A history, checked against the archive
Prediction markets kept dying for legal reasons rather than technical ones. The Iowa Electronic Markets survived by staying tiny; Intrade was sued out of the United States and then collapsed; DARPA's market was killed in a day. The crypto answer — Augur, Gnosis, Polymarket — was explicitly built to be unkillable, and the company that actually won, Kalshi, did it by registering with the regulator and suing. This is that history, and then what the early-crypto audio archive — now 50,432,534 words across sixteen searchable shows — recorded as it happened.
00 — Summary
Every claim below is established and sourced in the body; this is a compression, not an addition, and it carries the body's hedges with it. Sections 02–07 are the outside history, from primary documents and press. Sections 08–12 are the local transcript archive. The two piles are kept apart and are listed separately in the sources.
Interpretation, stated as such. Everyone named in this section is living and mostly still working in this field; claims about conduct are attributed to whoever makes them, and the readings below are about the evidence rather than about the people.
Two different kinds of material went into this report and mixing them would have been the easiest mistake available. Sections 02 to 07 are outside history: CFTC orders and press releases, court opinions, the Federal Register, congressional records, DARPA's own archived pages and contemporaneous reporting. Sections 08 to 12 are the local transcript archive. Nothing from the first pile supports a claim about what was said on air; nothing from the second supports a claim about law.
The archive half is searched with corpus.py against brain.db, and the
scope statement matters more here than usual because it has moved twice in two weeks.
50,432,534 words across 7,047 distinct broadcasts and 5,032 hours of audio. Sixteen of eighteen registered shows are searchable. Epicenter alone runs 2013-12-20 to 2026-09-09 unbroken, and Epicenter is the show that carries this subject: it did dedicated episodes with Paul Sztorc, Robin Hanson, Martin Köppelmann and Joey Krug inside twelve months of 2015–16, and it holds 99 of the 237 dated prediction-market broadcasts. Every previous report on this estate was written before Epicenter was indexed, so none of them could see any of that.
The two shows that are not searchable are KO (Countdown with Keith Olbermann, 2020–ongoing) and GQKO (The Closer / The Resistance with Keith Olbermann, 2016–2017). Both are outside the crypto corpus, both are held as commitments rather than files, and both are listed as risks by the coverage statement on any claim dated after they went on air — which, for the Kalshi and Polymarket firsts below, they are.
Bitcoin Uncensored is searchable and entirely undated. Its 369 broadcasts and 2,922,513 words were ingested on 22 October 2026 and not one row carries a date — the only surviving host is a flat alphabetical mirror, 4 of 371 filenames carry a date, and the channel that would date the rest is a 404. The register records this as deliberate, and it is right to: “a null date is honest; a guessed one would be a lie with a timestamp on it.”
The consequence for this report is specific and large. BU holds 18 prediction-market
broadcasts, 8 Augur broadcasts, 6 on Robin Hanson, and three of the most on-topic interviews in the
entire archive — Joey Krug of Augur at Blockchain Beach, Paul Sztorc of
Truthcoin at Blockchain Beach, and What is the function of Politics with Robin Hanson.
None of them can enter a single date claim in this report, because they have no
dates. They appear in the quotations and in the archive-wide totals, and nowhere in any
chronology, any spread table or any per-year chart — because
corpus.py spread keys on broadcast_id, which needs a date, so BU drops out
of those answers silently. That is why every count in this report is given twice: as a dated
figure and as an archive-wide figure. A single number would have hidden 23
prediction-market broadcasts, 30 Augur broadcasts and 3 assassination-market broadcasts behind a
tool's implementation detail. That asymmetry is the largest open gap here.
The corpus is Whisper output, and most of it came off whisper-base, the weakest model
in the family. Proper nouns are mangled systematically. For this subject the damage is the worst in
the archive:
| Thing | What Whisper writes | Effect of searching the right spelling |
|---|---|---|
| Augur | auger 235 · augur 41 · augre 2 | The correct spelling finds under a sixth of the material. Registered in aliases.json; auger is also a drill bit, so hits carry a low-confidence flag. |
| Intrade | in trade 70 · intrade 5 | The split form dominates 14 to 1. Most of the 70 are the ordinary English words “in trade”; the four real broadcasts were found by reading snippets, not by counting. |
| Sanjuro | sanjuro 1 · San Juro (in body text) · kuwabatake 0 | The pseudonym’s surname returns nothing at all across 50 million words. |
| Bitcoin (in the Bell interview) | midpoint | Whisper renders “Bitcoin” as “midpoint” through the whole second half of that episode. Quotes depending on it are flagged in place. |
| Kalshi | kalshi 1 · calshi 0 · “Kalishi” (in one Short) | Probed four ways. The count really is one. |
Every multi-word term in this report was quoted as an FTS phrase, because an unquoted one is an
AND across the whole transcript rather than a phrase — the trap that produced a seven-year
date error on this estate in October 2026, and the reason CONVENTIONS §4d exists. The terms that
needed it here were "prediction market", "prediction markets",
"assassination market", "assassination markets",
"idea futures" and "robin hanson".
./corpus.py first --any '"prediction market" "prediction markets"'
./corpus.py spread --any '"prediction market" "prediction markets"'
./corpus.py first --any '"assassination market" "assassination markets"'
./corpus.py first augur # alias group: augur | auger
./corpus.py first intrade
./corpus.py first polymarket
./corpus.py first kalshi
./corpus.py probe "idea futures" # 2 broadcasts: EPI 1, LTBX 1Attribution. Transcripts carry no speaker labels. Per CONVENTIONS §6 every quote below carries a tier and its grounds. A — self-identification in the passage, a hand-off this quote answers, or a single-voice programme. A− — the turn ends in a hand-off, or third-person elimination leaves one name. B — the roster is known and which person spoke is not. Nothing is named at tier B. This report is unusually well served: most of its best quotes are people saying their own names at a microphone.
