@adlrocha - Decision Models: A new primitive in the AI stack

If you’ve been following the space these past few weeks, you’ve probably seen decision models pop up everywhere in your feed. It started in mid-September with TypeSafe’s Jev, a model that, unlike LLMs, doesn’t write a single word as its output (it only picks between options you give it, and tells you how sure it is).Since the release of Jev, I haven’t gone a day without reading something new about it. LangChain plugged it into their agent harness, someone built a copy and paste that understands what you copied, and TypeSafe had it playing Doom at ten decisions a second (for about $7 an hour, if you were wondering).And at the start of October Cloudflare released Clef, an open-weight decision model under an Apache 2.0 licence that anyone can download and run at home (I just downloaded their flash version and I can assure that it can work locally quite well). I didn’t want to write about these models right away, because I wanted to understand the impact that they could have, but seeing…