Summary
Cole Medin shows how he uses Jev, a fast decision model, to power his AI second brain, focusing only on the use cases that have made a real difference after sustained daily use.
Unlike a large language model, Jev does not generate text. It takes a state and a list of multiple-choice questions and answers them in parallel with confidence scores. Medin says this makes it 20 to 200 times faster and 40 to 1,000 times cheaper than an LLM for decision tasks. He accesses it through OpenRouter’s decisions endpoint and reports 38,000 calls and 55.3 million tokens in a single week, at a total cost of about $21.
The main use cases center on filtering the flood of incoming information. Jev screens every email for prompt injection attempts, such as requests to forward data, edit memory or settings, or impersonate the owner, and quarantines messages above a confidence threshold. It also decides which emails merit a drafted reply and which to skip, all in under a second. Replies are then drafted with a lean open-weight model like GLM 5.3 Flash, keeping the system cheap and preserving rate limits for coding agents.
Viewers can borrow the patterns directly or use them as inspiration for their own automations.
📺 Source: Cole Medin · Published October 08, 2026
🏷️ Format: Workflow Case Study
