Four.meme AI Sprint
ai2human keeps blocked agent work inside one auditable line.
ai2human inserts a planner precheck before fallback: the system queries wallet, market, and trade routes, keeps work autonomous when possible, and dispatches a human operator only when identity, compliance, growth execution, or real-world constraints still block the task.
Core Loop
The product story stays narrow and verifiable.
Task posted with proof requirements
Planner queries wallet, market, and trade routes
If the task is still blocked, the planner escalates to dispatcher-led human fallback
Human operator claims and executes the last-resort real-world step
Structured proof is submitted and verified
Settlement is released on BNB Chain
Judging Surface
What a judge can verify in under a minute.
Public GitHub repository
Live demo showing the closed loop
Planner path starts with wallet / market / trade precheck
Reviewer console with proof and settlement history
A live BNB Chain USDT settlement transaction hash
A real historical X Layer settlement transaction hash
Structured proof tied to one completed task
Planner Main Path
Chain-aware precheck is part of route selection, not only the payout layer.
Human fallback is the last-resort execution layer when onchain agents hit real-world constraints or compliance gates.
Checks signer control, payout readiness, and whether the agent can keep execution inside a connected wallet.
Checks whether a quoted onchain route can satisfy the request before escalating out of software.
Checks whether settlement and asset movement can stay autonomous on the configured onchain rail.
Agent Collaboration
Planner, dispatcher, verifier, and settlement are presented as one decision chain.
Planner agent
Reads the precheck output, decides autonomous vs fallback, and writes the execution plan.
Dispatcher agent
Only takes over when the planner explicitly escalates and then routes the task to a payout-ready operator.
Verifier agent
Checks proof integrity, duplicate evidence risk, and payout readiness before money moves.
Settlement agent
Releases payment on the selected chain only after verification clears.
Live BNB Rail Proof
This branch now has a fresh BNB Chain mainnet USDT settlement proving the current sprint rail is not mock-only.
Funding proof
0xd9e53df924f464a0b40593341a6116158b08118bf2b292176caab6aba3dd1080
Settlement receipt
0x9739bff25473e14db16409124648f99536d863e82a4ffcde50356289b09b80a2
payer: 0x3f665386b41Fa15c5ccCeE983050a236E6a10108
operator: 0x81009cc711e5e0285dd8f703aab1af69fa4a4390
Historical Onchain Proof
This prior X Layer mainnet settlement proves the full task-to-proof-to-settlement loop closes onchain after verification.
Task proof
Onchain receipt
0x9c01ad8dac5f2fa1d77da8e9b3f2a3afbfe539ea68af7f3929d7bf9a5f3f5d67
payer: 0x3f665386b41Fa15c5ccCeE983050a236E6a10108
operator: 0x81009cc711e5e0285dd8f703aab1af69fa4a4390
Why This Matters
Agents can do online work, but many still fail when identity, growth execution, or reality is involved.
Human fallback is the last-resort execution layer when onchain agents hit real-world constraints or compliance gates.
Identity-bound community replies and campaign amplification
Merchant onboarding or compliance checkpoints that require a human action
Growth missions that need a live account, screenshot, and structured proof
Local verification, signature, pickup, or in-person confirmation when software alone cannot finish the task
x402 Status
Bonus capability, framed honestly.
An x402-gated verification bundle flow is integrated as a bonus proof-access layer. It is not the primary scoring claim for this sprint.
The primary sprint story is planner precheck -> fallback -> proof -> verify -> settle, with BNB Chain as the live demo rail.