Sequoia, OpenAI and Citadel Securities launch No Limit, a one-hour pokerbot tournament for students

Venture firm Sequoia Capital announced No Limit on Tuesday, Oct 27. It's a bot-building tournament where teams of college students and recent grads get 60 minutes to code a poker bot from scratch, and then the bots play thousands of simulated tournaments against each other. OpenAI and Citadel Securities are the hosts, and the top team at the Finals in San Francisco gets $100,000, according to the official rules.

It was the most-talked-about poker post on X yesterday. Sequoia's announcement passed 400,000 views within a day. Noam Brown, the OpenAI researcher who co-created the Pluribus poker AI, wrote that coding agents have sped up bot-building so much "that it's now possible to build a poker bot in an *hour*," and said he'll be at the Finals.

Dates and deadlines

Each regional runs from 6:00 to 9:00 PM local time, and you have to be there in person for the whole event. Applications opened Oct 27 and are reviewed on a rolling basis, with limited seats at each regional.

Who can enter

• Current students at U.S. universities, and graduates of U.S. universities from the class of 2025 or 2026.
• Everyone has to be 18 or older.
• Teams are one or two people, and teammates don't have to go to the same school. Each person can compete in only one regional.
• It's free to enter. The rules say participants put nothing toward the prize pool.

How the contest works

• The poker variant stays secret until the start of the build phase on event day, and the Finals use a different variant. The demos on the official site show simulated no-limit hold'em, but the rules don't promise that's what teams will get.
• Teams get 60 minutes to build a Python bot from a starter kit with an agent template, a local simulator, an Independent Chip Model (ICM) calculator, the payout structure and the blind schedule. The rules allow teams to use "Astra and Codex" to write and debug their bots.
• Bots run in sandboxes with no network access, so they can't call an AI tool or any outside service during play. Each bot has one millisecond to make a decision. A bot that errors or runs out of time folds, or checks if it can.
• Bots play only against other bots. Chips and points exist only in the simulation, no money is bet by the bots, and nobody at the event may bet on the results.

Scoring

There are six scored tournaments, T1 to T6, starting 20 minutes apart after the build phase. Teams can keep improving their bot between them. Each scored tournament is 10,000 simulated eight-bot sit-and-gos. Every bot starts with 2,000 chips, blinds start at 10/20 and go up every 25 hands, and a table stops when one bot has all the chips or after 200 hands. If the hand cap hits first, the remaining points go to the surviving bots by ICM.

Later tournaments count for more. A team's final score is T1 plus 2 times T2, 3 times T3 and so on up to 6 times T6. That means a bot that gets better over the evening can beat one that started strong and stood still.

Prizes

• Each regional: the top 5 teams get $5,000 per team and advance to the Finals. Up to five more wildcard teams can be picked based on their scores.
• Finals: $100,000 for first, $50,000 for second and $25,000 for third.
• Prizes are per team and split equally between teammates. Travel to the Finals is covered for teams that advance.

Where pokerbots come from

Bot-vs-bot poker contests aren't new to students. MIT Pokerbots (course 6.9630) runs every January, and teams spend a month building bots for a variant they don't know in advance. Brown's point is that coding agents now make a working bot possible in an hour, and No Limit puts that to the test. Brown is also one of the two authors, with Tuomas Sandholm, of the 2019 paper on Pluribus, an AI they showed was stronger than top human pros at six-player no-limit hold'em, published in Science.

For players, the takeaway is simple: this is a coding contest, not a poker game you can sit in. Questions about the event go to pokerbot@sequoiacap.com, per the official site.