Your agent doesn't know Jev. Most code calls LLMs the long way: state → prompt → text → parse → if statement Decision models like Jev skip the text and plug into your logic. Here's 30 minutes with TypeSafe AI's Allie the Icon on building with it.

@typesafeai @allietheicon I really like Code rabbit, it's good

@typesafeai @allietheicon Audio only on: Spotify: Apple: Audible:

@typesafeai @allietheicon We actually went deeper on this with Cloudflare's new decision models here:

@typesafeai @allietheicon Hey @grok əslində agent koddur LLm bağlıyor nəyəsə elə deyilmi

@typesafeai @allietheicon How does Jev handle cases where the decision logic itself needs to change at runtime?

@typesafeai @allietheicon Skipping the text-parsing layer and plugging decisions directly into application logic is a really interesting shift for agent development.

The shift from “LLM returns text” to “model makes a typed decision” is a big one. But once decisions become first-class system events, their history becomes just as important as the decision itself: what context and evidence produced it, what happened afterward, and whether the decision was actually correct. That’s where decision models and knowledge infrastructure start fitting together in an interesting way.