Proof of Human Work, explained

Bitcoin proves you burned electricity. Ethereum proves you locked capital. Worldcoin proves you have eyeballs. Proof of Human Work proves you did real, verified, attestable work — and the network mints $MANAV in proportion to it. It is the first consensus mechanism whose backing asset is what the world actually values: human contribution.

If proof-of-personhood is a passport, proof-of-human-work is a payslip. The first says "this person exists." The second says "this person showed up, contributed, and the network owes them." Different problems, different math, different incentives — and different downstream consequences for which token economies survive the agent age and which don't.

What every prior consensus actually proves

To see why PoHW is necessary, look honestly at what each existing consensus mechanism is actually backed by. The mechanism is the answer the chain gives to the question "why should I trust this ledger?" Each answer is a partial truth that paid off in its era and has aged in ways nobody quite wants to admit.

• Proof of Work rewards electricity. Bitcoin's miners are not contributing to anything outside the ledger; they are spending energy to make the ledger expensive to forge. Brilliant cryptography in 2008. Wasteful in a world that has now run out of "carbon-neutral hash" stories. The block reward is denominated in burned coal.
• Proof of Stake rewards capital. Validators with more tokens earn more tokens. Efficient relative to PoW, but it bakes plutocracy into the protocol — the rich get richer at exactly the rate the protocol allows — and it cannot tell you whether the staker behind the address is a human, a fund, a sovereign wealth pool, or a Cayman shell. The consensus is "money trusts money."
• Proof of Personhood rewards existing. Worldcoin's WLD grants are issued for being a unique human, verified once at an Orb. The token says nothing about whether you have done anything since you scanned. The consensus is "I am here," with no follow-on commitment.
• Proof of Stake of Reputation (the various social-graph experiments) rewards being known. They turn out to be re-implementations of LinkedIn endorsements with worse cryptography. They are gameable in exactly the ways LinkedIn endorsements are gameable.

None of these answer the question that matters in 2026: was a real human in the loop, and did that human contribute economic value the network can verify? A consensus mechanism that cannot answer this question is a consensus mechanism whose token will be repriced as the agent economy matures.

What PoHW is

Proof of Human Work is a consensus mechanism in which token issuance is bound to verifiable, attributable, human-supervised contribution. It mints $MANAV at a rate proportional to the volume and provenance of attested work that has cleared the protocol's anti-fraud filters. The five steps are simple to describe and surprisingly hard to game:

1. Register a Manav identity (Layer 1 of HATI: biometric + device + behavioural).
2. Do work — code, design, decisions, supervision, mentoring, agent management, peer review.
3. Attest the work via Manav SDK, browser extension, IDE plugin, or integrated tools (Layer 3 of HATI). Attestation is a signed credential, not a self-report.
4. Earn $MANAV proportional to the verified contribution, weighted by the work's provenance and your trust score.
5. Compound — your trust multiplier amplifies your mining rate as your verified history grows.

The mining rate is defined by a simple, auditable formula:

$MANAV_earned = base_rate × work_proof_count × trust_multiplier × scarcity_curve

Where base_rate is dynamic and adjusts to maintain the global emission schedule; work_proof_count is the count of verified Layer 3 attestations in the epoch; trust_multiplier ranges from 1.0× to 3.0× based on your Manav Trust Score; and scarcity_curve halves every two years — Bitcoin's halving rhythm, but indexed to attestation epochs rather than block heights.

What counts as work

Not all work earns equal weight. The protocol defines four tiers, calibrated to the economic value of the contribution and the difficulty of fraudulently producing it.

The asymmetric weighting is deliberate. High-weight work types share two properties: they require a counterparty signature (you cannot self-attest meaningful weight), and they produce an artefact whose authenticity can be re-verified later. Low-weight work earns roughly an order of magnitude less than high-weight work in the same epoch — enough to incentivise honest reporting of low-effort presence, not enough to make spam-attendance a viable mining strategy.

Why this is harder than it sounds

"Proof of useful work" is one of the oldest unsolved problems in cryptocurrency design. Three failure modes have killed every prior attempt, and PoHW takes each one seriously.

The Sybil problem. If a human can run ten identities, the protocol pays ten times for one person's work. Most prior PoUW attempts handed Sybil-resistance to ad hoc social systems and broke immediately. PoHW solves this by binding to Layer 1 of HATI — biometric and device-attested identity, with continuous re-verification. The cost of running a second Manav is not "spin up a wallet." It is "have a second body." We model that as cost-prohibitive at scale.

The Goodhart problem. Any measure becomes a target. If commits earn $MANAV, attackers commit empty pull requests. If meetings earn $MANAV, calendars fill with empty meetings. PoHW's countermeasure is two-fold: weight-by-counterparty (no signature, no weight) and the trust multiplier (low-effort attestation patterns are downweighted by graph anomaly analysis on the attestation network). Goodhart can only be partially defeated; the design goal is to make the gaming more expensive than the reward across reasonable horizons. Spam attestations become the protocol's tail-risk problem, not its dominant strategy.

The agent-laundering problem. A human assigns all work to an agent and claims credit. PoHW distinguishes authored, supervised, and directed work, and the gradient is steep: directed work earns a fraction of supervised work, which earns a fraction of authored work. Truthful labelling becomes the dominant strategy because over-claiming is detectable on review and over-claimed attestations are slashable. The system rewards the boring honest answer.

The anti-fraud stack

Three independent mechanisms compose to make sustained fraud unprofitable:

1. Layer 1 unicity — biometric template uniqueness with a small statistical false-rejection rate. Sybils require physical impersonation.
2. Bilateral attestation — high-weight claims require a counterparty signature. Self-signed claims contribute zero weight. Collusion rings become the dominant attack vector — and they are detectable graph-theoretically.
3. Slashing + reputation penalty — proven fraud results in $MANAV stake forfeit and a permanent T(h) penalty. The combination is designed to make sustained fraud unprofitable across reasonable horizons; one-off fraud is possible and survivable, sustained fraud is not.

None of these are new primitives. The novelty is in the composition: each mechanism alone is breakable; the three together raise the cost of breaking the system above the value extractable from breaking it.

Why now

PoHW becomes implementable only when three independent technological thresholds are crossed in the same year, and 2025–2026 is the first year all three were crossed at once:

• Cheap zero-knowledge proofs — selective disclosure (BBS+, SD-JWT) and range proofs (Bulletproofs, Plonk) are now production-mature with multiple audited libraries. Attestations can be private without sacrificing verifiability.
• Agent observability — MCP, ERC-8004, and the major agent frameworks have converged enough that the protocol can tell what an agent did, with what scope, in real time. This was not true in 2023.
• Biometric Layer 1 at scale — Apple Passkeys, government eID rollouts, and the maturation of multi-modal liveness detection mean Layer 1 unicity is no longer a research project. It is a deployment problem.

All three crossed the threshold in the last eighteen months. The window for a clean implementation is open now. It will close as platform-locked partial implementations entrench.

What it pays for

$MANAV earned through PoHW is not a grant. It is a receipt. A receipt that says: a real human, on a real day, did real work the network has cryptographically verified. That receipt is fungible across four use cases, each of which closes a loop in the protocol's economic flywheel:

• Verification gas. Every Layer 2 delegation, Layer 3 attestation, and Layer 4 trust query consumes a small amount of $MANAV. The gas is burned, not paid to validators. At protocol scale this is the dominant deflationary force.
• Trust amplification. Stake $MANAV against your handle to amplify your Layer 4 score; locked supply, slashable on fraud.
• Marketplace settlement. The Layer 5 marketplace settles in $MANAV. Smart contracts release payment only on Layer 3 attestation of the deliverable.
• Governance. Protocol parameter changes are voted on with quadratic-style weighting where 1 human = 1 vote and stake amplifies — but does not dominate — voting power.

Bitcoin's slogan was "1 CPU = 1 vote." PoHW's slogan is "1 verified human-hour = 1 unit of value the network can prove." It is a less catchy line, but a more honest one — and it is the only consensus mechanism whose token, ten years from now, will trace back to a thing the world was already paying for.

The first cryptocurrency mined by what humans do, not by what computers waste.

Common questions

Is this just a UBI airdrop? No. Universal basic income rewards existing. PoHW rewards contributing. A registered Manav with no attested work earns nothing. The two designs answer different questions; conflating them produces bad policy.

Can AI agents farm PoHW? No. Agent-only work earns directed-attestation weight (the lowest tier) and only when a named human delegated and supervises the agent. Pure machine output is uncountable as work. The protocol intentionally cannot mine itself.

Is $MANAV a security? The token is designed as utility — gas, access, governance, work reward — and is primarily earned, not purchased. Under both the SEC's Howey factors and the EU's MiCA framework, a token earned through productive activity, whose value derives from the holder's own work rather than the efforts of a centralised promoter, sits outside the canonical security definition. The full classification analysis lives in the whitepaper.

What's the emission schedule? 40% of the 10B fixed supply is allocated to PoHW mining, halving every two years over a roughly ten-year curve. Approximately 95% of the work-mining bucket emits in the first decade; the residual tail provides long-term incentive floor for late participants.

What stops a billionaire from buying their way to a high trust score? The trust multiplier is logarithmic in stake and bounded above at 3.0×. A 100,000× stake difference yields roughly a 5× voting/mining advantage, not a 100,000× advantage. Capital amplifies; it does not dominate.

What happens after the 10B is fully emitted? Inflation is capped at ≤1% per year by governance, available only by supermajority vote and only after Year 10. The default state is a fixed-supply asset.

The line that matters

Every consensus mechanism encodes a value claim about what the network is for. Bitcoin's claim was "scarce digital money is worth defending." Ethereum's claim was "programmable money is worth coordinating around." PoHW's claim is older and more obvious: human contribution is the only economic input that has held its value across every technology shift in history, and the agent age is the first one in which that claim is not implicit in every layer of the stack. PoHW makes it explicit again.

That is the whole thesis. The math is downstream of it.