Train a game bot and ship it as a Chrome extension with WebAssembly
A developer built a Pile-Up Poker (Puzzmo) bot scoring about 6,900 points a game. He tried 6 agent families, picked expectimax plus a value net, made it 1,450x cheaper with beam search and distillation and shipped it as a Chrome extension running in WASM.
Source
@therahuldapp on X · 2026-10-17
Written in our own words. Credit to the original author. Link checked 2026-10-22.
What they did
A developer built a Pile-Up Poker (Puzzmo) bot scoring about 6,900 points a game. He tried 6 agent families, picked expectimax plus a value net, made it 1,450x cheaper with beam search and distillation and shipped it as a Chrome extension running in WASM.
Result (as claimed)
Runs in your browser at 0.16 s per game; play against it and get a replay.
Tools they used
Cost: Pricing depends on the tools you pick; check each official pricing page.
Set it up
Gather your sources
Collect the papers, notes, links or files you want to work from in one place.
Load them into the tool
Open JAX, Python, WebAssembly, Chrome and add your sources. Keep the original files so you can check claims later.
Ask focused questions
Ask for summaries, comparisons or connections with citations to the source passages.
Verify and save
Spot check claims against the sources, then save the output as notes you can reuse.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Train a game bot and ship it as a Chrome extension with WebAssembly What the original author did: A developer built a Pile-Up Poker (Puzzmo) bot scoring about 6,900 points a game. He tried 6 agent families, picked expectimax plus a value net, made it 1,450x cheaper with beam search and distillation and shipped it as a Chrome extension running in WASM. Reported result (as claimed by the author): Runs in your browser at 0.16 s per game; play against it and get a replay. Source: https://x.com/therahuldapp/status/2111539795067277391 Tools they used: JAX, Python, WebAssembly, Chrome Category: Research & learning Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Train a game bot and ship it as a Chrome extension with WebAssembly What the original author did: A developer built a Pile-Up Poker (Puzzmo) bot scoring about 6,900 points a game. He tried 6 agent families, picked expectimax plus a value net, made it 1,450x cheaper with beam search and distillation and shipped it as a Chrome extension running in WASM. Reported result (as claimed by the author): Runs in your browser at 0.16 s per game; play against it and get a replay. Source: https://x.com/therahuldapp/status/2111539795067277391 Tools they used: JAX, Python, WebAssembly, Chrome Category: Research & learning Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Train a game bot and ship it as a Chrome extension with WebAssembly What the original author did: A developer built a Pile-Up Poker (Puzzmo) bot scoring about 6,900 points a game. He tried 6 agent families, picked expectimax plus a value net, made it 1,450x cheaper with beam search and distillation and shipped it as a Chrome extension running in WASM. Reported result (as claimed by the author): Runs in your browser at 0.16 s per game; play against it and get a replay. Source: https://x.com/therahuldapp/status/2111539795067277391 Tools they used: JAX, Python, WebAssembly, Chrome Category: Research & learning Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Guides for this use case
How to build a Chrome extension with AI · How to cut AI coding agent costs and token use · How to vibe code a game with AI
Related use cases
Have Claude Code email you files from your PC while you are on your phone
Away from his desk, a user asked Claude Code to mail him two videos sitting on his PC. Its Chrome extension wasn't connected and Codex's computer-use tool kept failing, so Opus 5.5 wrote its own small computer-use script in Python and drove the desktop to send them.
Make LLM agents team up in Pokemon Showdown battles
A researcher built a 2v2 Pokemon Showdown environment where two LLM agents must coordinate through team chat. You can watch battles with the chat shown, and the code is on GitHub.
Turn a stock watchlist into a 12-agent market research lab with Claude Code
Inspired by Boris Cherny's parallel-agent setup, a developer built a system with scouts, parsers, linkers and reviewers over 1,284 news sources. Reviewers test every signal on 3 years of data, against news changes and against doing nothing.
Have a Grok Bot apply to jobs with a tailored resume for each
A job seeker asked his Grok Bot to apply to 5 jobs and tailor his resume to each description. It handled captchas on its own and only asked for help, by showing him a browser, when a Google login was needed.