Thomas Hunt is a participant in this story, not an observer of it. He hosted Jeremy Gardner twice, interviewed Jim Bell, put Tony Sakich on air for an hour and Paul Sztorc on after that. Where this report uses his recollection it says so and marks it as recollection; where the archive contradicts or refines it, the archive wins and the difference is shown rather than smoothed over. Section 11 is the main case of that.
This page was reviewed against CONVENTIONS before it was finished and six real defects came back. They are recorded here rather than silently fixed, because CONVENTIONS §3's logic applies to a report's own errors as much as to dead links: a correction that leaves no trace makes the work look tidier than it was, and the next session inherits the illusion rather than the lesson.
| Defect | What was wrong | What it is now |
|---|---|---|
| A tool's implementation detail was hiding 23 broadcasts | Every headline count — 237 prediction-market broadcasts, 189 Augur, 38 assassination-market — came from corpus.py spread, which keys on broadcast_id and therefore silently drops every undated row. Bitcoin Uncensored's 369 broadcasts are all undated by design, so its 23 prediction-market and 30 Augur broadcasts were outside every figure on the page — while §01 claimed in terms that “they appear in the counts”. They did not. |
Every count is now given twice: dated, and archive-wide. 237 → 260. 189 → 219. 38 → 41. The §12 table has a column for each. |
| “Unindexed” was confused with “undated” | Three first-mention claims were called unprovisional on the grounds that nothing unindexed reaches past 2013. But BU is indexed and entirely undated, and an undated row cannot be ruled out as earlier. The coverage statement corpus.py prints does not model this case at all — it reasons about unindexed shows and says nothing about indexed-but-undated ones. |
Only the 2012 Intrade claim is now called unprovisional, and only against every dated show. The prediction-market and assassination-market firsts are qualified. The gap in the priority layer is named in §12. |
| Two tier-A attributions were attribution from role | Both Joey Krug quotes from LTB #222 were named at tier A because the answers were technical protocol detail and he had been introduced as the core developer. That is exactly what §6 forbids — “never attribute from writing style, known opinions, or who is famous for holding a view” — and the episode has two guests introduced together with the question put to both. | Both demoted to tier B and unnamed, with a .nameflag bar. Two properly attributable quotes from the same episode were found and added in their place: a tier-A− turn ending in a named hand-off, and the tier-A turn that hand-off lands on. The replacement material is better than what was lost — it documents Robin Hanson as an Augur adviser, on air, in June 2015. |
| A tier-A card whose own grounds described A− | The interviewer's turn in the Jim Bell interview was at tier A on grounds reading “this is the questioning voice, which hands back to the guest” — which is §6's definition of A−, backward-pointing. No quote on the page used A− at all, which should itself have been a warning. | Reassigned to A−. It is the only interviewer-voice quote card on the page, and now the only reason the A− tier appears at all alongside the LTB #222 hand-off. |
| Mentions counted as broadcasts | “114 of the 121 Gnosis hits are in that single episode” — 114 is a count of mentions inside one episode and 121 is a count of broadcasts. 114 of 121 broadcasts cannot sit inside one broadcast. The same error made “futarchy: 5 broadcasts” read as two. | Units separated and stated wherever both appear. |
| A figure corrected in the body and not in the summary | The Facebook like counts were measured properly (53 Gardner, 47 Krug, 11 Augur, 0 Hanson) and the body was updated, but the summary still carried the earlier eyeballed 45 and omitted Gardner entirely — the largest of the four. | Summary corrected to match the measurement. |
Two further things the review got right and this page now states: Paul Sztorc's position
on Augur was asserted as fact in §10 while the sources list said it was CoinGecko's
characterisation, and is now attributed; and the Epicenter and Bitcoin Game rows
claimed “no public link — local archive only” for shows that publish publicly. The
second produced the most useful finding of the review: the slug in Epicenter's own
source_id looked like a recoverable URL, and
epicenter.tv/episode/143 returns HTTP 200 with the title “Episode Not
Found”. A constructed link would have passed a status-code check and been wrong
— CONVENTIONS §4's trap, caught by reading the title instead of the status.
Two things happened almost simultaneously and independently, and the whole subject descends from both.
The Iowa Electronic Markets were initiated in April 1988 at the University of Iowa by economists Robert Forsythe, George Neumann and Forrest Nelson, with political scientist Jack Wright. Trading opened on 1 June 1988 in $2.50 portfolios of vote-share contracts on Bush and Dukakis, and closed at 9 a.m. on 9 November 1988. 192 traders took part in total, and the market's final price beat the polls by under one percentage point. The result was published in the American Economic Review in 1992 as “Anatomy of an Experimental Political Stock Market”.
Robin Hanson dates the idea-futures concept to the autumn of 1988 in his own account; the earliest written manuscript is 17 March 1989; he gave the talk “Could Gambling Save Science?” at a risk-and-gambling conference in London in August 1990, published it as a book chapter in 1992 and as a journal article in Social Epistemology in January 1995. The Foresight Exchange, originally called Idea Futures, went up as a web market in 1994–95 — play money, and the site said so: “Remember, this is not real money!” It won a Prix Ars Electronica Golden Nica in June 1995. Hanson's “Decision Markets” followed in 1999, and futarchy — “Shall We Vote On Values, But Bet on Beliefs?” — circulated as a working paper in July 2000 and was not formally published until 2013, thirteen years later.
A note on citation, because this is the most mis-cited item in the literature: “Could Gambling Save Science?” exists as a 1990 talk, a 1992 book chapter and a 1995 journal article, and all three years are correct for different artefacts.
Iowa is the only prediction market that has run continuously from 1988 to 2026, and the reason is that it agreed to stay too small to matter. Its first CFTC no-action position was issued in 1992; it was superseded by Staff Letter 93-66 on 18 June 1993, whose conditions the Commission quoted back in its own 2008 concept release: access limited to 1,000–2,000 traders per submarket, a maximum of $500 at risk per participant per submarket, relief premised on the market's academic purpose, and no profit or compensation to the operator. A 2014 letter added that participants must execute their own trades, with no brokerage, no commissions and no advertising.
