TESIGN / RADAR
2026.09.22
pokertools-arena

pokertools-arena

Watch AI models play poker at one table

WHO USES ITDevelopers comparing real-world judgement across models, and people building AI benchmarks.

Use in the browser

Seats several OpenAI-compatible models at the same poker table and lets you watch every decision as a spectator.

TRY

START IN 5 MINUTES ↓

The pokertools-arena screen: current decision and history on the left, and a green poker table with ten empty seats on the right.

TESIGN TAKE

Using an adversarial game instead of a question set as the benchmark is the different idea here, and being able to watch every decision live is a real part of that.

Activity and discovery evidence

★ 1▲ 2 HNAs of 09-22 03:00 UTC

1 stars · checked on GitHub 09-21 16:07 UTC · no 7-day star increase observed (GH Archive) (as of 09-22 03:00 UTC) · 2 Show HN points observed 09-21 15:09 UTC

NEW 0h OLD WHEN SEEN

AT A GLANCE

LICENSE: MIT
USAGE: Use, change and redistribute, commercially too. Keep the notice.
LANGUAGE: JavaScript
PLATFORM: web
ACTIVITY: last commit within 24 h (2026-09-21) · latest release [unconfirmed] As of 09-21 16:07 UTC
COMMUNITY: contributors [unconfirmed] · open issues [unconfirmed] · made by: [unconfirmed] Reference time unavailable
SOURCES: Show HN
OPEN SOURCE: YES
FIRST SEEN: 2026-09-21 15:07 UTC
CATEGORY: AI · GAMES

A summary, not legal advice.

WHY IT MATTERS

Most language-model benchmarks are a fixed question set, which makes it hard to tell memorised answers from real judgement. Poker has an opponent, hidden information and a different situation every hand, so bad reasoning shows up immediately. pokertools-arena turns that structure into a benchmark models play against each other.

BUILD FROM THIS

Poker legality, hole-card masking and legal actions all come from the PokerTools engine, and every move a model chooses is re-validated by that engine before it counts. It is fully client-side with no server or database — run it with npx pokertools-arena or build it into a single HTML file — and it exports decision logs and replay images.

WHO IT'S FOR

Developers comparing real-world judgement across models, and people building AI benchmarks.

START IN 5 MINUTES

# Run npx pokertools-arena to start a local server and seat a model.

CAVEATS

It needs Node.js 24 or later and an OpenAI-compatible API, and calling real models means paying that provider’s usual rate. It is still very early — one star.

RECEIPT

FIRST SEEN: 2026-09-21 15:07 UTC
AT SOURCE: 2026-09-21 15:00 UTC
KEPT: 2026-09-22 05:53 UTC
CREATED → FIRST SEEN: 0h
WHEN FIRST SEEN: 1 ★
SOURCES: Show HN

The same facts in machine-readable form — View as Markdown · JSON

SIMILAR TOOLS

Up to five from AI by ★ total: edited entries first, then repository cards; this entry is highlighted. Drawn from the same stored snapshot as the rankings — a comparison, not a recommendation. As of 09-22 03:00 UTC

NAME	★ TOTAL	▲ 7d	LICENSE	PLATFORM	LAST PUSH
eccSkills, memory and security for agents	★ 265k	▲ +413	MITpermissive	windows · macos · linux · cli	2026-09-21
Hermes AgentSelf-improving personal agent	★ 248k	▲ +108	MITpermissive	linux · macos · windows · android · cli	2026-09-22
DeepSeek HarnessPlugin agent harness for developers	★ 232k	▲ +487	MITpermissive	web · cli	2026-09-17
pokertools-arenaWatch AI models play poker at one tableTHIS ENTRY	★ 1	—	MITpermissive	web	2026-09-21
mattpocock/skillsCARD	★ 261k	▲ +317	MITpermissive	—	2026-09-04

TIMELINE

2026-09-18 23:21 UTCRepository created
2026-09-21 15:00 UTCShow HN post ↗
2026-09-21 15:07 UTCFIRST SEEN BY TESIGN
2026-09-22 05:53 UTCPublished on TESIGN

SAME DAY

01
ZCode
Run multiple coding agents from one app
A coding workspace that runs several AI agents at once and drives a goal to completion.
Use in the browser
★ 5.1k▲ +64 / 7d09-22 03:00 UTC
02
Try Omarchy for Windows
Try the Omarchy Linux desktop on Windows
An app that boots the real Omarchy Linux desktop inside a window on Windows — no VM software to install.
Install to use
★ 440▲ 5 HN09-22 03:00 UTC
03
hibi
A markdown notes app with a graph view
A desktop Markdown editor that also draws a graph of how your notes connect.
Install to use
★ 14▲ 5 HN09-22 03:00 UTC
04
tfm
Drag and drop files inside your terminal
A terminal file manager built for the mouse — drag, drop, dual panes, thumbnails.
Install to use
★ 136▲ +9 / 7d09-22 03:00 UTC
05
AgentMeasure
Audit your Codex and Claude Code logs
Reads your Codex and Claude Code session logs and surfaces repeated failures and retries, with the evidence attached.
Install to use
★ 216▲ 2 HN09-22 03:00 UTC
06
package-doctor
Find the Python dependency to fix first
Scans Python dependencies and tells you which are being exploited now — and which have no one left to fix them.
Install to use
★ 8▲ 4 HN09-22 03:00 UTC
08
Roastbook
Log espresso shots on your own server
A self-hosted coffee journal that reads a photographed bag’s label and fills in the details for you.
Self-host
★ 2▲ 1 HN09-22 03:00 UTC
09
Comic Chat for AI
Turn a ChatGPT chat into a comic strip
A revival of Microsoft’s old Comic Chat that automatically draws a ChatGPT conversation as a comic strip.
Install to use
★ 1▲ 3 HN09-22 03:00 UTC
10
2049 — Mines of Merge
2048 fused with Minesweeper in a dungeon
You are the tile: slide by 2048 rules while dodging Minesweeper-style hidden mines down eight depths.
Use in the browser
★ 0▲ 1 HN09-22 03:00 UTC