Typesafe + Taquet
Available over MCPFast, typed judgments instead of another paragraph of prose.
What Typesafe does
TypeSafe serves Jev, a System One model: it reads a piece of state you give it and returns typed answers (a choice, a yes/no probability, or a position on a scale) with their confidence, in about 100ms. It does not write text or reason out loud.
How it helps inside Taquet
Use `jev_judge` whenever a magik needs to decide instead of write: pick one option among known candidates, check whether a claim is actually supported by a source text, rate something on a scale you define, or choose a value your code already extracted. It is not for drafting copy, explaining something, or searching the web, that is what the writing model and search tools are for.
Send every independent question about the same state together in one call, they run in parallel and cost almost nothing extra. Read the returned probability and confidence before acting: a `noul` sitting near 0.5 means Jev does not know, and the magik should say so or ask rather than guess. A worked example lives in the 'verify a citation' demo magik: paste a source text and a claim, get back supported, not supported, or uncertain, with the probability behind it.
Mojo runs your recurring Typesafe tasks on a schedule so you never touch them.
Collapses everything new in Typesafe into a short list you can act on.
Turns a Typesafe run into content written in your saved brand voice.
Ready to plug Typesafe into Mojo?
One-click connect from inside Taquet. Your data stays in your account; Mojo only acts when you ask.
Open Taquet with Typesafe selected