Book a Free Strategy Call
Skip the read: talk to Walid in 30 min.
Free strategy call. We map your AI engineering team, you keep the notes.
If you want the idea fast, watch Caleb Writes Code's seven-minute explainer first. If you want to build, watch Greg Isenberg's conversation about using Jev for classification next. This page ranks and groups the 14 videos about TypeSafe's Jev worth your time, with each one embedded so you can watch here.
Jev launched on September 15, 2026, and the video response was immediate. We found roughly 370 videos published in the first five days that mention Jev in the title or description, and the biggest single video passed 350,000 views. Most of those are reaction clips. The ones below explain, teach, test or push back.
How we picked these videos
We searched YouTube through a search API for Jev and TypeSafe terms, kept videos published in the first week, and ranked by view count as of September 20, 2026. Then we removed clips that were only reactions, videos where Jev was a passing mention, and non-English videos, because we cannot judge their accuracy. View counts move fast in launch week, so treat every number here as a snapshot.
We describe each video from its title and description, and we say so where the framing matters. We have not audited every claim in every video. Where a title overstates what Jev does, we flag it. If you spot an error in one of our notes, tell us and we will fix it.
Related Reads
The short list
| Video | Channel | Length | Views | Best for |
|---|---|---|---|---|
| JEV Breakdown: The First AI Model Built For Code | Rob Shocks | 10:36 | 359,559 | A fast overview |
| Jev is HERE. How to use it | Greg Isenberg | 28:25 | 337,986 | Using it for classification |
| We need to talk about Jev | Matthew Berman | 12:31 | 299,834 | The mainstream take |
| Jev explained in 7min | Caleb Writes Code | 7:12 | 204,338 | The idea in seven minutes |
| Jev - The Ultimate Classification Model? | Sam Witteveen | 16:19 | 199,296 | Labeling and routing |
| Jev Explained: Demos and Use Cases | Syntax | 17:51 | 197,872 | Demos with code |
| I Tested Jev on 12 Real Use Cases. My Honest Thoughts | Nate Herk | 16:08 | 132,450 | Where it works and where it does not |
| Give me 9 minutes to show you why you really should not sleep on Jev | Maximilian Schwarzmüller | 9:34 | 127,635 | The bull case |
| JEV vs. ChatGPT & Claude | Lucas Montano | 19:34 | 111,444 | Jev vs ChatGPT and Claude |
| JEV: How It Works and What You Can Build | Riley Brown | 20:18 | 106,674 | Routing, classifying, filtering |
| Jev + Claude Code = The Cheapest Agentic Coding Loop Yet | Ray Amjad | 27:28 | 89,558 | Coding agents |
| This New Frontier Model Is Not An LLM | Tim Carambat | 15:26 | 69,549 | How it differs from LLMs |
| Jev by TypeSafe AI: The Model That DOESN'T Write Text | Fazt | 30:50 | 67,941 | Following along |
| Fake Jev demos are taking over the internet | Steve (Builder.io) | 9:31 | 64,213 | Spotting fake demos |
Free weekly brief
Steal our production automations
The exact n8n flows, Claude Code setups, and prompts we ship for clients, broken down step by step. No spam, unsubscribe anytime.
Start here: understand Jev in under 15 minutes
These are the fastest ways to get the idea. The core point, per TypeSafe's docs, is that Jev returns typed decisions with probabilities and does not write text.
Jev explained in 7min
Caleb Writes Code, 7:12, 204,338 views. The shortest video on this list, at about seven minutes. Its description calls Jev a new paradigm and mentions RLHF.
JEV Breakdown: The First AI Model Built For Code
Rob Shocks, 10:36, 359,559 views. The most-watched video so far. The title says built for code, but the description matches TypeSafe's own framing: Jev returns typed decisions and calibrated probabilities, not text. Keep that in mind while you watch.
Give me 9 minutes to show you why you really should not sleep on Jev
Maximilian Schwarzmüller, 9:34, 127,635 views. A nine-minute case that Jev beats LLMs on some jobs. Read "beats" as faster and cheaper on narrow decisions, which is what measured tests support.
Tutorials: how to use Jev
Pick one of these if you plan to build something. Together they cover the API shape, a walkthrough and worked demos.
Jev is HERE. How to use it
Greg Isenberg, 28:25, 337,986 views. A conversation with Ryan Vogel about using Jev for classification, where an input plus questions goes in.
JEV: How It Works and What You Can Build
Riley Brown, 20:18, 106,674 views. Timestamped, so you can jump to the part you need. The description lists routing, classifying and filtering as the main jobs.
Jev Explained: Demos and Use Cases
Syntax, 17:51, 197,872 views. What Jev is, how it works, several use cases and demos, with code shared in the description.
Jev by TypeSafe AI: The Model That DOESN'T Write Text
Fazt, 30:50, 67,941 views. The longest video on this list, at about half an hour.
Tests and use cases
These go past the launch demo and try Jev on tasks.
I Tested Jev on 12 Real Use Cases. My Honest Thoughts
Nate Herk, 16:08, 132,450 views. Twelve use cases tried one after another, with the presenter's own verdict. The title promises the presenter's honest verdict on each.
Jev + Claude Code = The Cheapest Agentic Coding Loop Yet
Ray Amjad, 27:28, 89,558 views. Jev used inside a coding loop with Claude Code. Watch this one if you build agents, and compare its claims with our note on confidence gating below.
Jev - The Ultimate Classification Model?