In 2026 it is still running, narrowed to three congressional-control markets, the $500 cap intact, and its director Thomas Gruca describes it to the Daily Iowan as “we're nonprofit, we're even non-revenue.” The CFTC's June 2026 rulemaking lists it, in the present tense, as one of only two outstanding staff no-action positions.
That is the shape of the entire pre-crypto settlement, and it is worth stating plainly before the next four sections: in the United States a prediction market could be legal or it could be large, and for twenty-two years it could not be both.
Intrade is the case that defines the period, and almost every popular detail about it is wrong.
John Delaney did not found it. Per Fortune's 2013 reconstruction, the company originated in 1999 with NYBOT floor traders John and Sean McNamara, trader Greg DePetris and Patrick Young, as Global Sports Exchange, with Ron Bernstein as CEO. Dublin was chosen because US legality was unclear. Delaney was hired in autumn 2000 as VP of Finance and became CEO of Tradesports when that entity was created around 2003. The New York Times's 2011 obituary headline — “John Delaney, Founder of Intrade, Dies at 42” — propagated the error, and both Fortune and The Currency state flatly that it is wrong. Wikipedia asserts the opposite ownership direction for the 2003 restructuring, so the sources conflict on that point and this report does not resolve it.
Intrade launched on 2 August 2001, raised $13.5m across two rounds — backers over time including Paul Tudor Jones, Stan Druckenmiller and Murdoch's Queensland Press — and lost $9m in its first two years. Its first real non-sports publicity came in December 2003, when its “Saddam Hussein captured or neutralised” contract spiked days before the 13 December capture.
This is the episode everybody half-knows, and the usual telling — DARPA was going to let people bet on assassinations — is false as a description of the project and true as a description of the project's own published artefacts. Both halves matter, and the report states both.
DARPA created the Information Awareness Office in January 2002 under Admiral John Poindexter. FutureMAP followed an existing SBIR topic, Electronic Market-Based Decision Support (SB012-012), first issued in May 2001; Phase I awards went to Net Exchange of San Diego ($97,750) and Neoteric Technologies in December 2001, with Net Exchange taking a $425,000 Phase II in autumn 2002. John Ledyard, Hanson's thesis advisor, was Net Exchange's CEO and founder; Charles Polk was its president. Hanson contributed the application area and the logarithmic market scoring rule that was to be the market's sole trading mechanism, and received about $30,000 of the funding as a subcontractor — a disclosure he makes himself.
The Senate Appropriations Committee had already zeroed IAO funding, FutureMAP included, on 10 July 2003 — before the story broke. Then:
/images/ directory is emptied. The archived Apache index shows
only “Parent Directory 29-Jul-2003 12:33”.What PAM said it would trade, from its own
pam_trade_info.htm: three types of contract — quarterly contracts on data indices
tracking economic health, civil stability, military disposition and US involvement in
Egypt, Iran, Iraq, Israel, Jordan, Saudi Arabia, Syria and Turkey; quarterly global
indicators; and “specific possible events”. Registration was to open 1 August
2003, live trading 1 October, capped at 1,000 traders rising to 10,000.
| Claim | Verdict | Grounds |
|---|---|---|
| “PAM was designed to predict terrorist events” | False | Every design document — including the ACM EC'03 paper written in June 2003, before the controversy, so not self-serving — describes quarterly geopolitical and economic indices. The words “assassinat”, “North Korea” and “special event” appear in zero lines of prose on any page of the preserved site. |
| “The assassination contracts were invented by critics” | False | The background screens deleted on 29 July survive in Hanson's own preserved archive. Contrast-stretched, one shows a panel headed “Special Event Securities” listing “Jordan King Overthrown 4th Q 03” (0.24), “N. Korea Missile Attack 4th Q 03” (0.15) and “Arafat Assassinated 1st Q 04” (0.17); a second shows all four at 0.23 — matching Daschle's reading of “23 cents” exactly — plus a global index panel with a standing contract labelled “Terror”, under the DARPA, Net Exchange and Economist Intelligence Unit logos. |
| “The senators invented the terror framing” | False | It originated with the Pentagon. Appendix B, p.B-8 of DoD's own 20 May 2003 report to Congress: FutureMAP “will answer predictive questions such as ‘Will terrorists attack Israel with bioweapons in the next year?’… The new approach is to set up, as it were, a ‘market’ in two kinds of futures contracts: One pays $1 if an attack takes place; the other pays $1 if there is no attack.” That is where the bioweapons example in the senators' letter came from. |
| “Decorative afterthought by someone else” | Substantially, not wholly | Hanson's 2007 account: the items were in “a small (less than 2 percent) section of two such screens”, were Polk's, and were unapproved per Poindexter. All true. But they were PAM's own published marketing, the category they illustrated was real and open-ended, the in-text worked example was a regime overthrow, and a standing global contract was labelled “Terror”. |
| “Congress banned terrorism futures markets” | False | P.L. 108-87 §8131 (26 Sep 2003) bars funds for the TIA Program and contains zero occurrences of “futures market”, “FutureMAP” or “Policy Analysis Market”. PAM was killed administratively on 29 July; the statute two months later closed the parent office. |
| “It cost $8 million” / “$750,000” | Neither | Both figures originate in the Wyden–Dorgan letter and are senatorial characterisations. Per Hanson, $750,000 was a planned Phase II ceiling and $8m an unfunded request. Use DoD's own answer for the record: $619,750 actually spent. |
PAM was designed to trade quarterly indices of economic health, civil stability and military disposition for eight Middle Eastern countries, plus global aggregates, plus an open-ended third category of nominated specific events. It was not designed as a terrorism or assassination market. But its own promotional screens illustrated that third category with an Arafat assassination, a North Korean missile attack and a Jordanian coup, and DoD's own report to Congress had already described the parent programme as a market in contracts paying “$1 if an attack takes place.”
