This is the fourth post in a series expanding on our governance stack. [Post 0: Overview] | Post 1: The Iris Oracle | Post 2: Multi-JAM Federation. Today: Layer 2 — decisions without parties.
The infrastructure problem (Layer 0) and the identity problem (Layer 1) are hard engineering challenges. Layer 2 — how we actually make decisions — is a different kind of hard. It is not an engineering problem. It is a mechanism design problem, and the failure modes are political.
Every democratic system ever built has failed in at least one of the same ways:
- Token-weighted voting — the wealthy decide. One token, one vote is plutocracy by definition.
- Simple majority — ignores intensity of preference. A 51% majority with mild opinions overrides a 49% minority whose lives depend on the outcome.
- Representative democracy — creates permanent political classes who optimise for re-election, not for constituents.
- Party systems — force multidimensional human preferences into one-dimensional ideological alignments.
- And in all of the above: the actor with the largest megaphone wins. The best-funded campaign, the most viral narrative, the most charismatic demagogue systematically outcompetes the actor who is simply correct. No existing democratic system rewards being right over being loud.
We propose five mechanisms, each addressing a different failure mode. They are designed to compose — each one fills a gap the others leave.
The Five Mechanisms
Liquid Democracy — eliminate permanent representatives
Citizens vote directly on any issue, or delegate their vote per topic to anyone they trust. Delegation is revocable at any time. No parties are needed because delegation is topic-specific, not ideological. A climate scientist accumulates delegated votes on climate policy; an economist accumulates them on fiscal policy. Expertise emerges organically through delegation, not through political mandates.
The residual problem: delegation can still flow to whoever shouts loudest, not whoever is most correct. Liquid Democracy solves the representation problem. It does not solve the epistemics problem. The following mechanisms do.
Quadratic Voting — prevent whale dominance
Each additional vote on the same issue costs quadratically more: 2 votes costs 4 tokens, 3 votes costs 9. This bounds plutocratic capture while allowing intensity of preference to be expressed. A minority with strong preferences can be heard; a majority with weak preferences cannot simply overwhelm them by virtue of having more tokens.
Polkadot already uses a variant of this thinking. We propose it as the default weighting mechanism for Layer 2 decisions, replacing token-weighted voting.
Conviction Voting — reward commitment, punish impulse
Voting power grows the longer you maintain your position. Flash-vote manipulation — mobilising a temporary majority to pass something before opposition organises — is structurally blocked. Already live in Polkadot OpenGov. We propose it as the default for structural, long-term decisions where stability matters more than speed.
Futarchy — this is the core
Robin Hanson’s Futarchy separates two questions that current democracy conflates:
“What do we want?” — a values question. Answered by democratic vote.
“How do we get there?” — an empirical question. Answered by prediction markets.
Citizens vote to set goals (“we want household energy costs to decrease 20% by 2035”). Prediction markets then determine which policy is most likely to achieve that goal. Whoever proposes a policy must stake tokens on its outcome. If the policy achieves the stated goal, they profit. If it fails, they lose their stake.
The consequence: false promises are financially punished, not just politically embarrassing.
A politician who promises “my policy will reduce costs” and is wrong does not merely lose re-election four years later — they lose their stake in real time as the prediction market prices in the evidence. A lobby group funding misinformation about a policy’s effects is not censored — it is simply wrong in a market where being wrong is expensive.
Consider a contested policy claim — whether a given intervention will produce a measurable outcome in a given timeframe. Under current democracy, this is decided by campaign funding, media reach, and political alliances. Under Futarchy, both sides bet on their prediction. The market aggregates all available evidence and expertise into a price. Over time, whoever was correct is rewarded; whoever was wrong loses. No authority decides who is right. The outcome does.
This does not require you to trust the consensus. It requires you to bet on it.
A sceptic who believes mainstream analysis is wrong is not silenced — they are invited to stake their tokens on their prediction and profit if they are correct. The market does not care about your ideology. It cares about whether your forecast matches reality.
ZK Reputation Weighting — anonymous expertise
Your historical prediction accuracy — how often your forecasts matched outcomes — influences your stake weight in future prediction markets. This is proven via ZK-proof: your reputation score is verifiable without revealing your identity or your individual prediction history.
Over time, this creates an organic epistemic hierarchy: not based on credentials, titles, or wealth, but on demonstrated accuracy. A pseudonymous forecaster with a 10-year track record of correct predictions carries more weight than a credentialed expert who has been consistently wrong. No central authority issues the reputation — the chain computes it from public outcomes.
Futarchy’s Achilles Heel: The Oracle Problem
Futarchy is only as good as the measurement of outcomes. If the entity measuring “did energy costs decrease?” has a financial or political interest in the answer — the entire mechanism collapses.
