usecases.sh
usecases.sh
Software & DevOps

Build and launch a browser game in 72 hours with a team of Grok Bots

Three xAI Grok Bot engineers shipped a game, Thursday Arena, from an empty repo in a 72-hour livestream: an implementer-verifier bot pipeline landed 433 PRs and launch day saw about 4,000 games.

Apps touched

What they did

Roshan Sadanani, Lauren Tan and Matt Palmer of the xAI Grok Bot team streamed a three-day build next to Dreamforce in San Francisco in September 2026, starting from nothing and pivoting from a pop-up restaurant idea to a game studio. Each Grok Bot had its own cloud computer and a narrow role: a chief-of-staff bot, implementer bots writing small pull requests, verifier bots that ran the game, clicked through it and fuzzed for bugs, and a playtester bot watching for green CI. Lauren's PStack skills drove the pipeline, with a swarm skill spawning Cursor cloud agents for QA. Merges were gated on reproduced bugs and attached proof. GitHub user unicodef1wn watched the stream and distilled the three days into rules, nine role playbooks, a failure log and an economics sheet.

How it went

The game, Thursday Arena, launched from a single X post at about 10am Pacific on day three. By end of day it had roughly 4,000 games, 6,000 public matches, 30,000 page views and 2,000 users, with about 8 percent of practice players signing in via X. Feedback was 71 percent bug reports and 16 percent praise. Revenue was zero; the ad auction was broken. The bot pipeline landed more than 100 PRs overnight and 433 in total, with 157 commits from one host. The failure log lists forty things that broke on air.

Worth knowing

This is a vendor team demonstrating its own product on stream, written up second-hand by a viewer from auto-captioned transcripts, so names and figures should be read as order-of-magnitude. The cost lessons are the transferable part: routines every 15 minutes and bots in group chats are where tokens go, browser-driven form filling costs more than an API path, and the team kept a human gate on deploys, migrations, money and permissions throughout.

Try it withGrok Bot

We are building [the product, e.g. a small multiplayer browser game] in [repo] over the next [three days]. Set up a team of bots with one narrow job each: a chief of staff I talk to, implementer bots that take one task each and open small pull requests, and verifier bots that run the app, reproduce the bug or feature, click through it and attach screenshots as proof. Break [the plan] into phases, run the implementers in parallel, and let a PR merge only when a verifier has reproduced the issue first and the fix passes CI. Run the check routine no more than every [hour]. Never deploy, migrate the database, spend money or change permissions without pinging me for approval. Report at the end of each phase with what merged, what broke and what is blocked.

Square brackets are the parts to fill in. The prompt is ours; the story is theirs.

Run this with your own accounts

Connect GitHub once on One and any agent you run can use it, with the scopes you choose.

Connect on One

Where they said it

Related use cases

All 6 Grok Bot use cases
All Software & DevOps use cases

Every entry is something a person reported doing with an agent, rewritten in our words, with a link to where they said it. Found across the web every hour; reviewed before it ships.

Maintained by an agent that uses One for every integration it touches.