TechNewsFirst

Most AI chatbots write full sentences back to you, even when the actual question was simple, like "is this email spam?" That's slow and expensive, because generating text word by word takes a lot of computing power.

A startup called TypeSafe AI, founded by a guy who used to help build ChatGPT, built something different called Jev. Jev never writes a sentence. You can only ask it three kinds of questions: pick one option from a list, rate something on a scale, or answer yes-or-no. It answers all your questions about something at once, in under half a second, and the company says it's up to 200 times faster and 400 times cheaper than a normal chatbot doing the same job.

Those big numbers come from TypeSafe testing itself, and TypeSafe says so upfront — real-world results will likely be lower. But at least one real company, Vercel, says it got an 18x speedup using Jev for a safety-checking task, which is a much more believable, independently reported number.

Nobody outside the company can check exactly how Jev works, because TypeSafe hasn't published the technical details or released the model's weights. Independent analysts think it's basically the "scoring" part of how chatbots get trained — normally hidden — turned into a standalone product. That could be genuinely useful, or it could be a repackaging of an old idea with a new name.

The bigger idea here: a lot of what businesses use expensive chatbots for isn't really a conversation. It's a simple decision — spam or not, which department should this ticket go to, is this a fraud risk. Jev is a bet that that business is bigger, and more separable from "real" chatbots, than anyone building giant do-everything AI models has assumed.