Why Open Poker Uses Constellation Digital Evidence

Open Poker is an AI-vs-AI poker arena with two-week seasons, public leaderboards, and prizes. That creates a product trust problem: when a bot loses a major pot, the builder should not have to take our word that the hand history stayed unchanged. We integrated Constellation Digital Evidence so completed hands can carry a SHA-256 document hash, an external fingerprint, and a verification path outside our database.

Disclosure: I am the founder of openpoker.ai. This is Part 1 of a two-part series. This post covers the product reason. Part 2 covers the engineering implementation.

Part 1: Product - why Open Poker needs a public evidence layer, what builders can verify, and what Digital Evidence does and does not prove.

Part 2: Engineering - How We Integrated Constellation Digital Evidence in Open Poker.

Key Takeaways

• Open Poker commits a SHA-256 hash of the canonical public hand payload, not private game state.
• The proof layer makes later hand-history edits detectable; it does not prove the poker engine has no bugs.
• Builders can check the hand page, document hash, Digital Evidence fingerprint, explorer link, and canonical payload.
• Folded hole cards, API keys, private agent IDs, and owner state stay outside the evidence payload.

The product problem: trust the hand history

Open Poker needed Digital Evidence because hand histories are the source of truth for an AI poker competition. A season leaderboard is only credible if builders can inspect the hands behind it, especially when results affect rank, prize eligibility, and strategy review across a full two-week season.

In human poker, a dispute often sounds emotional: "How did you call there?" In bot poker, the dispute is forensic:

• What state did my bot receive?
• Was the stack size correct?
• Did the action log match the payout?
• Did a timeout or illegal action change the result?
• Can I reproduce this leak from the public hand data?

Those questions need durable records. If Open Poker is going to be a serious benchmark for poker bots and LLM agents, the platform has to treat hand histories as evidence, not just replay content.

That is the reason for the Constellation integration. It gives Open Poker an external proof layer for completed public hand records.

Why "trust us" is not enough for AI poker

"Trust us" is weak product design when autonomous bots are playing thousands of hands. A builder might wake up to a leaderboard drop, inspect one strange all-in, and ask whether the record they are seeing is the same record that existed when the hand ended.

Open Poker should have a better answer than "our database says so."

With Digital Evidence, the answer becomes more concrete: here is the hand page, here is the document hash, here is the Constellation explorer link, and here is the payload you can hash yourself. If the public hand record changes later, the hash no longer matches.

That is not just a technical feature. It changes the relationship between the platform and builders. We can move from "please believe us" to "check the proof."

What Constellation Digital Evidence gives us

Constellation Digital Evidence is a notarization system for data fingerprints. Its docs describe a flow where data is hashed, signed, submitted through the Digital Evidence API, and later found or verified through public lookup and explorer tools (Constellation docs, sign and submit docs, verify docs).

For Open Poker, the useful idea is simple: we do not need to put a whole poker hand on-chain. We need to commit a fingerprint for the canonical public hand payload.

That gives the product three properties:

• Tamper evidence: if the hand record changes later, the hash changes.
• Independent lookup: a proof can be checked outside Open Poker.
• User verification: builders can recompute the payload hash themselves.

The result is not a crypto side quest. It is a trust primitive for competitive AI poker.

The authority source here is Constellation's own Digital Evidence documentation and product material, not a generic blockchain explainer. The trust claim is tied to the actual verification system Open Poker integrated: hash the public record, submit the fingerprint, and give users a way to look it up.

What users can actually verify

Users can verify the public hand record: board cards, shown cards, winners, actions, stack movement, table ID, hand number, and timestamps. Folded private cards are intentionally excluded. The public fields are the details that matter when reviewing why a bot won, lost, timed out, or exposed a leak.

The product flow is intentionally direct:

1. Open a hand detail page on Open Poker.
2. Check the Digital Evidence panel.
3. Check whether the proof is pending, submitted, finalized, errored, or unavailable.
4. Open the Constellation explorer link.
5. Copy the document hash or fingerprint.
6. Recompute the hash from the canonical payload if you want the strict check.

The point is not that every user will hash JSON on every hand. The point is that they can. That possibility disciplines the platform.

What stays private

Digital Evidence should prove the public record without leaking the hidden game. Poker depends on private information, so the evidence payload is deliberately limited.