Hanson's position is also narrower than it is usually rendered. In “Designing Real Terrorism Futures” (Public Choice, 2006) he argues affirmatively for genuine terrorism-event markets while maintaining that “PAM was not in fact designed to be terrorism futures”. He is not saying such markets would be wrong; he is saying this one was not that.
In 2008 the CFTC issued a Concept Release on the Appropriate Regulatory Treatment of Event Contracts — the Commission still dates its own use of the term “event contract” to that document. Two years later Congress legislated, and the text it produced is the hinge of everything that followed.
The language first appears in the Dodd–Lincoln Substitute of 29 April 2010; it is absent from the 15 April Dodd draft. Dodd-Frank §745(b), enacted 21 July 2010, added the Special Rule at CEA §5c(c)(5)(C), under which the Commission
…may determine that such agreements, contracts, or transactions are contrary to the public interest if [they] involve — (I) activity that is unlawful under any Federal or State law; (II) terrorism; (III) assassination; (IV) war; (V) gaming; or (VI) other similar activity determined by the Commission, by rule or regulation, to be contrary to the public interest.
That is the DARPA episode written into the United States Code. Seven years after a background JPEG cost an admiral his job, one of the five enumerated activities in the statute governing event contracts is assassination — and the sector it governs today is a $20bn-a-year industry in sports and election contracts.
The only legislative history is a Feinstein–Lincoln floor colloquy of 15 July 2010. Lincoln, whom the CFTC identifies as the Special Rule's author, said the purpose was “to prevent the creation of futures and swaps markets that would allow citizens to profit from devastating events and also prevent gambling through futures markets.” The CFTC's own 2026 reading is that the two senators “appear to be talking about two different types of derivatives contracts”, and that the colloquy does not show an intent to revive the pre-CFMA economic-purpose test. The Commission also concedes that the clause's cross-reference to “section 1a(2)(i)” is “a typographical or drafting error” — no such section exists.
Regulation 40.11, adopted 27 July 2011, implemented it: §40.11(a)(1) flatly bars listing event contracts involving the enumerated activities, with a 90-day review under §40.11(c). The Commission expressly declined a comment asking it to define “gaming” — which is the word it then used to kill everything for the next twelve years.
| Date | What the Commission did with “gaming” |
|---|---|
| 2 Apr 2012 | The Nadex Order prohibits political event contracts. Reasoning, per the pin cites in the 2026 rulemaking: the contract is judged “considered as a whole”; the legislative history showed Congress intended to restore the economic purpose test; “gaming” means gambling, which is linked to betting on elections; the contracts “cannot reasonably be expected to be used for hedging”; and they “could be used in a way that could potentially adversely affect the integrity of elections.” |
| Dec 2020 – Mar 2021 | The near-miss on sports. ErisX self-certified NFL contracts for eligible contract participants hedging sportsbook exposure; the CFTC opened a review; ErisX withdrew one day before the 90-day period expired. Then-Commissioner Quintenz later disclosed that staff had drafted an order prohibiting them. |
| 22 Sep 2023 | The Kalshi Order finds the congressional-control contracts involve gaming and activity unlawful under State law, and are contrary to the public interest. “Involve” is read to capture both contracts whose underlying activity is an enumerated activity and those with “a different connection” to one; chamber control “does not, in and of itself, have sufficiently direct economic consequences” to support hedging; and the contracts “could incentivize the spread of misinformation.” The vote was not 3–2: Behnam supported, Mersinger dissented, and Pham abstained — citing the Fifth Circuit's finding in Clarke v. CFTC that the Commission had violated its PredictIt injunction and “only narrowly avoided sanctions… because of the mercy of the court.” No document states the arithmetic. |
The thesis was simple and, on its own terms, correct: every prediction market that had ever mattered was killed by a regulator rather than by a bug, so build one with no company to sue and no server to seize.
Founded in 2014 by Jack Peterson, Joey Krug and Jeremy Gardner — Gardner is consistently named as the third co-founder and is usually omitted. The whitepaper went to arXiv in January 2015. Its advisers included Ron Bernstein, Intrade's founding CEO, and Vitalik Buterin, per Bloomberg's 2018 reporting; Augur's own press kit also listed Robin Hanson. That adviser roster is largely sourced to Augur's own material, and this report says so rather than treating it as independent.
The REP crowdsale ran 17 August to 1 October 2015 — 45 days, 11 million REP, 80% sold publicly, about 4,800 accounts at roughly $1,100 average, hitting $1.3m in the first 26 hours. The total genuinely conflicts across sources from $5.09m to $5.3m, almost certainly BTC/ETH conversion-date variance over a 45-day sale, and this report gives the range rather than picking one. Forbes retrospectively calls it Ethereum's first ICO. Beta shipped on Microsoft Azure in March 2016; v1 mainnet launched 9 July 2018.
The assassination markets were real, and nobody traded them. By 23 July 2018 the Trump market had acquired 50.3 shares. On 24 July Trustnodes could not find the market in Augur's own front end and saw no bets at all. MIT Technology Review and Bloomberg covered it on the 26th. On the 31st Forbes reported roughly $1.53m staked across 800-plus bets platform-wide in the whole first month — which is not death-market volume. The press cycle vastly outweighed the trading.
The Forecast Foundation's position, quoted by Forbes, was that it “has no power to censor, restrict, or curate markets, orders, trades, positions or resolutions on the Augur protocol contracts”, and Krug's was that “the Augur protocol is nothing more than open source software.” That is the censorship-resistance thesis working exactly as designed, and it is also the thing that put the word “assassination” back in the headlines for the second time in fifteen years.
Then the users left. The Economist, 9 August 2018: daily users fell from 265 in early July to 37 on 8 August.
