Jev is HERE. How to use it
Jev is a new AI classifier model created by Dooo Almeida that makes fast, low-cost decisions on structured inputs rather than generating text like traditional LLMs. It processes information through a defined schema to return probability-based classifications across multiple categories, enabling applications from email triage to content clipping to flight booking automation.
Summary
The podcast features Ryan Vogle from the founding team of Open Code explaining Jev, a fundamentally different type of AI that functions as a decision-making classifier rather than a conversational language model. Unlike LLMs that predict the next token and stream responses, Jev takes a defined input, evaluates it against a schema of possible outputs, and returns probability scores for each option—for example, categorizing an email as 80% spam, 10% marketing, 10% finance.
Ryan demonstrates Jev's capabilities through a practical email classification example where 1,700 emails were categorized across four dimensions (category, priority, spam score, and reply percentage) for just 18 cents. This illustrates Jev's key advantages: exceptional speed (approximately 200 milliseconds per query regardless of complexity), extremely low cost (demonstrated with a $5 credit sustaining heavy usage for 2 days), and type-safe structured outputs that integrate directly into code.
The conversation explores multiple business applications: email lead filtering for a graphic design agency, support ticket triage and routing, form-based instant quoting for service businesses, YouTube video clip extraction by identifying interesting moments in transcripts, and browser automation for travel booking. Ryan emphasizes that any workflow involving human decision-making on incoming data could benefit from Jev's instant classification.
Ryan acknowledges limitations, noting that Jev performed poorly on Bitcoin trading signals compared to GPT-4o, suggesting it excels at routing and classification decisions rather than complex reasoning tasks. He characterizes Jev as a decision model specifically, not a general-purpose reasoning system, and demonstrates creative applications like using letter-by-letter classification to generate text, though noting this isn't Jev's primary strength.
The episode concludes with access information: Jev is available through the Vercel gateway for immediate testing, and users are encouraged to identify workflows in their own businesses that could benefit from fast, cheap classification automation.
About this episode
In this episode, I talk with Ryan Vogel about Jev, a new type of AI built for classification. Ryan shows how Jev takes an input plus an output schema and returns a probability for each choice in about 200 milliseconds. He demos Jev sorting 1,700 emails for 18 cents total, then covers lead scoring, support routing, video clipping, and browser control. I push him on the startup angle: find a business with an expensive queue of incoming information and put Jev at the front of it. You leave with a clear mental model, real use cases, and a simple way to try it today. Links Mentioned: Jev/Typeface AI: https://typesafe.ai AI Gateway: https://vercel.com/ai-gateway Timestamps 00:00 – Intro 02:27 – What Jev Is and Why It Matters 04:32 – Email Triage Demo 07:19 – Jev as an AI Decision Maker 15:46 – How to Use Jev in a Business 20:48 – Startup Idea: Local Services Matching and Instant Quotes 22:51 – Use Case 1: Bitcoin Signal Test and Limits 24:03 – Use Case 2: Auto-Clipping Long Videos 25:27 – Use Case 3: Browser Control: Flight Pick in 7.1 Seconds 26:18 – How to Get Access 27:25 – Closing Thoughts Key Points • Jev is a classifier: an input and an output schema go in, and a probability for each choice comes out. • Ryan's demo scores 1,700 emails for 18 cents total. • Each Jev query takes about 200 milliseconds, whatever the input and output structure. • Use Jev at any point where a business makes fast, repeatable decisions on incoming data. • Keep Jev in an advisory role, and save frontier models for high-intelligence tasks like trading. • Instant access runs through the Vercel Gateway, and a waitlist covers direct access. Numbered Section Summaries 1. What Jev Actually Is Ryan explains Jev as a classifier. You define an input and an output schema, and Jev returns a probability for each option, for example 80% orange, 10% red, and 10% blue for the color of an iPhone. 2. The 18-Cent Email Demo Ryan runs Jev on 1,700 of his own emails and scores each one for category, priority, spam score, and reply likelihood. The run uses 4.2 million input tokens and 500,000 output tokens and costs 18 cents total. 3. A Decision Model, Distinct From a Chat Model Jev returns only structured output, meaning the numbers and categories from your schema, with zero visible reasoning. Developers can drop that type-safe output straight into code. Ryan also tests Jev as a letter-by-letter text generator to show how a decision model differs from a chat model. 4. Cheap Enough to Experiment A $5 intro credit lasts Ryan's team two days of heavy use, and he estimates $10 could last about three months. His girlfriend's graphic design agency uses Jev to score contact-form leads on a scale from 0 to 1, so high-value leads get a fast reply. 5. The AI Traffic Cop I frame Jev as a traffic cop: information comes in, and Jev decides what it is, how important it is, and what happens next. A high-confidence lead goes to a human, a lower score goes to automation or an LLM, and the lowest scores get ignored. Ryan adds support routing, where 200-millisecond answers replace slow streaming responses. 