Does it ever emit code or copy from the model?

No. Jev only returns typed picks over bounded option sets while the app renders real shadcn primitives; table and card words come from small curated banks, with deterministic row variation rather than per-row model calls.

How many API calls does one description cost?

Usually two: one speculative fan-out plus top-2 family refinement, with extra calls only for thin-margin tiebreaks, variable layouts like cards and charts, or bounded text fields needing content fills.

What breaks or looks wrong most often?

Overlapping candidates can produce near-duplicate columns since ranking caps rather than merges them, the English word banks are small, and generic catalog entries are typechecked and code-reviewed rather than all visually checked.

Who pays when I deploy the demo publicly?

Your server key covers the first 5 successful generations per visitor, then the UI asks for the visitor's own console key stored in their browser and billed to them; failed calls consume no quota, and the cookie counter is a cost cap rather than a security boundary.

Where does Instinct get the component names for its Code view?

lib/element-to-jsx.ts walks the real React element tree the renderer produces and serializes real tag names, props and nested elements, resolving imports from the exact bindings the canvas uses, so recharts' Tooltip comes out under its imported name.

How does Instinct fill table cells and card stats?

For fields Jev types as free-form text, lib/generated-content.ts asks it to pick a category and then the word or words from small curated word banks by Choice, never free text; row 0 is what Jev picked and further rows vary deterministically.

What does the free generation counter on Instinct actually count?

Only successful decisions on the shared key; the first 5 generations per visitor come out of the deployer's key, after which the UI asks for the visitor's own key, whose calls are not metered, and failed calls do not consume quota.

CommunityJev Demos & Experiments

UI Generator Instinct Jev

Demo that turns free-text UI descriptions into real shadcn components using only Jev typed Choice, Noul, and Score answers, with no generated code on the decision path.

See highlights

Open github.com

Stars: 7
Language: TypeScript
Published by: Community
Author: joevidev
Last verified: Oct 2026

joevidev/ui-generator-instinct-jev

User repository on GitHub

View on GitHub

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

Describe it. Jev decides. A demo of Jev — TypeSafe's "System One" model — as a UI generator. You describe a case in free text; Jev never generates a line of code or copy back. It only ever answers typed questions (Choice, Noul, Score) over bounded, real option sets, and the app renders whichever real shadcn/ui component (or full page block) it picked, configured the way it decided. Content artifact for a Buivo post about System One models. Working name, pending final brand sign-off.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Routes across 48 real shadcn components in 7 families plus 17 page blocks sourced from installed shadcn output.
Resolves cases hierarchically with speculative fan-out, top-2 beam refinement, and conditional tiebreak calls.
Fills text fields by picking categories and words from curated banks, never free text, capped per relevant field.
Serializes the actual rendered React tree into copy-pasteable JSX with correct imports instead of templates.

Watch out

Not every catalog entry is visually checked, near-duplicate fields can slip through, word banks are small English sets, and an unused Claude flavor route ships in the tree. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-browser

TypeScript browser harness where an OpenRouter planner splits goals into subgoals, Jev picks each concrete action, OpenRouter fills text fields, and Playwright drives the page.