Where does Jev fit into the bird brains?

Nowhere documented in the excerpt. Brains are feedforward nets evolved by selection, crossover, and Gaussian mutation with sigmoid firing over 0.5, not a Jev decision call.

Can human reflexes beat early generations?

That is the Human vs AI mode premise: press Space or click to flap against Generation 1, while Generation 10-plus evolves higher flight efficiency.

Does this need an API key, or is everything in-browser evolution?

No key: nothing in the excerpt calls an API, and the flocks are feedforward nets shaped by selection, uniform crossover, and Gaussian mutation with the sigmoid output firing above 0.5. It is Vite and TypeScript, so Node v18 or newer, tested on v24, with npm v9-plus, then npm install and npm run dev on port 5173; the lab tunes population, mutation, elitism, crossover, wind, and speed from 1x to 32x.

How do I keep a good pilot?

Export the champion weights as JSON in the DNA vault, re-import brains later, or inject the Acrobat Alpha preset to demo high-scoring flight.

What does the simulation actually evolve?

Three feedforward species compete: Scarlet on a 4-4-1 network, Ash on 4-8-1, and Rust on 4-16-1. Inputs are the bird's altitude, horizontal distance to the pipe, gap top, and gap bottom; the sigmoid output neuron above 0.5 means FLAP. Selection, uniform crossover, and Gaussian mutation shape the weights, with no datasets and no backpropagation anywhere in the loop.

Can I tune the evolution while it runs?

Yes, through the Hyperparameter Lab. It adjusts population size per species from 10 to 100 birds, base mutation rate from 5% to 80% with its variance, the elitism count that protects champions across generations, crossover probability, and the wind and turbulence physics. Simulation speed is separately adjustable from 1x to 32x with the arrow keys.

What license is it under, and what is the stack?

Vite 6 with TypeScript 5 and HTML5 Canvas 2D for rendering, no backend and no framework. The simulation lives in the browser and needs Node 18 or newer, tested on Node v24, with npm 9 or newer, then npm install and npm run dev. Audio is synthesized at runtime through the Web Audio API, so no audio files are downloaded.

CommunityJev Demos & Experiments

TypeSafe Mario AI

Browser neuroevolution game where flocks evolve to fly through towers with no datasets or backprop, plus human-vs-AI mode and DNA vault.

See highlights

Open github.com

Stars: 0
Language: TypeScript
Published by: Community
Author: GunaTeja777
Last verified: Oct 2026

GunaTeja777/typesafe-mario-ai

User repository on GitHub

View on GitHub

Real-time Super Mario game autonomously controlled by TypeSafe AI's "Jev" System One Decision Architecture. Features sub-100ms decision loops, live prompt & probability telemetry, multi-provider support (OpenRouter Jev & Groq LPUs), and an offline neural simulator.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Challenge the evolving AI flock directly! Press Spacebar or click the canvas to flap as the Human Challenger. Test whether your human reflexes can beat Generation 1, and watch as Generation 10+ evolves superhuman flight efficiency. Inspect live brain activity for each competing species: Input layer: Bird altitude ($y$), horizontal distance to pipe ($x$), gap top, and gap bottom. Hidden neurons with real-time activation values and bias rings.

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

Tagscommunity github

Works forJev for Routing

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Three species with 4-4-1, 4-8-1, and 4-16-1 nets; inputs are altitude, pipe distance, gap top and bottom.
Live visualizers show activations, color-coded weights, pulses, and FLAP versus WAIT gauge with procedural Web Audio.
Hyperparameter lab tunes population, mutation, elitism, crossover, wind, and speed from 1x to 32x.
Exports champion weights as JSON, imports brains, or injects Acrobat Alpha preset; runs on Vite with npm run dev.

Watch out

README excerpt describes the bird simulation and controls without documenting a Jev call; verify Jev wiring before listing as Jev-powered. 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

jevry

123

Electron desktop browser where Jev picks the next executable action from a page's observed controls while a separate text model plans; deterministic code builds the action space, validates responses, dispatches input and keeps receipts.

Jev Demos & Experiments#community#jevbest

Community

michaelswissa/jevryTypeScript

playwright-recast

Turns a Playwright trace into a narrated demo video through a chainable pipeline - speed control, subtitles, TTS, click and cursor overlays, zoom - with ffmpeg doing the render.

Jev Demos & Experiments#community#jevbest

Community

ThePatriczek/playwright-recastTypeScript

Pi-Jev-Guide

Pi guard plugin that checks inputs and tool calls with local rules plus Jev semantic judgements, with redaction and verification reminders.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

zszz3/Pi-Jev-GuideTypeScript

JevTown

Pixel-art simulation engine where agents pick structured next actions through Jev while a separate language model handles dialogue, memory, and beliefs, all running in one browser tab.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

NevaMind-AI/JevTownTypeScript

flue-jev-demo

Standalone Flue example that routes support chat through Cloudflare Workers AI, using gpt-4o-mini for chat and Jev for typed intent and urgency routing.

Jev Demos & Experiments#community#awesomejev#catalog-import

Community

matthewp/flue-jev-demoTypeScript

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.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

joevidev/ui-generator-instinct-jevTypeScript

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai