Proof of Human Work — the technical spec

PoHW is not mining and not staking. It is a continuous attestation that a verified human contributed measurable, attributable work — and it pays them in $MANAV for it. This is the spec that turns that sentence into something an engineer can implement.

The five primitives

Every PoHW emission is the product of five signed objects. Identity commitment — a Manav DID anchored to a hardware-attested device, the human's root of trust. Work artifact — a hashed reference to the thing produced (a commit, a document, a closed ticket, a delivered design). Role attestation — the human's claimed role on the artifact: authored, supervised, or directed, with descending PoHW weight. Witness signatures — at least one independent party (a reviewer, a customer, a co-author) signing that the role claim is accurate. Time anchor — a notarized timestamp from the Manav verification ledger, ensuring the artifact was committed before any rewards were calculated.

The emission formula

The reward for a single artifact is approximately:

R = E_t · w_role · log(1 + S_human) · A_artifact · (1 - d_decay)

E_t is the epoch emission rate, which halves every two years on the schedule published in the whitepaper. w_role is 1.0 for authored, 0.5 for supervised, 0.2 for directed. S_human is the human's current Trust Score, log-scaled so reputation amplifies but never dominates. A_artifact is the artifact's measured magnitude — lines reviewed, words written, dollars closed — capped at the 95th percentile to resist gaming. d_decay is a freshness term: artifacts older than 90 days earn nothing.

Signatures and chains

Every signed object uses Ed25519 by default, with BBS+ available where selective disclosure of attributes matters (think: prove you reviewed the commit without exposing which file you touched). The artifact, the role attestation, and the witness signature form a chain. The chain's Merkle root is anchored to the verification ledger every block. Anyone — auditor, employer, regulator — can verify a payout was earned by walking three hashes.

Anti-Sybil design

The single hardest problem in PoHW is preventing one human from spinning up ten attested identities and farming themselves. Manav's defense is layered. Hardware attestation binds each DID to a unique secure element. Liveness checks at registration confirm a real human, not a recording. The witness graph requires real reciprocal relationships — three witnesses to one identity is a yellow flag; the same three witnesses across forty identities is a guaranteed blackball. Behavioral biometrics fingerprint typing cadence and editing patterns; a single human running ten "humans" produces statistically identical work patterns and the network rejects the duplicates. The Worldcoin orb-farms taught us exactly which checks were missing; PoHW ships them by default.

Consensus and finality

PoHW does not need its own L1. The verification ledger runs as an application chain on the Manav settlement layer, with attestations finalized in 6-second blocks and emission payouts settled hourly. The chain only stores hashes and signatures; the artifacts themselves live in the customer's existing systems (GitHub, Notion, Salesforce) or in encrypted user-owned storage. This keeps the chain small and compliant by default.

What's verifiable, what's private

Verifiable: the human existed at the time of work, the role claim is signed by witnesses, the magnitude is within bounds, the payout matches the formula. Private: the contents of the work (a payroll spreadsheet's numbers stay in the customer's tenant; only the hash is on-chain), the human's identity (selective disclosure means a recruiter can verify "shipped 200+ commits" without learning the human's name).

What this enables

Once PoHW exists, three things become possible that today are not. Hiring becomes cryptographically auditable: a candidate's claim of "shipped at Stripe" is a verifiable artifact, not a LinkedIn line. Compensation becomes attribution-aware: when an agent ships work alongside a human, the protocol resolves who gets the credit and who gets the bonus. Reputation becomes portable: a designer leaving an agency takes their attestations with them, not the agency's CRM record of them.

Common objections

The honest objections: protocols without products fail, and tokens without revenue are speculative. We answer the first with a hosted commercial layer and design partners shipping today; the second with a fee model where the protocol earns from real agent verifications, not from token velocity.

Frequently asked questions

Is this about a token, or a protocol? A protocol first. The token exists to align incentives — humans earn for attested work, relying parties spend to verify, the network captures a small fee. The protocol stays useful even if the token's price is volatile; the token gets useful only when the protocol is.

How is this not just another crypto identity project? Most crypto identity projects answer 'is this a unique human?' and stop. This one answers 'what did this human authorize?' which is a different question. The substrate (hardware-attested device, recoverable, revocable) is also designed for enterprise compliance, not just consumer Sybil resistance.

What happens if the chain has an outage? Verification continues. Signatures verify locally; the chain anchors audit roots periodically, not per-action. A multi-day chain outage would delay forensic anchoring but not block any agent action that already had a valid delegation.

Where to start

Move from this to proof of human work for the technical design and manav token explained for the economic shape. The protocol and the token are designed to be read in that order — design first, incentives second.

Bitcoin proved you burned electricity. PoHW proves you did the work. The first paid the world for compute. The second pays the world for being human in front of agents.