Skip to content

Latest commit

 

History

6 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MA Showcase

Real, reproducible demos showing MA turn a prompt into a verified result. MA is a coding agent that works in a repository, runs what it produced, and keeps the evidence.

MA creates and verifies a real artifact

Each demo keeps four things together:

  1. The prompt and task boundary.
  2. The exact command used to run MA.
  3. The generated artifact and a direct verification step.
  4. A terminal recording made from that same command.

This repository does not use synthetic agent output or staged completion claims. A demo is only listed after its artifact exists and the recorded command has completed.

Demos

Demo What it proves Artifact
01 Terminal Status MA creates and verifies a standalone UI artifact through its filesystem tool. artifact/index.html
02 Reaction Time MA builds a browser interaction, then a real browser test reaches its result state. artifact/index.html
03 Versioned MemoryPatch MA drives a local synthetic-rule memory through metadata-verified mount, a new version, and rollback. EVIDENCE.json

02 Reaction Time

A real browser run of MA-generated Reaction Time

The artifact, browser recording, and verifier are committed together. The GIF shows a Playwright-run browser opening the generated game, waiting for its live "go" state, clicking the panel, and reaching the result screen.

Shareable Evidence

Vertical short video packages the actual tool-call evidence and browser result into an 11-second 1080x1920 clip. It is ready for YouTube Shorts and Douyin; posting copy and the reproducible render script are included.

Versioned MemoryPatch evidence adds a bounded local-memory proof: two synthetic rules became two auditable patch versions, then the active profile rolled back to the first version. Its vertical short, images, raw evidence, and rerun script are committed together. It deliberately does not claim unlimited context or general cloud-model parity.

Run Locally

The recorded runs use MA with an Agora local setup. That runtime detail is only needed when reproducing this exact environment; the user-facing proof remains the same: MA created a real artifact and the browser verified it.

AGORA_ROOT=/path/to/agora bash demos/01-terminal-status/run.sh

# For the browser interaction demo
AGORA_ROOT=/path/to/agora bash demos/02-reaction-time/run.sh
npm ci
npx playwright install chromium
npm run verify:reaction

# Verify the recorded MemoryPatch evidence and render its social assets
npm run verify:memorypatch
npm run render:memorypatch

The replay scripts exit successfully only after the generated file is present and non-empty. The reaction-time verifier also proves that the artifact reaches a result screen after a real click sequence.

Status

This is an early showcase repository. Every recording is a real local run; model speed and output can vary by hardware and model version. The initial reaction-time run took roughly two minutes before its file-tool action; it is evidence of a working artifact, not a claim that every local machine will finish at that speed.

About

Real MA + Agora coding-agent demos, recordings, and reproducible artifacts.

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages