@PredgeAI
Account based inBulgaria
About this account
- Account based in
- Bulgaria
- Connected via
- Georgia App Store
Account-level information from X, not a live location or the device used for a specific post.
Trust & dispute layer for AI-agent payments: a signed verdict on whether the paid work arrived, verifiable offline. Free key: https://xitter.cf/t.co/1VxRg2u5Xe
Joined October 2021
- Tweets70
- Following189
- Followers18
- Likes42
Pinned Tweet
Settlement risk on prediction markets, measured.
1 Jan to 2 Oct 2026: 3,005 disputes on 2,666 Polymarket markets. Of 2,543 settled disputed markets, 853 (33.5%) settled differently from the disputed proposal.
Signed, verifiable records. Free API: predge.io
What an agent's settlement logic should take from this:
- a price near 0.97 or 0.9995 is a price, not a resolution
- after a second dispute, plan for days
- do not spend collateral that is still in a vote
Predge signed a record of this dispute while the vote was open and again after it resolved. Both snapshots pass our verifier.
Full issue, tx hashes and files: data.predge.io/dispute-of-th…
Check a record: data.predge.io/verify
Built for @monad Metropolis: Predge Guard tells you, or your AI agent, whether a disputed Polymarket answer will hold before you trade it.
1 in 3 disputed answers changed in 2026. 26 verdicts already on Monad mainnet.
Tap a market, no typing: guard.predge.io
Sign in with email (@privy_io), the wallet you already use (@dynamic_xyz) or just a passkey (mera by Category Labs). No USDC on Monad? Aurora Intents tops up from 20+ chains.
On Android it's the same flow, and your fingerprint is the wallet.
Why Monad: measured on mainnet, a verdict lands 1 to 2 s after its request and costs about $0.001 (38 markets, 66 tx). Cheap and fast enough to require before every single trade.
Code: github.com/predgeAI/erc8004-…
Step 2, who's exposed: @nansen_ai data, paid per call over x402 on Monad.
Trump "Ammunition" is in a vote right now. Top holders: $91,170 on the disputed No vs $36,208 on Yes (12 Oct). That decides the next step: hedge, size down or wait.
Step 3, act in one tap: a @KuruExchange note that pays $1 if the answer flips (testnet, live book), a BTC short on Perpl margined in Agora's AUSD, or wait.
Agents get the same rule: our @MetaMask agent wallet plugin refuses the bet while the vote is open.
1 in 3 disputed Polymarket markets that settled this year ended up different from the disputed proposal (853 of 2,543).
Settle Watch alerts you when a market you hold is disputed, and verifies every record on your phone.
Built for #ClockIn 👇
github.com/predgeAI/predge-s…
Polymarket builders: Data API v1 is retired Oct 24 (/trades -> /v2/trades?condition=).
We moved our ingest and diffed 27,179 fills read from v1 and v2 at the same moment: 0 differences.
Gotcha: outcome_index 999 = outcome not labeled. Treat it as missing.
We scanned every Polymarket dispute in 2026 from on-chain logs.
2,666 markets were disputed. In 34% of the settled ones, the dispute flipped the outcome.
Market creators also edited rules 26,921 times after markets opened. In 223 of 303 disputed markets with edits, the first edit came after the dispute.
Method and a signed evidence pack: data.predge.io/samples/settl…
We post numbers like these in t.me/predge_news