Bitcoin Hivemind

Bitcoin Hivemind, formerly Truthcoin, is a peer-to-peer oracle protocol that absorbs accurate data into a blockchain so that Bitcoin holders can speculate in on-chain prediction markets. Built as a Bitcoin sidechain, it lets anyone holding Bitcoin take part in those markets.

The project's elevator framing is that Hivemind is "Three existing ideas…combined!" Each contributes one stated core advantage:

At a glance, the design uses two coins: CashCoins (redeemable 1:1 for Bitcoin) for users, and VoteCoins (equity in an "oracle corporation") for the people who resolve outcomes. These are paired with a singular-value-decomposition (SVD) voting algorithm and a logarithmic market-maker (LMSR). Those mechanics are unpacked in How it works.

The founder

The protocol is the work of Paul Sztorc. Sztorc is a Yale Statistician with degrees in the blockchain-relevant fields of econ, psychology, and mathematics, who has been passionate about prediction markets since long before Bitcoin was created.

The rename: Truthcoin to Hivemind

A blog post dated 21 Sep 2015 announced that the project formerly called Truthcoin had been renamed Hivemind, for two stated reasons: to emphasize unity (the new site remained essentially the Truthcoin webpage, only with different sidebar colors), and to dodge "Altcoin stigma". As the post puts it: The name was originally 'truthcoin' and I changed the name (to avoid Altcoin stigma), without really telling anyone. The six-slide deck characterizes the whole system as the same as Bitcoin in every respect, apart from a few additional transaction types and data structures.

The motivating argument is laid out in Sztorc's manifesto, Market Empiricism (Statement of Purpose, v1.3), which opens on a Buddha epigraph: All suffering is caused by ignorance. Its thesis is that the defining problem of the information age is one of aggregation rather than availability.

Sztorc argues that conventional sources fail to inform and are therefore unacceptable. Formal research, he claims, is unavailable (paywalled even to Harvard), incomprehensible, slow, and, most provocatively, inaccurate: he contends that for university-grade research, somewhere from 60% to as high as 90% of published claims later turn out to be false. Informal media, in his account, rewards content that is flattering, amusing, frightening, or emotionally distracting.

The root cause he names is rational ignorance: when a single vote or voice cannot matter, the rational person invests no effort in the decision at all. A prediction market, the project argues, inverts this incentive, forcing a clear definition and paying people to become informed.

The recurring metaphor is a "Future Wikipedia": an open record where editing earns or costs money, and where, crucially, a price of 40 implies a 40% probability.

Sztorc frames all of this as a "Second Revolution" extending the Scientific Revolution and the Royal Society's motto Nullius in Verba: Take no one's word for it. The TABConf framing opens on Milton Friedman: …make it politically profitable for the wrong people to do the right thing.

The thesis's headline figures, that 60% to 90% of research is false, that no cable-news show reaches 1% weekly viewership, and that multiple Nobel Prize winners endorse the idea, are presented without in-text citations. More fundamentally, the source material does not quantify how accurate prediction markets are in practice: it offers no prediction-market-versus-poll accuracy data and no calibration figures. The case for their accuracy rests on design reasoning rather than on demonstrated results.

A prediction market is, in Sztorc's first-principles definition, a venue for buying and selling predictions. A valid prediction pays its owner $1 (the "unit price"); a false one is worth nothing, so the current price can be interpreted as the likelihood of the prediction coming true.

Binary markets alone can estimate the likelihood of any event; scaled markets can estimate the expected value of any quantity; the project holds that PMs deliver their most powerful insights only when the two are combined. Sztorc's taxonomy labels every market with three parameters, namely K (Decisions), N (States), and D (Dimensions), across eleven worked examples:

The market-maker is what makes this tractable: an LMSR lets a trader act selectively on only the information they possess (marginal, joint, total), while also preventing uninformed traders from throwing their money away. On the recreation side, the project claims a market has no 'house edge', and with only a 1% trading fee this is possibly the fairest proposition in the history of gambling.

The whitepaper, "Truthcoin: Peer-to-Peer Oracle System and Prediction Marketplace" (v1.5), opens on the problem it sets out to solve: Bitcoin can support financial derivatives and smart contracts, but the main benefits are lost if a trusted third party is required to inform those contracts. Its proposed answer is a proof-of-work sidechain that collects information on the creation and state of Prediction Markets, resolved through what the whitepaper calls an "oracle corporation" model, designed so that a group of self-selected, anonymous, greedy users will always resolve contract-outcomes honestly.

1 · Two coins, two layers

Like equities, VoteCoins are tradable and pay dividends over time, which the project argues removes the incentive for an "exit scam". Holders collect listing fees and half of all trading fees; the other half goes to Decision & Market Authors.