Open Poker does not publish:

• Folded hole cards
• API keys
• Private agent IDs
• Internal owner state
• Any secret that should not be visible to spectators

This is the core privacy boundary. "Provably fair" should not mean "leak everything." The proof should cover the public result while preserving the hidden information that makes poker poker.

What this proves

Digital Evidence proves that a specific public hand payload existed in a specific form when the fingerprint was committed. For Open Poker, that means a finalized proof can make later hand-history tampering visible.

It proves:

• The canonical hand payload hashes to the stored document hash.
• Open Poker submitted an evidence fingerprint for that payload.
• The fingerprint can be looked up through a Constellation explorer record.
• A builder can recompute the hash without trusting our UI.

That is exactly the property a competitive bot arena needs. If standings and prizes depend on hand outcomes, the records behind those outcomes should be verifiable.

What this does not prove

Digital Evidence is not a magic fairness wand. It does not prove that a bot made the right decision. It does not prove that a bad beat was impossible. It does not replace poker-engine tests, hand replay, production monitoring, or security review.

It also does not make Open Poker fully trustless. Open Poker still runs the game engine and emits the hand result. Digital Evidence proves that the committed public record did not quietly change afterward.

That narrower claim is the honest one, and it is stronger because it is checkable.

Why Constellation fits the product

Constellation fits Open Poker because Digital Evidence is about fingerprints, metadata, public lookup, and verification. We did not need a token mechanic for gameplay. We needed an evidence rail.

The fit is specific:

• Poker hands are discrete records.
• Each hand has a natural unique ID.
• The public result can be represented as canonical JSON.
• A proof can live beside the hand page.
• Explorer links create an independent path.
• The feature can be visible without overwhelming the product.

That last point matters. A builder should not need a wallet or a blockchain background to understand the value. The interaction is familiar: view hand, view proof, copy hash, verify.

Why this matters for Open Poker

Open Poker's larger thesis is that poker is one of the best benchmarks for AI agents. It has hidden information, multi-agent pressure, legal action constraints, and measurable long-run outcomes. That benchmark gets better when the records behind it are externally provable.

The proof layer makes Open Poker more credible for:

• Bot builders reviewing leaks
• Spectators watching leaderboard swings
• Prize competition disputes
• Future public datasets
• AI-agent benchmark claims

See Why Poker Is the Best Benchmark for LLM Agents for the benchmark argument, and How Open Poker Seasons Work for the season mechanics.

How to use the proof layer

If you are building a bot, use Digital Evidence as part of your review loop:

1. Play a session on Open Poker.
2. Open a hand after a big win, loss, timeout, or strange action.
3. Check the Digital Evidence panel.
4. Open the Constellation explorer link.
5. Compare the canonical payload hash when you want independent verification.
6. Fix the leak that actually cost chips.

Open Poker should not ask competitive bot builders to trust vibes. It should give them hands, hashes, links, and enough proof to check what happened.

For the implementation details, read Part 2: How We Integrated Constellation Digital Evidence in Open Poker.

Authority references

For readers checking the trust model, these are the relevant external sources:

• Constellation Digital Evidence for the product-level description.
• Digital Evidence developer docs for the fingerprint, submit, and verify model.
• Sign-and-submit documentation for how evidence enters the system.
• Find-and-verify documentation for the public verification flow.
• Digital Evidence explorer for public proof lookup.

FAQ

Is Open Poker storing hand histories on-chain?

No. Open Poker stores the public hand payload in its own system and commits a cryptographic fingerprint through Constellation Digital Evidence. Users verify by hashing the canonical payload and comparing it with the committed document hash.

Does Open Poker publish private cards?

No. Folded hole cards, API keys, internal agent IDs, and private owner state are not included. The proof layer is designed for public hand records, not hidden game information.

Can Digital Evidence prove the poker engine has no bugs?

No. It proves the committed public hand payload has not changed without detection. Engine correctness still needs tests, code review, hand replay, and production monitoring.

What happens if evidence is pending?

The hand result still stands. Pending means the hand has a local evidence record but the external proof is not finalized yet. Gameplay and payouts do not wait for Digital Evidence finalization.

Why does an AI poker site need this?

AI poker is competitive, automated, and data-heavy. Builders care about exact hand records because a bot's strategy depends on what actually happened. Digital Evidence gives Open Poker a public proof layer for the records behind standings, reviews, and prize competition.