How is this different from holding ⌘Tab and picking myself?

Holding ⌘Tab shows a static recency list you scan yourself. Here Jev reads your recent switches, dwell times, time of day, and last manual ⌘Tab picks, then activates its top choice in one keypress.

What happens when the model is slow or picks wrong?

Any model failure or a 1 s timeout falls back to the last-used app. Turning on Developer mode in Settings notifies you whenever that fallback path is taken.

What does setup actually require on a fresh Mac?

Run ./build.sh to build the Rust core and Swift app, sign, install to /Applications, and launch. On first launch grant Accessibility, then paste a TypeSafe key from Settings into the login keychain.

Can I inspect exactly what was sent to the model?

Yes. The menu-bar menu shows the last decision with Last Request and Last Response submenus; hovering reveals the exact text sent or received, and clicking a row copies it.

What data goes into the choice?

On ⌘`, the 10 most recently used running apps become options in one Jev choice question, and the request includes recent switches, how long you stayed in each app, the time of day and your last 10 manual ⌘Tab switches.

Can I test the core without the app?

Yes. cargo test runs the Rust core tests, and running TYPESAFE_API_KEY=... cargo run --example predict from core/src/lib.rs hits the live API directly.

Is tab switching on the tilde key supported?

Not yet. The README lists tab switching on ⌃` as planned but not implemented, so only the ⌘` window prediction works today.

CommunityJev Tools & CLIs

Smart Switch

macOS window switcher that predicts the app you want on ⌘` from recent switching history, asking Jev to pick from the 10 most recently used running apps with a 1 s fallback.

See highlights

Open github.com

Stars: 3
Language: Swift
License: AGPL-3.0
Published by: Community
Author: reycn
Last verified: Oct 2026

reycn/smart-switch

User repository on GitHub

View on GitHub

Reimagined window switcher for macOS using frontier artificial intelligence. Predicted by TypeSafe's Jev model

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Reimagined window switcher for macOS using frontier artificial intelligence. It just knows which window you wanna use. Predicted by TypeSafe's Jev model from your recent switching history, in realtime. Tab switching on ⌃` is planned, not implemented yet. Every app activation is recorded (last 200), tagged as a ⌘Tab switch, a SmartSwitch switch, or other.

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

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Swift

Highlights

Records the last 200 app activations with switch source, dwell time, time of day, and recent ⌘Tab history.
Builds one Jev choice question per ⌘` press; failure or 1 s timeout activates the last-used app.
Rust core with a 4-function C ABI plus a SwiftUI menu-bar app; ./build.sh builds, signs, and installs it.
First launch walks through the Accessibility grant and a TypeSafe key kept in the login keychain.

Watch out

Tab switching on ⌃` is planned but not implemented, and prediction needs a TypeSafe key with Accessibility access granted. 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-recall

Python memory filter that judges every stored memory against the current request with one calibrated yes/no per memory in a single Jev request, keeping everything above a threshold.