2 · Branches, Decisions, Markets & the Vote Matrix

Truthcoin can host many "oracle corporations" called Branches, each with its own VoteCoins. Decisions are questions that must be resolved by Voters (Binary or Scaled; a vote of 1=TRUE, 0=FALSE, .5=unclear). A Ballot is the set of all matured Decisions on a Branch; stacking ballots by row builds the Vote Matrix (n Voters × m Decisions), the object the consensus algorithm operates on. Markets partition the world into states, drawing their Decisions from any column(s) of any Vote Matrix of any Branch.

3 · Liquidity: the LMSR market-maker

A Logarithmic Market Scoring Rule (Hanson's LMSR) takes the other side of any and all PM-trades to ensure market liquidity, so every market has a tradable price at all times, even at zero volume. A user-chosen liquidity parameter b governs depth, and there is no need for Bids or Asks, or other order book artifacts. The exact numbers behind this live in The numbers.

4 · Resolution: an SVD Schelling game

Outcomes are computed by the matrix factorization known as singular value decomposition (SVD) on the Vote Matrix, a role the project likens to its use in principal components analysis, effectively a weighted PCA. The procedure produces a normalized reputation vector in which the most-deviant reporter always receives the greatest punishment. Critically, if (and only if) the votes are 100% unanimous, reputation-values do not change.

The mechanism that, by the project's account, makes this safe is framing resolution as a Schelling coordination game (Schelling's noon at the information booth at Grand Central Station). Because the right answers are known to all honest voters, any wrong answer must have been inserted deliberately; with sealed ballots, a "double-agent incentive" makes collusion self-defeating, since the project argues it is paradoxical to require a coalition of liars to communicate truthfully, in what amounts to a massive prisoner's dilemma. Afterward, RBCR (Reputation-Based Coin Redistribution) reallocates VoteCoins toward the consensus, smoothed by a parameter α (suggested new-weight 20% / old 80%).

The project itself downplays this machinery. Its FAQ says it is a common misconception that the SVD-resolution algorithm is the main innovation behind Hivemind. The real driver, by its own account, is the tradable share-token with owner-operators, together with the operationalization of economic costs and benefits.

5 · Security: the Φ = 65% band

The supermajority threshold Φ (Phi) = 0.65 (chosen, the whitepaper notes, because 2/3rds is a standard democratic threshold) defines three regimes. Below the floor, attacks fail and burn the attacker's stake; in the middle, an honest dissenter can force an Audit; only above Φ can an attacker rewrite a Branch's Decisions:

Three escalating safeguards sit atop the SVD: the Audit (every Ω = 6 months, disputed ballots re-resolved with CashCoin votes, and, the whitepaper notes, not a core requirement to the protocol design), the Miner Veto (if >50% of Miners Veto a Ballot, it has no effect), and the Miner Override. A dissenter holding between 35% and 65% has the ability to trigger an Audit.

6 · Cadence, branching, and free security

Voting runs on a fixed Intervote Period of ten weeks. Because merged-mining is free, the project says Truthcoin can run for free, with no coinbase rewards, with miners earning Bitcoin transaction fees and unable to censor market creation. The network scales by Branching, a controlled fork of the VoteCoin set, so that, as the project describes it, the network grows organically, branching in the same way that a healthy tree splits new branches.

These figures are design projections rather than operating results: the sources contain no deployed-mainnet metrics, on-chain data, audited security analysis, or independent benchmark of the SVD consensus, so every security and performance figure reflects the author's own estimates.

The project's claim of always-on, permanent liquidity is grounded in a single rich artifact: "Trading Under MSRs", the project's 27-sheet LMSR trading workbook (v2.1.1, July 18, 2020) built to demonstrate how trading works under Market Scoring Rules. Under an MSR, the workbook explains, issuance, price determination, and trading all take place in a single combined step.

The workbook never prints the LMSR formula algebraically, but its labeled columns and exact numbers imply the standard cost function C(q) = b·ln(Σ exp(q_i/b)), prices p_i = exp(q_i/b) / Σ exp(q_j/b), and a seed-capital identity k = b·ln(n). The headline trade-off is straightforward: greater liquidity requires a higher upfront cost.

The voting workbook does not model the consensus

A second workbook, the project's voting workbook, has a name that might suggest a demonstration of the SVD resolution, though its focus lies elsewhere: it does not include the SVD, a multi-voter vote matrix, or reputation weighting. It is instead a combinatorial study of how a single voter's relevance and sovereignty change as the number of other voters grows. The workbook identifies the point at which research-based voting begins to fail, a condition it labels rational ignorance, at 8 other voters, and it proposes a Rule of 20.