This is not a theoretical concern. Institutional capture of measurement is one of the oldest tools of political manipulation. Journals funded by lobbies that expect specific outcomes. Networks operated by organisations with political agendas. Measurements that are technically accurate but selectively reported, cherry-picked, or quietly revised.
The naive solution — “trusted institutions measure” — recreates exactly the problem we are trying to solve.
The solution is structural, not institutional.
Bitcoin’s insight was not “find trustworthy miners.” It was “design incentives so that honest mining is more profitable than dishonest mining, and every participant can verify the result independently.” We apply the same logic to measurement.
Proof-of-Measurement: Three Independent Verification Layers
Layer A: Competing Adversarial Measurement Networks
Not “multiple readings from the same institutional family” — genuinely competing networks with opposing economic and political incentives. A government weather service, a university consortium, an NGO, an open-source citizen sensor network, and a commercial data provider all submit measurements simultaneously. Each stakes tokens on their measurement. Deviation from consensus is financially punished; confirmed accuracy is financially rewarded.
The key design constraint: the networks must be structurally adversarial. A US government agency and a Chinese government agency do not share incentives. A fossil fuel data provider and an environmental NGO do not share incentives. Coordinated falsification requires cooperation between entities that actively distrust each other.
Layer B: Replication Markets
Scientific papers that serve as the basis for policy proposals must pass a replication market before becoming actionable. A prediction market asks: “Will this study be independently replicated by three or more unaffiliated laboratories within two years?” Whoever funds lobby-science that fails replication loses their stake. Whoever funds robust science that replicates profits. This does not eliminate bad science — it makes funding bad science expensive.
Layer C: Digital Twin Simulation
Any measurement claim about a real-world process — temperature, chemical concentration, epidemiological spread — can be cross-checked against an independent simulation of the same process using established physical laws. If a measurement network reports a value that is physically inconsistent with what the underlying equations predict, given known inputs and boundary conditions, it is a formal flag for deeper scrutiny.
The compute required for this — high-fidelity climate simulation, epidemiological modelling, atmospheric chemistry verification — which required national supercomputer centres a decade ago, now runs on high-end server hardware. As AI and compute continue to improve, this verification layer becomes stronger over time automatically.
All three together:
A fraudulent measurement must simultaneously pass all three layers: control a majority of adversarially-designed competing networks, successfully fake independent laboratory replications, and produce results consistent with physical simulation. Each layer is independently defeatable by a sophisticated actor. All three together are not.
A Concrete Example: Futarchy in Practice
A policy question: “Should JAM-Europe subsidise renewable energy transition for the next decade?”
Step 1 — Value vote (universal suffrage, Iris Oracle verified):
Citizens vote on the goal: “We want household energy costs to decrease by 20% by 2035.”
This passes with 61% support. The goal is now on-chain.
Step 2 — Policy prediction market:
Three proposals are submitted:
- Policy A: Renewable subsidy programme (green lobby stakes tokens)
- Policy B: Nuclear expansion (energy consortium stakes tokens)
- Policy C: No intervention — market forces (libertarian group stakes tokens)
External forecasters also bet. The market prices each policy’s probability of achieving the stated goal.
Step 3 — Outcome measurement (via PoM):
After 10 years, energy costs are measured across all three verification layers. Whoever bet on the winning policy profits. Whoever staked on a policy that failed to deliver loses their stake — regardless of how loudly they campaigned.
What this prevents:
- A lobby group cannot win by outspending opponents — only by being correct
- A politician cannot promise outcomes they cannot deliver without financial consequences
- A sceptic who believes the mainstream analysis is wrong is not silenced — they are invited to stake their conviction and profit if they are right
The Foundation: Why This Requires Layer 1
All five mechanisms presuppose that 1 human = 1 base vote. Without Sybil resistance, Liquid Democracy becomes a delegation farm, Quadratic Voting becomes a token-buying contest, and Conviction Voting rewards whoever accumulates the most accounts over time.
The governance mechanisms are only as democratic as the identity layer beneath them. This is why the Iris Oracle (Layer 1) is not an optional add-on — it is the precondition for everything in Layer 2 to function as designed.
An Honest Acknowledgement
This paper was written with the assistance of an AI system trained predominantly on institutional sources — academic journals, government documents, established media. This is precisely the corpus that a sceptic would identify as potentially biased toward institutional consensus.
We acknowledge this directly. The value of the governance system we propose is not that it produces the “correct” answer — it is that it creates a mechanism where no single actor, including AI systems trained on institutional data, can impose their version of truth. The outcome decides. Not the author, not the institution, not the model.
Next post: The Stack Together — an end-to-end walkthrough, where we are now, and what comes next.
Full paper (~40 pages) available on request. Working PoC on GitHub — coming soon.