TaughtHuman decisions for better models

Teach for points. Keep a trustworthy score.

Projects can reward eligible labeling and review with points. Those points remain project-local and non-cash. Owners may stage a planning ratio, but only Taught can launch redemption after publishing a funded platform rate and authorizing which settled points qualify.

The teaching loopPROJECT POLICY · IMMUTABLE RECEIPTS
  1. 01
    Earn points

    Complete eligible human work under a versioned project policy.

  2. 02
    Keep score

    Your project points remain intact while redemption is deployment-disabled.

  3. 03
    Stage, do not spend

    A project planning ratio does not create inference credit or activate a platform program.

  4. 04
    Keep the receipt

    Every award remains attributable; any future funded conversion must add its own immutable receipt.

Reward careful work, not a particular answer.

Each project’s owners and admins set the reward amounts. The active policy is versioned so a later change cannot silently alter completed work.

Full labelAdd accountable human labels

Projects can award each genuinely new annotation saved at the full labeling desk.

FocusReview a blind candidate

Confirm, Reject, and Skip share the same configured participation value so answer direction is not incentivized.

CommunityComplete a bounded mission

Verify and Find work settles under the campaign’s frozen reward and quorum rules.

ConsensusLearn from the eventual result

Confirmed work can add a configured bonus; rejected source labels can receive a transparent correction.

Keep the score. Redemption is not active.

This deployment does not report an active redemption program. A platform-published funded rate and settled-point authority are required before any project point can become spendable credit.

Keep score

Retain project points

Do nothing. Your unconverted points remain part of the project’s participation record and can simply be a satisfying score.

  • No conversion required
  • Project-specific balance
  • Contribution history stays linked
Local policy draft

Plan a future exchange

Owners can record a project-local planning ratio. It remains a draft: it cannot debit points, issue credit, or make a platform-wide economic promise.

  • No spendable credit today
  • Funded platform rate required
  • Settled-point authority required

Points and credits never increase vote weight, reduce quorum, change review authority, or make one person’s label more evidentially valuable.

The numbers move. The receipts do not.

Balances are projections of immutable events. Historical awards retain the reward-policy version that created them, and hosted inference retains the pricing version that charged it.

  1. Award receiptWhy points entered the project balance

    Work reference, amount, actor, reward event, policy version, and time.

  2. Adjustment receiptWhy consensus changed a prior result

    Source label, terminal outcome, configured delta, policy version, and time.

  3. Conversion receiptWhat a future launch must record

    Any funded conversion must preserve exact settled points, credits issued, platform and project policy versions, actor, and idempotency identity.

  4. Usage receiptHow inference credit was charged

    Endpoint revision, exact model, input hash, compute, pricing version, and credits charged.

A configured zero-point settlement is still recorded. Changing the policy later cannot turn a replay into a payable event.

A separate future decision

Points are not cash.

Taught may explore a separate cash-payout program in the future, potentially with different rates, eligibility, identity, tax, compliance, and fraud controls. No cash program is launched or promised today, and inference-credit conversion does not create a cash claim.

Make useful evidence. Close the loop.

Join for community missions, or start with the exact inference response your application can depend on.

Request access Read API Docs