Augur v2 launched 28 July 2020 — DAI-denominated, 24-hour resolution, a tradeable “Invalid” outcome, and US and UK residents excluded. Augur Turbo followed on Polygon in May 2021, sports-focused. By mid-2021 v2 was averaging about $35,000 a day while Polymarket held roughly 90% of all decentralised prediction-market volume. There is no dated shutdown statement; the best marker for the Foundation's exit is a 17 November 2021 post, and what the “most decisive step toward decentralization” it announced actually was could not be verified. “Augur died” is also wrong: a new steward, the Lituus Foundation, announced a reboot in March 2025 and reports a completed fork mechanism in August 2026 — the roadmap conceding in its own words that “REP was once listed across nearly all major centralized exchanges — only Kraken remains today”. That reboot is almost entirely project self-reported.
Founded in 2015 inside ConsenSys by Martin Köppelmann and Stefan George (employees three and four), with Friederike Ernst also named a co-founder in Gnosis's own retrospective, and an original mission “to build trustless, decentralized prediction markets”. The GNO Dutch auction of 24 April 2017 raised $12.5m and cleared in under eleven minutes — and the top three contributions were 50% of the funds raised, executed through “BiddingRing” contracts. Share of supply sold and clearing price both conflict between sources, and the widely cited ~$300m implied valuation is arithmetic rather than a sourced figure.
Gnosis dates its own move away from prediction markets as a core product to 2017. Its prediction-market front end, Omen, was Gnosis-originated and DXdao-operated, and its cumulative volume was about $6.4m by July 2021 — roughly $15,000 a day. What survived instead was infrastructure: Gnosis Protocol, Gnosis Chain, CoW Protocol, and Safe. And the Conditional Tokens Framework, which is Gnosis's most consequential legacy for this story, because per Gnosis's own account Polymarket is built on it.
This is the structural point the whole section exists to make.
| Polymarket | Kalshi | |
|---|---|---|
| Founded | June 2020, Shayne Coplan (b. 14 June 1998, so 22 at launch, not 21), as Blockratize, Inc., New York. Pivoted from his earlier Union Market. | 2019, Tarek Mansour and Luana Lopes Lara, who met at MIT; Y Combinator Winter 2019 batch. |
| Regulatory strategy | Launch first. CFTC settlement order 3 Jan 2022, Docket 22-09: off-exchange event-based binary options without DCM designation, $1.4m penalty, and an order to wind down all non-compliant markets. More than 900 event markets since inception. | Register first. Order of Designation as a Designated Contract Market issued 3 November 2020. $30m Series A led by Sequoia Feb 2021, with Charles Schwab and Henry Kravis. Public launch July 2021. |
| How it got into US election markets | It didn't, for the 2024 cycle — and still took over $3.5bn on election day offshore, per Dune. The “Théo” whale made roughly $47m profit on about $45m staked across four accounts, confirmed by Polymarket to CNBC: “My intent is just making money.” | By suing. Election contracts went live days after the D.C. Circuit refused a stay; about $430m on the presidential race and over $700m platform-wide for the cycle. |
| The low point | 13 November 2024: the FBI searches Coplan's Manhattan apartment and seizes his phone and electronics. No arrest, no charges. Polymarket called it “obvious political retribution by the outgoing administration… for providing a market that correctly called the 2024 presidential election.” Both the DOJ and CFTC probes closed with no charges in July 2025, and neither agency announced it publicly. | March–April 2025: cease-and-desist orders from Nevada, Montana, New Jersey, Ohio and Maryland over sports contracts, and the start of the 142-matter litigation campaign in section 07. |
| How it got back / scaled | It bought a registration. July 2025: acquires QCX LLC (a DCM) and QC Clearing LLC (a DCO) for $112m. CFTC Staff Letter 25-28 (2 Sep 2025) gives reporting relief only — not a designation, not an approval, and expressly drawing no conclusion about the contracts' legality, though Polymarket's CEO characterised it publicly as clearance to go live. ICE invests up to $2bn on 7 October 2025 at roughly $8bn pre-money. | It kept winning. Robinhood distribution from March 2025, later driving roughly half its volume; $185m at $2bn (June 2025); $300m at $5bn (Oct 2025); $1bn at $11bn (Nov–Dec 2025); $1bn Series F at $22bn led by Coatue, 7 May 2026. A Form D filing of 26 August 2026 shows $1.12bn raised since April 2026 — the best primary evidence of 2026 scale. |
| Where it stands | Feb 2026: over $7bn monthly volume and 679,000 monthly actives. March 2026 peak figures conflict badly across trackers, from $10.15bn to $25.7bn; this report does not pick one. | The larger venue by volume through 2026 — in the week of 18 May it took about 70.8% of combined weekly notional. Reuters reported advanced talks at a $40bn valuation on 30 September 2026; not closed and not announced as of this compilation. |
The CFTC is also policing these markets while defending them. Four 2026 actions: a US service member charged with insider trading in Maduro-related contracts (23 April); a Google employee charged over search-result contracts (27 May); George Santos ordered to pay $35,000 for manipulative trading of a State-of-the-Union contract (31 July); and Gabriel Perez ordered to pay $172,000 for insider trading of “mention market” contracts (28 August).
Kalshi filed its APA challenge on 1 November 2023. What followed is the most consequential run of litigation in the subject's history, and it is still unresolved.
This is the part that makes the whole history legible. Michael Selig was sworn in as CFTC Chairman on 22 December 2025, and since that day the Commission has had exactly one member. In February 2026 it withdrew the Biden-era proposed rule, Selig describing it as the previous administration's “frolic into merit regulation”. On 17 February 2026 it filed as amicus for Kalshi's side in the Ninth Circuit. From 2 April to 23 June 2026 it became a plaintiff against the states — Arizona, Connecticut, Illinois, New York, Wisconsin, Minnesota, Rhode Island, New Mexico, Kentucky — obtaining a TRO blocking Arizona criminal enforcement on 10 April. Selig: “The CFTC will continue to safeguard its exclusive regulatory authority… against overzealous state regulators.” Connecticut Attorney General William Tong: “These contracts are plainly unlicensed illegal gambling.”
And in the live rulemaking of 10/12 June 2026, Prediction Markets; Public Interest Determinations (91 FR 35806), the Commission states in a footnote that “the reasoning in a Commission order in 2012 prohibiting certain political event contracts was incorrect”, and that the Special Rule shows no congressional intent to apply the pre-CFMA economic-purpose test. It proposes at last to define “gaming” — the word it declined to define in 2011 — and reads “may determine” to mean prohibition now requires an affirmative, individualised written determination rather than operating automatically. It concedes the consequence in its own words: “some event contracts that are contrary to the public interest may be traded during the period of time required for the Commission's review.”
It remains a proposed rule. The comment deadline was 27 July 2026 and 1,459 comments were filed; no final rule had issued as of 23 October 2026. There are no commissioner statements or dissents to cite, because Selig is the sole member.
At the Supreme Court: three petitions, none granted. Flaherty v. KalshiEX, No. 26-299, was filed 2 September 2026 and docketed on the 8th, with the response due extended to 9 November 2026 — Sundeep Iyer for New Jersey, Neal Katyal for Kalshi. Amici supporting certiorari include Ohio and 38 other states and DC, 145 Indian tribes and tribal organizations, the NFL, former Senator Christopher Dodd, former CFTC Chairman Gary Gensler, and the American Gaming Association. Also docketed: Robinhood Derivatives v. Dreitzer, No. 26-338. As of 23 October 2026 a litigation tracker counts 142 matters across 28 states and DC and 30 appellate dockets. The newest front, opened 16 October 2026, is QCX LLC d/b/a Polymarket US v. Wilson in the Southern District of Ohio — the first to add a dormant Commerce Clause theory.
Federal legislation: none enacted. Bills in committee in the 119th Congress include the Prediction Markets Are Gambling Act, the Prediction Markets Security and Integrity Act and the Prediction Market Act of 2026. The only measures to move are self-dealing bans: S. Res. 708 was agreed to by the Senate on 30 April 2026, barring Senators and staff from prediction-market contracts.
Everything above came from outside the archive, because the archive could not have supplied it. What the archive supplies is different and in places irreplaceable: people in rooms, before the press releases.
The headline archive result is a priority claim, and it is one of the few in this corpus that comes back unprovisional.
$ ./corpus.py first intrade
FIRST MENTION IN THE INDEXED CORPUS - 2012-11-30 BJF (Brother John F - The Bitcoin Report)
Bitcoin Report Volume 26 (Bitcoin and Intrade)Brother John F is the only pre-2013 voice in this archive, and Volume 26 went out on 30 November 2012 — four days after the CFTC filed against Intrade. It is a full editorial on the case. It quotes John Stossel via an article by Jeff Berwick, it correctly identifies the jurisdictional absurdity at the heart of the complaint, and then it does the thing that makes it worth a section of its own.
“…now we know that the c_f_t_c_ the whole thing is a joke because there is no uh… delivery of anything on in trade… it's not the c_f_t_c_ is a commodity futures trading uh… commission so they regulate the trading of commodity futures there are no commodities delivered on in trade the in trade does not deliver anything these are simply predictive bets about future events so to say that a future event whether gold is going to be above two thousand dollars on a certain date or below it to say that that is a regulated option on commodities is ridiculous and absurd”
c_f_t_c_, u_s_) and renders Intrade as two words; both are preserved rather than silently corrected.And then, having worked out that the regulator had removed a market that was serving its customers well, he tells the audience what to do about it:
“now the big story related to that of course is going to be the bitcoin market that's why this is so important for bitcoin now we have a site i covered on the blog bets of bitcoin and uh… i thought of forming my own but honestly i don't have time so but uh… you can imagine there will be plenty of sites popping up where you can do exactly what you did on in trade uh… using bitcoin so my suggestion would be that if any of you have uh… a big mailing list of u_s_ intrade customers that you would mail them a link to the bitcoin client… and suggest that they uh… continue their business with uh… the bitcoin instead which is completely legal there's nothing illegal about using bitcoins to bet”
That is 30 November 2012. Augur would not be founded for about two years, its whitepaper was two years and two months away, and its token sale two years and nine months away. Gnosis did not exist. Polymarket's founder was fourteen. And the entire logic of the crypto prediction-market sector — the regulator has removed a market people wanted, so the replacement will be built on Bitcoin because that one is legal and cannot be removed — is stated on air, as a prediction, four days after the complaint that triggered it.
The claim is unprovisional. corpus.py returns
“No unindexed show could affect this answer. It is as firm as the archive gets”
for 2012 dates, because nothing held-but-unindexed reaches back past 2013.
The honest limit is what the archive is, not what was broadcast: this is the
earliest mention in sixteen searchable shows, which in 2012 means essentially one show. It
is not a claim about who covered Intrade first in the world.
237 broadcasts, 2013-09-26 to 2026-08-29, across ten shows, from
corpus.py spread --any '"prediction market" "prediction markets"'. The shaded bars are
the years in which the subject was theoretical. The two vertical markers are the events that made it
real — and the archive is quietest at both.
