Read More
Can Nvidia afford to lose Jensen Huang? | Zero Shot Inference | Allen Au
19-11-2026 00:01 HKT
'Human' named Cambridge Dictionary's Word of the Year
19-11-2026 03:01 HKT




Jev, by TypeSafe AI, has been getting hot recently, with about US$40 million (HK$312 million) in seed funding. Jev is not a large language model trained to generate text one token after another. Jev does not write tokens at all!
Founder Diogo Almeida, a former OpenAI researcher, essentially built a decision engine. Feed Jev an email or a ticket and a short list of questions, and the model does not write but returns a choice, a score, or a yes-or-no probability. AI communities are excited because artificial intelligence wastes expensive chat-model tokens on tiny calls, and users often want only the conclusion: Is this user angry? Which tool next? Spam or real? TypeSafe claims it answers up to 200 times faster and 400 times cheaper.
From a technical perspective, it is niche and will be hard-pressed to compete with frontier models by Anthropic or OpenAI.
First, frontier models such as Claude and GPT perform well because they think in steps. Chain-of-thought is not decoration, because hard judgment means noticing the detail that matters, discarding a plausible misread, and then committing. The process costs tokens and the cost is not waste. Strip the trace and you do not get a cleaner judge but a shallower one. On that view, Jev is a model that gave up the thing that makes frontier judges good.
Further, OpenAI and Gemini already sell Flash and Mini models for this layer. They are fast, cheap, and can be locked to a schema. However, Flash models are still token-by-token generators. Jev’s claim is that it scores the options you define in one pass, so it can be quicker and does not charge for a stream of output words. However, that difference only matters if you are making an enormous number of tiny decisions and cannot afford an extra fraction of a second on each one.
Jev may survive as a cheap filter in front of a reasoning model, but it cannot replace frontier models as a reliable judge. If good judgment needs thought, thought needs tokens, and Flash already covers the cheap calls, then Jev looks like new architecture chasing a niche problem the labs already priced down.
Allen Au is a tech startup founder, AI architect, and YouTuber