Sam Witteveen, 16:19, 199,296 views. Frames Jev as a classification model and looks at its System 1 style of speed. The most useful angle if your workload is labeling or routing.
Comparisons and critical takes
Watch at least one of these before you commit. They cover what Jev cannot do and how to spot demos that overstate it.
JEV vs. ChatGPT & Claude
Lucas Montano, 19:34, 111,444 views. Jev set against ChatGPT and Claude. Read it as different jobs, not a winner.
We need to talk about Jev
Matthew Berman, 12:31, 299,834 views. One of the three most-viewed videos of launch week, at 299,834 views.
This New Frontier Model Is Not An LLM
Tim Carambat, 15:26, 69,549 views. Covers why models like this need to exist and how they differ from LLMs in use, with chapter timestamps.
Fake Jev demos are taking over the internet
Steve (Builder.io), 9:31, 64,213 views. About fake or misleading Jev demos. Its description links Jev and Builder.io's own Agent-Native project.
What the videos leave out
Most launch videos show one impressive demo and stop. Four things are worth knowing before you build.
Jev's confidence score is what makes a confidence-gated cascade possible. In our own test across 791 labeled decisions, accuracy on the items Jev was at least 90 percent sure about was 95.5 percent on an 8-way routing task, against 83.8 percent overall. A confidence gate that sends the uncertain items to a larger model is a better plan than trusting every answer. AY Automate has not used Jev on client work, so these are benchmark results, not client outcomes.
The cost story is real but smaller than the headlines when you compare like for like. On short prompts we measured Jev at 40 to 49 times cheaper than a frontier model per decision, and 4.7 to 7.5 times cheaper than the two cheapest small models, not hundreds of times. Median latency was 0.33 seconds against 0.67 to 1.17 seconds for the four models we tested. The raw results are public in the benchmark files.
Jev can be confidently wrong when your labels overlap. In our test all five of its most confident 8-way mistakes were the same pair of intents, and a larger model made the same call. That is a labeling problem, and a confidence score cannot fix it.
Numbers in demos are the builder's own claims. Our index of 1,300+ Jev builds marks them that way and includes the tests where Jev fell short, so you can see the limits as well as the wins.
Which video to watch for your goal
- To decide whether Jev matters to you: Caleb Writes Code, then Lucas Montano's comparison.
- To build your first integration: Greg Isenberg, then Syntax, then our step-by-step guide with working code.
- To wire it into agents: Ray Amjad, then our Jev use cases guide.
- To avoid being fooled: Steve from Builder.io, then the counterpoints group in the builds index.
FAQ
What is the best video to learn Jev?
For the idea, Caleb Writes Code's seven-minute explainer. For building, Greg Isenberg's conversation with Ryan Vogel, which covers using Jev for classification, and Riley Brown's timestamped tutorial.
Are there Jev tutorials for developers?
Yes. Syntax shares code in the description of its demos video, Fazt's video runs about half an hour, and Greg Isenberg's video covers the input-plus-questions pattern. For the API itself, TypeSafe's documentation is the source of truth.
How popular are Jev videos?
The top video passed 350,000 views within days of launch, and twelve videos in our search passed 100,000 views, ten of them listed here. Those figures are snapshots from September 20, 2026.
Do these videos show real results?
Some do and some do not. Tests with a stated method, like the twelve-use-case video, are more useful than single demos. Treat every speed and cost number as the presenter's own claim unless the video shows the run, and compare with independent tests.
Are there Jev videos that criticize it?
Yes. Lucas Montano's comparison and Steve from Builder.io's video about fake demos are the two most useful, and our builds index has a group of tests where Jev underperformed.
Can Jev replace ChatGPT or Claude?
No, and the better videos say so. Jev returns typed decisions with probabilities. It does not write text, read images or reason at length. It works best in front of a larger model, deciding when the larger model is needed.
Where can I see what people actually built with Jev?
In our index of Jev use cases and examples: browser agents, triage, game and drone control, trading loops, eval scorers and open-source repos, each scored by Jev for how specific and checkable it is.
How much does Jev cost?
TypeSafe lists $0.042 per million input tokens with free output. Prices and free windows change, so check TypeSafe's page before you build.
Continue Reading
How to Use Jev: A Tested Step-by-Step Guide With Working Code (2026)
A practical guide to TypeSafe Jev where every request was run against the live API. Setup, the three question types, eight recipes with curl, Python and JavaScript, decision-policy tables, cost math, error handling, a production checklist and when not to use it.
Jev Pricing: What It Costs Per Decision and Where It Is Free (2026)
Jev costs $0.042 per million input tokens with free output. Measured cost per 1,000 decisions against GPT, Claude and Gemini, where it is free until September 25, and how to estimate your own bill.
We Tested Jev on 791 Labeled Decisions Against Four LLMs: Speed, Cost and Accuracy (2026)
An independent test of TypeSafe Jev against GPT-5.4 nano, Gemini 3.5 Flash-Lite, Claude Haiku 4.5 and GPT-5.6 Terra on intent routing and prompt-injection detection. Jev was faster and far cheaper, level with the small models on accuracy, and its confidence score made a cheap cascade work. Code and raw results included.
Book a Free Strategy Call
Building this in production?
Walid runs a 30-min call to map your AI engineering team. Free, no slides.
Free weekly brief
Steal our production automations
The exact n8n flows, Claude Code setups, and prompts we ship for clients, broken down step by step. No spam, unsubscribe anytime.
Robel engineers production-grade automation pipelines at AY Automate, focused on integrations, reliability, and the systems that keep client workflows running.
