OpenAI just opened its Decisions API to every developer and bolted on image input, a direct shot at a three-week-old startup called TypeSafe that somehow talked Andreessen Horowitz into an $870 million check at a $7.5 billion valuation before most of the industry had heard its product’s name.
The product in question is Jev, TypeSafe’s “decision model,” and it’s worth explaining what that actually means because it’s not another chatbot. A decision model doesn’t generate free-form text. It picks, with a confidence score, from a fixed list of outcomes you hand it: approve or deny this loan application, route this support ticket to one of six queues, flag this transaction or don’t. That narrower job sounds less flashy than writing essays or generating video, but it’s the exact shape of task that insurance underwriting, fraud review, content moderation and a huge share of enterprise software already run on, usually with brittle rules engines or expensive human review behind the scenes.
TypeSafe pitched Jev as faster and more reliable than asking a general-purpose large language model to do the same job and then regex-parsing its answer out of a paragraph of hedging. Bloomberg Government, reporting on the a16z round, described Jev as winning early enterprise pilots specifically because its outputs are structured and auditable by design, not because a prompt engineer managed to coax the right format out of GPT-5 or Claude on a good day.
OpenAI’s response arrived fast, and it’s hard to read the timing as coincidence. The Decisions API had been in limited preview; it’s now generally available, and OpenAI added image input so the same fixed-choice logic can run against a photo, a scanned document or a screenshot rather than just text. That’s a meaningful gap closer, since a lot of the real-world decision tasks TypeSafe is pitching, claims adjustment, document verification, retail returns, involve looking at an image before deciding anything.
The size of TypeSafe’s round is the part generating the most talk among people who track enterprise AI funding, and for good reason. $870 million for a company that’s been shipping product publicly for roughly three weeks is an eye-watering number even by 2026 standards, and it says more about investor anxiety over missing the next category than it does about Jev’s current revenue. a16z has made a habit this year of writing large, fast checks into narrow, defensible AI product categories before the giants can route around them with a feature update, and this is arguably the cleanest test case yet of whether that strategy works: OpenAI shipped a competing capability within the same news cycle as the funding round closed.
“The entire bet behind backing a company like TypeSafe is that the foundation labs are too slow and too distracted to copy a narrow, well-executed product,” one investor familiar with the round told Bloomberg Government. “OpenAI just tested that bet in real time.”
It’s worth being precise about what OpenAI did and didn’t do here. Opening a preview API to general availability and adding image support is a real product shift, but it isn’t necessarily a clone of Jev’s specific architecture or its enterprise integrations. TypeSafe’s pitch to customers has leaned heavily on compliance tooling, audit trails formatted for regulators, and support for industries like insurance where a wrong automated decision carries legal exposure. OpenAI hasn’t announced anything matching that layer yet. The two products currently look more like they’re converging on the same category than copying each other feature for feature.
That distinction matters for how this actually plays out. If the real moat in decision-model products turns out to be the compliance and audit tooling wrapped around the core model, rather than the raw accuracy of the decision itself, TypeSafe has room to defend its position even against a well-resourced competitor. If the moat is just “fast, structured output from a fixed list of choices,” OpenAI’s scale and distribution make this a very short-lived advantage for the smaller company.
There’s a broader pattern here that’s become familiar to anyone tracking the AI funding cycle this year: a narrow, well-scoped product category gets validated by a splashy raise, and within days one of the big labs ships a version of the same capability bundled into an existing platform most enterprise buyers already have a contract with. We saw a version of this dynamic play out when Anthropic partnered with private equity firm Hg to spin up a dedicated European AI consultancy rather than trying to build that go-to-market motion in-house. Speed to market and distribution keep beating raw technical novelty as the deciding factor in who actually wins a category.
For enterprise buyers evaluating either product right now, the practical advice is unglamorous but accurate: pilot both against your actual decision data before signing anything. Structured decision accuracy varies enormously by domain, and a model that performs well sorting support tickets won’t automatically transfer to underwriting insurance claims. Ask both vendors for their error rates on edge cases, not just their marketing accuracy numbers, and get clarity in writing on what happens, legally and operationally, when the model gets a decision wrong.
It’s also worth asking why a16z moved so fast on this one specifically. The firm has backed plenty of AI infrastructure plays this year, but the speed of the TypeSafe deal, term sheet to close in roughly three weeks according to people familiar with the round, suggests a fear of losing the deal to a rival fund more than it suggests unusual conviction about the product. That dynamic, large funds racing each other into deals before diligence can catch up, has been a quiet undercurrent of the 2026 AI funding market, and it’s part of why valuations keep detaching from any conventional revenue multiple. Jev may well deserve the attention. It’s harder to say it’s deserved a $7.5 billion price tag after three weeks on the strength of pilot programs alone.
The bottom line: TypeSafe proved a narrow AI product category can be worth $7.5 billion three weeks after launch. OpenAI proved it can ship a credible competing feature within the same news cycle. Neither fact settles who wins the decision-model market, but it’s a preview of how fast every narrow AI niche is going to get contested going forward.
The Teck Hustlers Editorial Team covers technology, business, finance and gaming news for readers who want the story behind the headline. Every article is researched, written and edited by our in-house team, checked against company announcements, filings and reputable reporting, and updated when new facts emerge. We focus on what each development means for founders, professionals, investors and everyday users. Have a tip or a correction? Reach us through our contact page.
