Orgo co-founder explains what Jev is and how it cuts costs
Podcast notes on Jev as a multiple-choice decision maker that returns confidence scores, using GPT-6 as the brain and Jev as the execution layer, with cost comparisons for Zoho scraping, Minecraft and X monitoring.
Jev clearly explained (and how to make money with it): The 80/20: keep your frontier LLM as the brain and swap in Jev as the hands. Jev doesn't write text. It picks from multiple choice options you define, much faster (and MUCH cheaper) than an LLM. @nickvasiles is the co-founder of Orgo, and he came on the pod to break down what JEV actually is and where it makes you money. Here's what I learned: 1. Jev is not an LLM. It's a decision maker. You give it the question and choices A, B, C, D, and it answers with a confidence score. It can't output text at all. 2. It's a money-saving model, not a money-making model. If you're running an AI offer at 0% or 50% margin today, Jev can push that to 90 to 99%. 3. His proof: a computer use agent that scraped 60 products into a Zoho CRM cost $30/hour and ran 2 to 3 hours per batch. Rebuilt on JEV, he estimates under 10 seconds for practically free. 4. It played Minecraft for six pennies. Full computer use, tools defined by GPT-6 Astra, executed by Jev. 5. The "look now" detector: 16 Orgo computers watching X in real time, 4 criteria, Jev fires an alert when a model release matches 3 of 4. Same job on a frontier model would cost hundreds of dollars a day. 6. The architecture is GPT-6 as the brain, Jev as the execution layer. That's the harness Orgo is building around it for computer use. 7. Where it fits: high volume, structured inputs. Insurance claims by the thousand. Monitoring every skincare brand in your category for price drops. Online arbitrage scanning Kohls and Ulta for pennies. 8. On its own it's dumb. He asked it to play chess and it made exactly one move, instantly, then stopped. You have to build its brain out before it can act. 9. If you're not already making money with AI, Jev is a distraction. Go build an agent that's really good at one workflow for one customer first. 10. Ask your existing agent to figure it out. His Hermes agent Dewey studied the Jev docs, integrated the API, and built the demos himself. His 2 key takeaways: 1. Bet on cost going to zero. He sold $5K/month managed agents with unlimited tokens at break-even because model costs only move one direction. Jev is that bet paying off. 2. Don't get FOMO. It's already been open sourced and the big labs will package it into the products you already use. Build the workflow now, swap the engine later. Nick is building at the edge of computer use and we had a blast going deep on it. Go follow @nickvasiles Full video below. https://www.youtube.com/watch?v=5WlnNuVduKw (Also available on the Build With AI podcast wherever you get your pods)