. 4. Jev for classifying GitHub issues and pull requests to route tasks and select appropriate models within Archon workflows. The post also provides links to related GitHub repositories ('jev-ultrafast', 'jev-use') and documentation ('TypeSafe Jev docs').
  3. Jev is the FIRST of a Whole New Class of AI Models (Here's How to Actually Use It)This post introduces Jev, a novel AI model class focused solely on decision-making. It details Jev's capabilities, including taking state and typed questions to output decisions with probabilities in ~200ms for a fraction of a cent. Practical applications demonstrated include playing an indie game at 3 decisions/sec, triaging PRs, and routing models for AI agents. The post also links to related projects like Firecrawl, Dynamous's Agentic Coding Course, TypeSafe Jev, Archon, and various Jev demos.