Jev, Cowork, and the future of AI tools

This month we're back to building, and we put that checkpoint on the screen.

Two tools, two live demos, no code required.

Cowork is Claude’s “do the work” mode. You hand it a folder and a task and it goes: reading files, drafting documents, sending the summary, coming back when it’s done. It’s the closest thing most of us have to the intern we talked about in August.

Jev, launched in September, is a different kind of AI. It doesn’t write anything. You give it a situation and a bounded question (“should this email go out: yes, no, or ask a human?”) and it returns a decision with a confidence score.

Developers are already using it as the gate in front of their agents, sorting every action into go ahead, stop, or escalate.
Put the two together, and you get a preview of where AI tools are heading. One tool does the work. Another decides what the work is allowed to do. The final say still belongs to a person.

On the table:

A live Cowork run on a real recurring task, with the checkpoints visible
The Jev playground: plain-English decisions, watched end to end
Where “ask a human” has to stay in the loop no matter how good the gate gets
What’s real about a brand-new tool and what’s still vendor claims

Everyone is welcome, from AI enthusiasts to the deeply unconvinced.

Bring your questions and a friend.