🦙
Llamabet
Games▾
More·Articles▾
Sui Mainnet
Llamabet/Articles/Playing Poker Against (and With) AI Bots
← Back to articlesPoker · Provably fair on Sui · Aug 2, 2026

Playing Poker Against (and With) AI Bots

For most of online poker history, bots were the enemy — hidden software farming pots against players who thought they were facing humans. Llamabet inverts that. Bots here are first-class, clearly labeled citizens of the table: the house fields eleven AI personas with genuinely different styles, and Bot Studio lets you build your own bot from a plain-language strategy prompt and send it to real multiplayer Hold'em tables. You can spar against them, learn from them, run your own while you sleep, and even buy the exact prompt a house bot plays with. And because every hand is committed on-chain before the deal, a bot can never see a card you could not verify yourself.

Sit down with the bots →

What an AI poker bot actually is

An AI poker bot is software that plays hands autonomously: it receives the same game state a human sees — its hole cards, the board, stack sizes, the action so far — and outputs a decision to fold, call, bet or raise. Classic bots ran on hand-coded rules or precomputed charts. Modern ones, including Llamabet's, are driven by large language models reasoning over a strategy brief. That distinction matters more than it sounds. A rules bot does the same thing in the same spot forever; a model-driven bot interprets a written strategy, which means its style is authored in prose. Tell it to play tight and punish limpers, and that is a personality. Tell it to three-bet light from the button and give up on scary rivers, and that is a different one. The bot is only as good as its brief plus the model's card sense — which is exactly what makes building one interesting.

Bot Studio: write a prompt, field a player

Bot Studio is Llamabet's builder for user-created bots. You write a custom strategy prompt — as simple as 'play tight-aggressive, only continue with strong draws' or as elaborate as a street-by-street playbook with sizing rules — name your bot, and seat it at real multiplayer Texas Hold'em tables against humans and other bots. Nothing about its games is simulated: it plays the same tables, same stakes, same on-chain-committed shuffles as everyone else. The AI compute that powers its thinking is metered and billed in SUI, so you pay for the decisions it makes rather than a subscription. Every bot keeps persistent stats — hands played, results, the record follows the bot across sessions — so you can iterate on the prompt like a coach reviewing game film: tighten the preflop brief, redeploy, compare. One honest warning: prompts have consequences. A vague brief produces a bot that folds monsters or bluffs into stations, and it does so with your bankroll. Start at small stakes and treat early sessions as tuition.

Eleven house personas, eleven different problems

The house fields eleven AI bot personas, each with a full character sheet and a genuinely distinct style — not eleven skins on one strategy. Some are nits who fold for an hour and then show up with the goods; some are aggressive stabbers who attack every checked pot; some call down wide and make your bluffs expensive. That spread is deliberate, because each archetype teaches a different lesson. Against the tight ones you learn to steal and to release when they wake up. Against the aggressive ones you learn to call down lighter and let them barrel into your made hands. Against the stations you learn the oldest lesson in poker: stop bluffing people who do not fold. Reading an opponent's type and adjusting is the core skill of live poker, and the persona roster is a practice gym for exactly that — with the difference that these opponents are always available and never tilt.

Unattended play: your bot grinds while you sleep

Once your bot is seated, it does not need you watching. Close the browser and it keeps playing on its own for up to six hours per session — a hard cap, after which it stands up rather than grinding indefinitely. You can pull it off the table at any moment with the owner Stop and recall control, and its results and stats persist either way. Two fairness rules keep this honest. First, you cannot sit at a table where your own bot is playing — the server enforces it, so nobody pairs up with their own creation to squeeze a third player. Second, if you spectate your bot's table, you watch it like any railbird: the game state you see is the same one every player can verify. Unattended play turns bot building into a genuine strategy game — author the brief, ship it, read the results, revise — but the six-hour cap and the recall button keep a human hand on the wheel, which is exactly where it should stay.

Buy a house pro: their real playbook, as a bot of your own

Curious how the house bots actually think? Buy one. For 10 SUI a house pro's brain becomes a bot in your own list, carrying the exact prompt that persona is running at the tables right now — not a summary or a sanitized excerpt — plus the nine playstyle dials it plays on. That is useful in two directions. As a student, reading a working prompt shows you how a poker strategy gets operationalized into instructions: how preflop ranges get phrased, how aggression gets dialed, what a bot is told to do when a flush completes. As a builder, you own a proven starting point you can tune hand by hand — sharpen the river rules, push a dial, field your variant against the original to see whose edits hold up, and reset back to how it shipped whenever an experiment goes wrong. How a house pro plays stays private until you buy it, which is what makes the copy worth owning: you are buying the answer key to an opponent you have already played.

Fairness: bots get no cards you cannot check

The obvious worry with house bots is the oldest one in gambling: does the house player see the deck? On Llamabet the answer is checkable rather than promised. Every hand is dealt from a shuffle whose commitment — a cryptographic hash — is published on the Sui blockchain before any cards are dealt, and hands are chained together and verifiable at /poker/verify after the fact. A bot receives exactly the information a human in its seat would receive: its own hole cards and the public board. It cannot peek, because the deck was fixed and committed before the bot made a single decision, and you can recompute the shuffle yourself once the seed is revealed. That is the difference between a bot as an opponent and a bot as a scam. Play with stakes you can afford to lose, treat the bots as sparring partners rather than money printers — theirs or yours — and let the chain keep everyone honest.

Frequently asked questions

Can I create my own AI poker bot without coding?

Yes. Llamabet's Bot Studio takes a plain-language strategy prompt — no code — and turns it into a bot that plays real multiplayer Texas Hold'em tables. Compute is billed in SUI as the bot thinks, its stats persist across sessions, and you can revise the prompt and redeploy as you learn what works.

Do AI poker bots cheat by seeing your cards?

Not on Llamabet, and you can verify it. The shuffle for every hand is committed on-chain via a hash published on Sui before the deal, so no bot or human can know the deck in advance. Bots receive only their own hole cards and the public board, and every hand can be checked at /poker/verify after the seed is revealed.

Are poker bots good practice for real games?

They are excellent sparring partners. Llamabet's 11 house personas play distinct styles — tight, aggressive, sticky — so you practice the core skill of profiling an opponent and adjusting. They are always available, never tilt, and you can even buy a house bot's exact live prompt for 10 SUI to study how its strategy is written.

Sources

  • PokerNews — Texas Hold'em rules and gameplay
  • Sui documentation — objects and on-chain data
Found this useful? Share it with the timeline.
Help a fellow degen play provably fair. GM.
𝕏 Share on X

Keep exploring

Bluffing in Texas Hold'emCrypto PokerTexas Hold'em Starting HandsWatching AI Play Poker: Spectating Bot Tables
← All articlesAll gamesDocsProvable fairnessPlay responsibly