Although the design places considerable weight on SVD resolution, the available sources do not include a numeric, chartable "vote matrix to reputation-weighted outcome" example. The whitepaper describes the algorithm in prose and references figures (a 7-voter and 6-voter matrix, and a 256-ballot example), but only the captions survive, not the underlying cell values. No chart of the resolution step can therefore be drawn from sourced numbers. The closest real artifact is the deck's binary vote-tally example shown in How it works.

The clearest way to understand a prediction market is to trade in one. The sandbox below is a faithful implementation of the LMSR described in the project's LMSR trading workbook. Adjusting the liquidity parameter b and trading YES and NO shares moves the price; at every moment the YES price equals the market's probability that the event happens. Every number it shows is computed from the formulas the workbook uses, and reproduces the workbook's own figures exactly.

The sandbox reproduces the workbook's own numbers exactly, for example, the opening cost C(0,0) = 7·ln 2 = 4.8520302639196169 at b = 7, alongside the seed-capital values from the k = b·ln(n) table. The cost function is C(q) = b·ln(Σ exp(q_i/b)) and the price of each state is p_i = exp(q_i/b) / Σ exp(q_j/b).

Paper 3 ("Prediction Market Uses (Other Than Prediction)", v1.4, 16pp) sorts uses into three buckets: applied prediction, financial services beyond forecasting, and five cryptocurrency "Big Ideas."

On the financial side: insurance (hedge a NYC earthquake), portfolio replication of indices like Gold/DJIA/FOREX, put/call derivatives (which the author says can be combined to form any modern financial derivative), and shorting anything. The five "Big Ideas" extend PMs into governance and money:

Futarchy: governing by markets

The most ambitious application is futarchy (Robin Hanson's term): pick a metric of success, then let a conditional market choose the policy that maximizes it. Two worked examples carry the idea.

San Francisco housing (2020 blog). A 2×2 market crosses whether the mayor is re-elected in 2023 with 2024 one-bedroom rent, bounded at $2 to $10 per square foot. At onset all four cells price at 0.25, implying a 50% re-election chance and an estimated 2024 rent of $6.00. The author argues that the market removes political disorganization merely by existing, and that manipulation is effectively impossible, since the mayor can keep her money only by keeping her word.

The Sumner-Hanson nGDP Quartet (2020). A 2×2 LMSR market crossing an nGDP target with a money-supply lever is presented as a countercyclical Fed that always hits its targets, with the bottom-left contract designed to make money either way. The pitch leans on cost: the author estimates at most $1 million to administer and $5 million to subsidize it handsomely, against a Federal Reserve Board that runs at more than $790 million a year and a 2008 recession he prices at $1 to $10 trillion.

Every figure in this section (the exact target hits, the zero final ROIs, the +5% SF return, and the $5.73 rent) originates as a stylized worked example or thought experiment.

Paper 4 ("Prediction Market Myths") addresses three common beliefs about prediction markets.

Paper 5 ("Defeating Market-Manipulators…via 'Augmentation'") argues that PMs are highly resistant to manipulation: the author frames uninformed trades as risk-free arbitrage opportunities and contends that noise trades actually increase the returns to informed trading. The one durable exploit, the paper says, is the self-fulfilling prophecy, which it counters through augmentation: a new market that simultaneously predicts the target outcome and any number of influential intermediate decisions, exposing conditional prices.

The essay argues that the worry does not withstand a sober examination, resting its case on a four-part thesis: (1) assassination markets (AMs), as originally proposed by Jim Bell, are irreconcilably different from prediction markets; (2) the public-goods method is incompatible with crime; (3) markets are an excessively complex, risky, and convoluted form of criminal financing; and (4) Truthcoin offers a peaceful alternative.

Why an AM isn't a market. In Bell's design the assassin alone secretly picks the death date; in a real PM, anyone can buy any amount of 'Yes' at any time. The essay concludes that Bell's 'Assassination Markets' aren't really 'markets'…there's no price, no trading activity, and no speculators…you can't even sell.

The self-correcting contradiction. Because prices react to wagers (a "centrifugal governor"), spending to push "Die" drives its price toward zero, and ~40 million near-free "Die" shares would be snapped up by anyone with actuarial tables. Hence the fundamental contradiction: it is impossible for PMs to both [1] accurately measure the likelihood of something, and [2] change the current likelihood…to something else.

The author's own caveat. There's nothing suggesting it will work at all. The T-DAC and SI's are completely new and theoretical concepts invented by me. They have not been tested by anyone, they have not even been peer-reviewed. And the "Anti-AM Trump Card": the Victim is…in arguably the best position to cause this death…[and] will almost always be able to fake his own death or disappearance.

