TypeSafe AI and Diogo Almeida
A ChatGPT co-inventor spent two years in stealth asking why superhuman chat models never produced superhuman automation. Jev is his answer.
- Company
- TypeSafe AI
- Founder
- Diogo Almeida
- Background
- ChatGPT and RLHF co-inventor, formerly at OpenAI
- Launched
- September 15, 2026
- Time in stealth
- 2 years
- Access
- Early access, waitlist at typesafe.ai
Need a model that does generate?
Jev AI returns decisions, never pixels or prose — that is the whole point of it. When you need the other half, Pollo AI runs the frontier video and image generators side by side in one place, so you can compare outputs before committing to any single vendor. Different tool, opposite job to Jev AI.
- Free videos
- 4
- Free images
- 20
- Models
- 5+
Free plan is watermarked; paid plans start around $10/month.
We may earn a commission from this link.
The founder behind Jev AI
Diogo Almeida is credited as a co-inventor of ChatGPT and of RLHF, having worked at OpenAI on the methods that turned a research capability into something people could use. Jev AI is what he built next. That is a genuinely unusual starting point for the company behind Jev AI whose entire pitch is that conversational models were the wrong tool for the job everyone tried to use them on — the critique carries more weight coming from someone who helped build the thing being criticised.
His framing of the Jev AI problem, repeated across the launch materials, is a question rather than a claim: if models have been superhuman at chat for years, why has automation not followed? The answer he spent two years building toward is Jev AI: the industry kept reaching for deliberation when what the work needed was fast, cheap, calibrated judgement, and Jev AI is an attempt to supply exactly that.
Why the name TypeSafe fits Jev AI
The name is a borrowing from programming languages, and it is doing real work rather than sounding clever. For Jev AI the borrowing is literal. In a type-safe language, a whole class of errors cannot be expressed: you do not check at runtime whether a string snuck into an integer field, because the compiler made that state unreachable. TypeSafe makes the same claim about Jev AI output, and the Jev AI schema is what enforces it. You do not validate that a Jev AI answer is one of your categories, because a value outside them is not representable.
Anyone who has wrapped a language model call in a parser, a validator, a retry loop, and a fallback will recognise what Jev AI promises. All of that scaffolding exists to defend against output that might be anything at all. Jev AI removes the possibility, and the scaffolding has nothing left to do.
It is worth being precise about the boundary of the Jev AI guarantee, since the name invites overreading. Type safety in a language does not mean your program is correct; Jev AI type safety means certain errors are impossible. The same applies to Jev AI. Fabrication is off the table for Jev AI. A wrong Jev AI answer inside your schema is not.
Advertisement
What to watch next for Jev AI
Three things will tell you whether Jev AI holds up. The first is independent evaluation: launch numbers come from the vendor, and until outside teams publish Jev AI accuracy and calibration figures on their own data, the performance story is a claim rather than a result.
The second is whether Jev AI pricing survives contact with real load. An early-access rate card is a hypothesis about unit economics, and the test is what it looks like a year after the waitlist opens.
The third is whether anyone else builds a rival to Jev AI. A category with a single occupant is a product; a category with three is a shift in how this class of work gets done. That question will resolve itself over the next year without anybody having to argue about it.
Reading a stealth launch carefully
Two years of silence followed by a launch with a named category, a named training method, and a rate card that undercuts the field by two orders of magnitude is a well-constructed debut. That is not a criticism — it is what a serious company does — but it is worth separating what has been demonstrated from what has been asserted, because at launch almost everything is the latter.
Demonstrated: Jev AI exists as a product, developers are being let in off a waitlist, and the published Jev AI interface does what it says. Everything else about Jev AI is assertion. Asserted: the Jev AI accuracy, the calibration quality, the per-decision cost comparisons, and the claim that the economics hold without subsidy. None of those are unreasonable, and none of them are yet checkable by anyone outside the company.
The good news is that Jev AI is unusually easy to evaluate yourself. You do not need a benchmark suite or a research team. You need a few hundred cases you already have labels for, an afternoon, and the willingness to look at a calibration curve. For a fraction of a cent per Jev AI case, the cost of finding out is lower than the cost of the meeting where you would have debated it.
The one Jev AI risk that is not resolvable by testing is dependency. A single-vendor, closed, waitlisted model with no alternative supplier is a concentration of risk regardless of how well it performs, and it is worth designing the Jev AI integration so the decision layer can be swapped — a thin interface, your thresholds in your own code, and your labelled evaluation set kept somewhere you control.
Who funded Jev AI, and who built it
- Founded
- 2024, San Francisco
- Funding
- $40M seed round
- Lead investor
- DCVC
- Co-founders
- 3
Diogo Almeida
Co-founder and CEOCo-inventor of RLHF and ChatGPT at OpenAI
Erik Gafni
Co-founderNamed in the funding announcement
Sasha Sheng
Co-founderNamed in the funding announcement
Most coverage of Jev AI names only the CEO, which understates the Jev AI team. Gafni and Sheng appear as co-founders in the funding announcement and in independent reporting, and the company has published no broader roster — no headcount, no research team, no advisors. For a lab whose central Jev AI claim is a novel training method, the absence of a named research team is the gap most worth watching.
The $40M is a vote of confidence in the architecture and the founders rather than evidence about the economics. The announcement disclosed no revenue, no named customers and no valuation. What would actually move the Jev AI picture is a deployment with published error rates — Jev AI holding up on data the company did not choose.
Why Jev AI is called Jev
Jevons observed in 1865 that more efficient steam engines made Britain burn more coal, not less: cheaper use unlocked uses that had never been economic. Naming the model after that is a thesis statement. TypeSafe is not arguing that Jev AI lets companies spend less on AI — it is arguing that a hundredfold drop in the cost of a decision makes companies automate decisions they never bothered to automate, and that total spend rises rather than falls.
That framing is worth holding onto when you read the Jev AI pricing. The argument is not that Jev AI shrinks your AI bill; it is that Jev AI changes which problems are worth deciding at all. With Jev AI, sampling becomes full coverage, nightly batches become inline checks, and questions nobody asked because the answer was not worth the compute start getting asked on everything.