Ten shows carry the subject. The order of adoption, and where each one stopped:
| # | Show | First | +days | Last | B’casts | Then silent for |
|---|---|---|---|---|---|---|
| 1 | LTB Network | 2013-09-26 | 0 | 2020-03-02 | 23 | 14 d (show ran to 2020-03-16) |
| 2 | Let’s Talk Bitcoin | 2013-09-28 | +2 | 2014-07-22 | 5 | 200 d (show ran to 2015-02-07) |
| 3 | The Bitcoin Talk Show | 2014-04-10 | +196 | 2018-08-18 | 4 | 1,012 d |
| 4 | Mad Bitcoins | 2014-06-22 | +269 | 2026-01-31 | 5 | — |
| 5 | World Crypto Network | 2014-07-03 | +280 | 2026-01-31 | 51 | 246 d |
| 6 | Epicenter | 2014-07-12 | +289 | 2026-07-03 | 99 | 68 d |
| 7 | SF Bitcoin Meetup | 2014-12-16 | +446 | 2017-09-27 | 9 | 972 d |
| 8 | The Bitcoin Game | 2015-04-16 | +567 | 2019-07-18 | 3 | 2,311 d |
| 9 | The Bitcoin Group | 2015-06-26 | +638 | 2026-08-29 | 34 | 48 d |
| 10 | Speaking of Bitcoin | 2019-05-26 | +2,068 | 2021-12-22 | 4 | 241 d |
| Not in this table: Bitcoin Uncensored, which holds 18 prediction-market broadcasts and cannot be placed in it, because none of its rows carries a date (§01). Brother John F holds the Intrade episode above and never uses the phrase “prediction market” at all. | ||||||
The very first phrase hit is not a prediction-market story at all — it is Daniel Larimer pitching Keyhotee on the LTB Network on 26 September 2013, listing prediction markets among the things a blockchain could do: “You can do prediction markets. You can revolutionize so many different industries.” The identical passage appears two days later on Let's Talk Bitcoin #44, because the network republished it; that is one broadcast counted in two places and the table shows it as two, which is a limit of per-show counting rather than a finding.
219 broadcasts mention Augur — 189 dated and 30 undated. That figure is
only reachable through the alias group: Whisper writes auger 235 times against augur 41, so the correct
spelling finds under a sixth of the material, and auger is also a drill bit, which is
why every hit in this section was read before it was counted.
The earliest dated Augur mention in the corpus is not an interview or a news item. It is the open-mic announcements slot at a meetup at Geekdom in San Francisco, 16 December 2014 — before the whitepaper went to arXiv, eight months before the token sale. “Dated” is load-bearing. 30 of the 219 Augur broadcasts are undated Bitcoin Uncensored rows, two of them Joey Krug interviews, and any of them could in principle be earlier. This is a first-mention claim about the dated corpus, not about 50 million words.
repetition_loop flag is not set on this row.“Hello. Hi. My name is Jeremy Gardner. I am part of a team called Auger. You can check it out. It's Auger. It's Auger. Decentralized prediction markets. Got a lot of potential. Really fun. I'm not going to say anything more right now. But I really encourage you to check out our white paper and our prototype, which you can download.”
Earlier in the same evening a lawyer on the regulatory panel — working through which
Bitcoin 2.0 businesses look like regulated activity — reaches for prediction markets as his
example and looks around the room for him: “For example, with the law in the 2.0 check,
which is a big interest of mine, we have, say, prediction markets. Is Jeremy here with
Auger?” That line is also what the phrase query
"prediction market" matches first for this show; the speaker is not established and is
not named here.
This is where Thomas's recollection — “also friend Jeremy and Joey from Augur etc. knew all the augur guys” — stops being recollection and becomes record.
| Date | Show | What it is |
|---|---|---|
| 2015-05-29 | The Bitcoin Game #18 | Rob Mitchell runs Gardner's twenty-minute Texas Bitcoin Conference talk in full, introducing him as co-founder of the College Cryptocurrency Network and “Director of Operations” at Augur. no episode URL established — see §13 |
| 2015-06-21 | LTB Network — Let's Talk Bitcoin! #222, Understanding Augur | Matthew Zipkin interviews Tony Sakich and Joey Krug for a full episode. Recorded 16 June 2015. watch |
| 2015-08-15 | Mad Bitcoins — CoinCongress 2015 | Gardner's own presentation, uploaded twice to Thomas's channel: a14Npzc1XFM and fS_lONNEQ4w. Two days before the REP sale opened. |
| 2015-08-21 | The Bitcoin Group #77 | Augur is a billed story of the week: “Bitcoin Civil War — Augur — Flash Crash — Jeb vs. Encryption”. watch · transcript |
| 2015-08-27 | Money & Tech | Ethereum-Based Augur and the Wisdom of the Crowd. watch |
| 2015-09-24 | World Crypto Network — The Bit Panel | Gardner on Thomas's own network as a panellist, billed in the title: “Amazon Discounts, Prediction Markets, Crypto for Artists and Weed”, alongside Tatiana Moroz. watch (re-uploaded 2015-10-02 as e5wwv3_2YrE) |
| 2016-07-01 | World Crypto Network — #Blocktalk | A full hour with Tony Sakich, Augur's marketing lead, as guest of honour. watch |
“My name is Jeremy Gardner. My first kind of survey to the Bitcoin space was as the co-founder of the College of Tripp and Currency Network, which is a global educational nonprofit, comprised of Bitcoin clubs all over the world. I then went on to co-found the model, which is a decentralized prediction market platform. … And then lastly, I worked as an entrepreneur and residents at blockchain capital.”
Worth noting against the external record: the subagent research found Gardner's tenure at Augur dated inconsistently — his own site says Director of Operations 2014–2017, while Crunchbase gives August 2014 to October 2015. The archive does not settle the end date, but it does independently confirm he held the title in May 2015 (The Bitcoin Game #18's introduction) and was describing the role in the past tense by September 2015 (above, “I then went on to co-found… and then lastly…”).
Augur's single most consequential engineering decision is explained by its co-founder in a podcast interview in June 2015, three years before launch. This is the kind of thing the archive has and nothing else does.
“So the difference between a prediction market and gambling are, so with a prediction market, they're designed to elicit information. So a prediction market is a very general financial market. Any other financial market can be implemented as a subset of a prediction market. … And it turns out, according to a bunch of research, that you actually get really good odds estimates for future events, more accurate than almost anything else. They beat polls, they beat pundits, even sometimes they beat based in statistics.”
“I would bet all my money on the Ethereum side chain. So what we looked at is we looked at building it on Bitcoin. We actually started building on Bitcoin. But we discovered that it was really complicated to add things to Bitcoin.”
Two turns later the same interview produces something that no secondary source establishes independently. Robin Hanson — section 02's Robin Hanson — is named on air as an Augur adviser, in June 2015. Outside this archive that fact rests almost entirely on Augur's own press kit; here it is a named team member saying it in an interview, with a date on it.
