Did OpenAI just clone Jev? Everything you missed from OpenAI DevDay Transcript, AI Summary & Key Points

00:00 Last week, Sam Alman sat frustrated in his bunker as he watched Zuck unveil Muse, a free personal assistant inspired spy that trains on everything you tell it in exchange for doing your shopping. Not to be outdone, Sam got on a stage of his own in San Francisco just a few days ago and announced that for $100 per month, they'll give you another cute and totally harmless personal assistant that can respond to all of those AI generated emails in your inbox and rewrite your backend in whatever language DHH thinks is fast

00:27 that day. In today's video, we'll break down everything else you missed at OpenAI Devday, including a model that's almost as smart as GPT6 Astra, an API that looks suspiciously familiar. >> My name is >> Sam Alman, >> and a new subscription tier that quietly castrated the $200 per month plan. It is October 1st, 2026, and you're watching the code report.

00:46 But back in January, a bored and recently retired Austrian developer named Peter Steinberger released Claudebot, a first of its kind AI personal assistant for people who like to appear productive without the hassle of actually doing the work. It quickly became the fastest growing repo in GitHub history, which was enough to get him hired by OpenAI to continue to work on the idea.

01:05 Then just last week, all his hard work finally paid off when Zuck took the concept and stuffed it into Muse. And just a few days ago, his own employer did the same thing, but now with a $100 per month price tag that they're calling DOTs with a singular DOT being a personal agent powered by GPT6 Astra. So, it's basically just a bot with the B flipped around.

01:25 You just give it a name, then its cute squishy little avatar will get to work ruining the internet as we know it. In their keynote demo, Sam told his Alfred to remove an old inventory API before it got shut down. So Alfred then followed modern software best practices of tracing dependencies, updating integrations, running tests, and spamming the repo with three PRs from the CEO that some poor dev will feel obligated to review because they know it'll be the first time human eyes have ever looked at that code.

01:53 Then Holly Lee came out to prove that DOTs work outside of a pre-recorded video and instead just proved that live demos are still hard even in the age of slop. But my favorite part of Dots is when the internet realized that Dots.com is owned by some woman's clothing brand that died in 2014 and dot.com is owned by XAI and now redirects to Grockbots's landing page.

02:15 It's the best domain prank I've seen since whoever got their hands on nex.js.dev and remix.dev a few years back. But if dot sounds like your slob cannon of choice, you'll need a pro subscription which OpenAI revealed some new pricing for. The new Pro 500 plan costs $500 per month and gets you 25 times the usual usage of Chat GBT Plus along with access to Ultraast, which is a new speed tier that runs Astra at 300 tokens per second.

02:41 In the API, Ultraast costs $60 per million tokens in and 300 out. And when Sam said those words out loud, the audience of grown men moaned, >> "You know what? It's worth it." >> Which apparently isn't the first time he's caused that reaction. Meanwhile, the $200 plan went back on sale, but had its usage chopped in half. Luckily, if you're broke, they announced GPT 6.1 Soul, which they're describing as near Astro Intelligence for a fifth of the price.

03:06 You get $2 per million tokens in and $10 out, which is the exact same price of GPT6 Soul, which launched a week ago. Then Sam announced the new decisions API where the idea is that instead of asking a big model to write a paragraph and then parsing the JSON out of it with a retry loop, you hand their tiny Luna model a fixed list of answers and it picks one in a fraction of a second.

03:28 For example, imagine you're building a dating app for gay donkeys. As a way to fund the venture, every message needs to be categorized by how useful it would be for blackmail. Historically, you'd send each message to a full language model, wait a few seconds while it pretends to be ethical, then hope the JSON comes back in the right shape. With a decision model, you give it a finite set of answers up front.

03:47 It returns one of them in a few hundred milliseconds. And since it physically can't answer anything outside your list, there's nothing to parse or retry. And if all this sounds familiar, it should because just 2 weeks ago, we released a video about a company called Typesafe AI, founded by an OpenAI researcher who just released Jev that does the exact same thing.

04:04 But they didn't stop there. Later in the event, they announced a Google Docs clone called Pages, a Notion clone called Space, and a Slackbot that works a lot like Claude Tag, showing what it looks like when you have infinite tokens. But the one thing in the keynote that could actually make you money is signin with chat GPT, which lets your users log into your app with their chat GBPT account and burn their own tokens instead of yours.

04:27 It launched with 16 partners including Devon, Notion, Verscell, and OpenClaw. And comes with an implicit understanding that if you get too popular, they'll just clone your app, too. Which is why if you need to build a decision model into your own app, you might want to own the weights yourself. And you can do that with Fastino Labs, the sponsor of today's video.

04:44 Jev has been going viral for giving developers fast type decisions, but it's closed weight and only runs on the company's cloud. The Fastino's Gliner 2.5 to side is an openweight model that makes those same calls locally and up to eight times faster. This so you can train your own Jevlike models and own the weights. It comes from the Gliner family which has been open- source since 2024 with over 50 million downloads.

05:08 I added their skill to open code in about 5 minutes so my agent can route each developer request to a coding model or internal tool and pull out the repo file path and function names. Other developers have used it to run a browser agent 36 times cheaper than Jev. And Fastino just released a brand new model called Glide that beats Jev on accuracy across intent routing, fact-checking, and hallucination checks by a pretty wide margin.

05:33 You can host these models anywhere, but Fastino's inference API is the quickest way to replace your Frontier model. Try it out for free with our code at the link below. This has been the code report. Thanks for watching and I will see you in the next one.