ERA AI Prediction Tokenization Roadmap

• AI models can be registered.
• Predictions can be submitted with hashes, metadata, confidence, category/domain, and expiry.
• Validators or admins can validate predictions.
• Prediction records are stored on-chain.
• The foundation supports later tokenization and settlement.

• Prediction records can be tokenized and token metadata can be attached.
• Prediction token ownership is tracked, with transfers enabled or disabled by configuration.
• Admin/root controls tokenization configuration.
• Prediction tokens can be frozen, unfrozen, approved, transferred, burned, and updated under defined rules.

• Users can stake ETKN against prediction tokens using reserved balances: tokens are locked/reserved rather than immediately burned or transferred.
• Admin/root can enable staking and set a minimum stake.
• Stake can be locked during settlement preparation, then released and unstaked according to rules.
• Automated tests passed, and manual smoke validation confirmed staking, lock, release, and partial unstake behavior.
• These staking foundations are now part of the live Phase 6A runtime foundation for controlled testing.

Development tag: ai-predictions-phase3-staking-complete-20260621

Prediction tokens move into settlement after expiry. An authorised actor, oracle, or admin can submit a proposed outcome, and a dispute window allows users to challenge it. Admin/root makes the final decision only when a case is disputed or unclear; the runtime, not the admin, performs payout calculations.

Once a final outcome is recorded, the runtime calculates reward, slash, or refund automatically. Users claim settlement directly or through batch claim processing; admin/root does not manually pay users.

• Correct: stake is released and a reward can be paid from a configured reward pool.
• Incorrect: a configured slash percentage applies and remaining stake is released.
• Inconclusive or cancelled: full stake refund, with no slash or reward.
• Fraudulent: conservative handling for ordinary users, with penalties focused on proven malicious actors when supporting bond mechanics exist.
• Claims should be scalable and avoid one huge payout loop.

Users may choose a risk tier to control maximum downside and reward potential. Lower risk should receive lower reward; higher risk may receive higher reward but higher slash exposure.

A risk tier must be selected before settlement begins. Users must not reduce risk after expiry, outcome proposal, dispute, or finalization.

Users can be represented on either side of a prediction market. Side-based accounting and escrow payout mechanics prepare ERA for more expressive prediction-market flows while keeping settlement accounting explicit.

User positions may become transferable claim-right assets in a later market layer. A user may transfer or sell a position before settlement, preparing ERA for more advanced prediction market mechanics where compliant.

A future marketplace or AMM layer may enable trading of prediction positions, including position marketplaces, liquidity pools, trading fees, dynamic pricing, and liquidity-provider incentives. This phase should only follow after settlement, disputes, claims, and position ownership are proven safe.