How Truthcoin is designed to suppress it. The design aims for censorship resistance rather than a guarantee against all censorship. Sztorc plans to launch the first Branch with a guideline that defines anything violent as off topic, keeping violence out of the initial Truthcoin system. Re-allowing it would require buying the whole corporation, after which honest users migrate away and the malicious branch's value collapses to zero.

The essay argues in prose rather than through data tables, and its figures (up to 40 million dollars, roughly 5 million dollars, one hour) are illustrative hypotheticals rather than measured quantities. Its closing Nirvana Fallacy argument, that prediction markets may also prevent deaths on the order of millions, is one the author himself frames as an unknown estimate.

The "What People Are Saying" page is a project-authored reception page spanning five sections (Highlights, In The Media, Forums, Academia, Websites). Its three headline endorsements appear below, carrying the home page's dates:

A fourth, on the home page, from Dr. Adam Back (Dec 2015): Hivemind is a real project with interesting use cases. It's great to see them using sidechains - anyone with Bitcoin can participate in their markets. Beyond the highlights, the page collects sixteen forum handles. A sampling:

The page dates only a single June 2014 media interview rather than the individual quotes, and the Hacker News post is listed without engagement metrics. Its attributions and descriptors are the project's own. Two of the forum quotes, from ppcman and BldSwtTrs, are missing their closing quotation marks in the source and are preserved here exactly as found.

Few projects publish a self-authored risk register; Hivemind does, in the first person, describing the section as a continuing work in progress that invites outside criticism and noting that the design has already been improved three times. Concerns are sorted into three explicit tiers, each with the project's own rebuttal attached.

Might be a problem

The most severe failure case is a Φ% Branch takeover. The project's recourse argument is economic: if buying 20% of a Branch costs x, buying 40% necessarily costs more than 2x, and the cost of reaching Φ% (always above 50%) is highly uncertain and could even be infinite if the remaining holders refuse to sell. Damage is also bounded: once one ballot is attacked, no one expects future ballots to resolve honestly, which limits the attack to a single surprise. The other tier-1 concerns are stated candidly, spanning complexity (a larger codebase implies more bugs, surfacing over several years, and developer onboarding that may prove insufficient) and competitive displacement (Hivemind may struggle to compete with a prediction-market business operating in a pro-business jurisdiction).

Probably not a problem Definitely not a problem

Lower tiers cover VTC price collapse (the project holds that markets are very robust to rumors), miner laziness (it treats the Miner-Veto as in no way a requirement), Sybil and ballot-stuffing (which it likens to buying all the shares of a bank and then voting to hand yourself its money), and slow resolution (which it considers a nearly irrelevant detail). On viability, it points to InTrade, which saw trading volumes in the tens of millions of dollars, and argues that even 50% of 0.1% of that would still be $10,000.

The adjacent polemics

Three combative blog essays sharpen the worldview. In "The Case Against Augur" (16 Dec 2015), Sztorc, who opens by noting that he created Augur, argues that the competitor has no future, dismisses Ethereum, and names individuals the essay accuses of having knowingly defrauded the public. "Against Crowdsales" (14 Dec 2015) is blunter still, urging readers not to crowdsale and calling it immoral to sell dreams. "Hostility Toward Election Betting" (8 Nov 2016) explains why markets unsettle people: a 40% plurality in a poll suggests a candidate will win, whereas 40% in a prediction market implies the same candidate will probably (60%) lose.

Because the Weaknesses page is the project's own risk register, each rebuttal it contains is self-assessment rather than independent verification; the author even flags one of his rebuttals as a cop-out. The Augur essay is a polemic, and its references to fraud and scam are accusations rather than established fact. The success percentages it assigns to projects other than Hivemind's 42% appear only within untranscribed images and are labeled as completely subjective opinions.

The nine blog posts span 21 Sep 2015 to 15 Jul 2020 (with a near-four-year gap after Nov 2016); the nine talks run Jul 2014 to May 2024.

Across this arc, the public record documents design and planning milestones rather than a shipped product to date. The available sources do not identify a confirmed mainnet launch date or a shipped release; the "could be live this autumn" (2019) and "rewriting in Rust" (2024) statements describe intent, not delivery. The TABConf, Scaling Bitcoin Montreal, and Anarchapulco YouTube transcripts are auto-generated captions carrying recognition errors, and so inform the themes above without supplying verbatim quotation.

Glossary

Two terminology notes help when moving between the sources. The whitepaper refers to the system as Truthcoin throughout, while the FAQ and decks use the later name Hivemind; and the term futarchy appears in the applications and blog material rather than in the FAQ, whitepaper, or Outcomes deck.