6. Put Jev at the Front of the Queue I ask how to find a business with an expensive queue of incoming information and put Jev at the front of it. Ryan describes a local services platform that matches a request like "I need my driveway power washed" to the best nearby business and turns "instant quote" forms into truly instant quotes. 7. Where Jev Struggles Ryan's Bitcoin test, which asks Jev to buy, hold, or sell every minute, performs poorly. OpenAI's latest frontier model does a little better because it cross-references news. Ryan's advice: use Jev for routing-style decisions and keep it away from your portfolio. 8. Clips, Browser Control, and Getting Started Ryan's video clipper scores 17 moments in about three seconds, and he builds it in about 10 minutes. A Browser Use demo shows Jev picking a flight from Zurich to London in 7.1 seconds. To start, use the Vercel Gateway and ask your AI agent which of your daily workflows could use a decision maker like Jev. The #1 tool to find startup ideas/trends - https://www.ideabrowser.com LCA helps Fortune 500s and fast-growing startups build their future - from Warner Music to Fortnite to Dropbox. We turn 'what if' into reality with AI, apps, and next-gen products https://latecheckout.agency/ FIND ME ON SOCIAL X/Twitter: https://twitter.com/gregisenberg Instagram: https://instagram.com/gregisenberg/ LinkedIn: https://www.linkedin.com/in/gisenberg/ FIND RYAN ON SOCIAL X: https://x.com/ryanvogel Youtube: https://www.youtube.com/@vogeldev/videos
Key Insights
- Jev processes 1,700 emails with complete categorization across multiple dimensions for 18 cents, with 4.2 million input tokens and 500,000 output tokens, demonstrating the dramatic cost advantage over traditional LLM-based solutions
- Jev returns probability distributions rather than binary decisions—for example, 80% orange with 10% red and 10% blue—allowing for confidence-weighted classifications that reflect uncertainty rather than false certainty
- All Jev queries complete in approximately 200 milliseconds regardless of input/output complexity, compared to traditional AI that can take 30 seconds and requires streaming listeners, enabling real-time decision routing in applications
- Jev doesn't generate text or perform internal reasoning—it only returns structured data matching the defined schema, which distinguishes it from LLMs but also limits its application to decisions that don't require complex explanation or reasoning
- A browser automation agent using Jev selected a flight from Zurich to London in 7.1 seconds, compared to minutes for other browser-use AI agents, demonstrating speed advantages in sequential decision-making tasks
Topics
Transcript
[0:00] Jev is here and it's a big deal. It was created by Dooo Almeida. Yes, that's the same guy whose research built chatbt. Now, it's such a big deal because it's a whole new way to do AI. So, I brought on my friend Ryan who's on the founding team of Open Code to just come on and clearly explain what Jev is, what are some insane use cases, and break down some startup ideas that are now unlocked. As of publishing this, Jev is invite only. But good news, by the end [0:31] of the episode, you're going to see how you can get access today. So, you're going to want to like, comment, and subscribe right now…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Greg Isenberg
Instinct AI is For Real. What You Need to Know.
Hosts Greg and Remy review Instinct AI, an invite-only personal agent app raising at a $10 billion valuation, demonstrating its capabilities for booking flights, restaurants, obtaining visas, and other personal tasks. They highlight its impressive simplicity, human-like communication style, and network effects through its trusted agent connectivity, while flagging privacy and data retention concerns.
Building a Software Factory that actually works (Full Course)
Ross Mickey explains how to build a software factory—a systematic workflow using AI agents to ship high-quality software at scale. The factory concept uses markdown files to define agent behavior and follows four key steps: isolate (branching), build (with code structure guidelines), prove (with before/after testing), and ship (with automated code review).
Build Bigger with GPT-6 Astra (here's what to do)
This episode explores practical applications of GPT-6 Astra beyond flashy demos, featuring nine actionable prompts for productivity and business, and demonstrating how Astra enables hardware projects through AI assistance with design, code generation, and supplier sourcing.
What are Agentic Loops?
Professor Ross Mike explains agentic loops — AI systems that self-iterate without human input — and argues they are largely impractical for most developers due to high token costs and poor assumption-making. He distinguishes between 'human in the loop' (iterative, guided development) and fully automated loops, recommending loops only for constrained, binary-feedback processes like code review. He shares his own GP loop use case as a rare valid example.
Clearest Explanation of AI Agents as Customers
The speaker argues that AI agents are becoming the primary 'customers' of the internet, replacing human users as the dominant force driving web traffic and commerce. This shift requires entirely new infrastructure — from agent-native inboxes and wallets to machine-readable websites — creating massive startup opportunities. The speaker frames this as a bifurcation of the internet into a human layer and an agent layer.