“Then there are some that have, you know, joined up. us as advisors, like Robin Hansen, who really is, you know, the original thought leader of this entire scene, if you want to call it that… He's a researcher. He actually coined the term idea futures. I believe it was in the early 90s. And I'm actually going to let Joey talk a bit about him, too… But he was actually a part of that whole cypherpunk scene in the early to mid-90s. He posted on the mailing list, and he's also a professor at, I believe it's George Mason University”
And the hand-off lands, which makes the next turn the one quote in this episode that reaches tier A — and it reaches past Hanson to the economics the whole subject descends from:
“Yeah, Hanson's pretty much the father of prediction markets. Guy he has first proposed by Hayek in like the 40s, I think. And then Hanson picked it up again in the 90s and really ignited the deal of academic research on it.”
A Bitcoin Uncensored interview — undated, and therefore absent from every chronology in this report — has Krug saying the same thing in the opposite direction, to Krista Rose of the Counterparty Foundation at Inside Bitcoins New York: “Yeah, so we're using Ethereum. And so what we do is we have smart contracts on Ethereum. … And the way Bitcoin comes in is it'll be sidechained to the Bitcoin blockchain.” The sidechain never happened.
Thomas's network gave Augur's communications lead a full hour on #Blocktalk five months before the 2016 US election, and the whole episode is about the launch date.
“Obviously, auger is the center of what I'm doing, but I'm more than happy to talk about whatever else is out there. … We're aiming for fall. We would like to have it out by the presidential election. The full release or everyone can use it with crypto. But you know, one of those things with what just happened with the Dao. We want to make sure that our security audits are as thorough as possible.”
Augur v1 shipped on 9 July 2018 — two years and one presidential election after that. The DAO hack is cited in the same breath as the reason, which is accurate: the security audits were real and the delay was, in engineering terms, the right call. It is also the decade's central irony in one sentence. The platform built to be unstoppable was too careful to ship, and when it did ship, the first thing it was famous for was the markets it could not stop.
Epicenter — invisible to every previous report on this estate — did the interviews that nobody else did, four of them inside eleven months:
| Date | Episode | Why it matters |
|---|---|---|
| 2015-09-21 | Paul Sztorc: Truthcoin & Prediction Markets | Truthcoin, later renamed Hivemind, is designed as a Bitcoin sidechain rather than on Ethereum. On the relationship to Augur this report defers to Augur's own account, quoted below: a named Augur guest on LTB #222 calls Truthcoin the closest of its predecessors, says Sztorc “didn't really have time to implement it fully”, and says Augur “picked it up in August” — which is a stronger and better-dated source than any secondary characterisation. no episode URL established — see §13 |
| 2015-09-28 | Robin Hanson: Futarchy | The man from section 02, in the crypto corpus, one week later. Only two shows in the whole archive use the word futarchy at all — Epicenter, 4 broadcasts, and The Bitcoin Game #68 — plus one undated Bitcoin Uncensored row. no episode URL established — see §13 |
| 2016-07-11 | Martin Köppelmann: Gnosis | 114 separate mentions of Gnosis inside this one episode — the densest single broadcast on any term in this report. 114 is a count of mentions, not of broadcasts; the 124 in §12 is a count of broadcasts, and mixing the two units is the mistake this row previously made. no episode URL could be established — see §13 |
| 2016-08-09 | Joey Krug: Augur | The origin story in his own words, 69 mentions. no episode URL established — see §13 |
“Yeah, so, so like the idea for Auger is kind of, it's really kind of a combination of like three different ideas proposed by other people. We just kind of combine them in an interesting way. So there was a paper in 2014 called Decentralizing Predictor Markets and Limit Order Books. And that was paper written by some guys at Princeton. And it was basically the idea to make a decentralized prediction market. And so, like our order book mechanism, the mechanism for issuing shares, things like that, all came from there.”
unknown and may_serve_media: false; it is quotable under the archive owner's 15 Oct 2026 editorial decision and its media is not served.“Today we have a special guest joining us from Washington. His name is Dr. Robin Hansen. And for those of you who don't know him, we could think of him approximately as the Nick Sabo of the prediction market world. He has played a role in the technology of prediction markets for the past 20 years, developed some of the main inventions that underlie modern prediction markets…”
“A prediction market is really just a betting market, but we give it a different name…”
One last piece of corroboration for the personal connection, from a different part of the same
archive. The Facebook reaction export holds 53 separate likes on Jeremy Gardner's posts and
photos, 47 on Joey Krug's, and 11 on Augur's own page — and none at all on
Robin Hanson's, which is the sort of null that tells you the first three are about people he
knew rather than about the subject. The Twitter archive holds Thomas's
retweet of World Crypto Network's post of the Jim Bell interview, tagged
#hcpp17 and mentioning @Paralelni_polis — which is how this report
establishes where that interview was filmed.
Those are social-graph records, not evidence of friendship, and the report does not stretch them. What they do establish is sustained attention over years to a specific person, which is consistent with the recollection and is the sort of thing that normally leaves no trace at all.
This was asked as a question, not asserted as a fact — “I think they knw or one of their guys went o to make calshi or the other oen” — and it is answerable. The answer is no, and the real overlap is more interesting than a founder link would have been.
No founder link exists. Kalshi was founded in 2019 by Tarek Mansour and Luana Lopes Lara, who met at MIT and had worked in quantitative finance (Goldman Sachs, Citadel, Bridgewater among the firms named). Polymarket was founded in June 2020 by Shayne Coplan, who had previously run a DeFi project called Union Market and pivoted it to prediction markets. No Augur co-founder, contributor, Forecast Foundation employee or Augur adviser is documented as having founded, joined or been employed by either company.
And there is no cap-table overlap to find on the Augur side, because Augur never had a cap table: the 2015 REP sale was a public crowdsale of about 4,800 accounts, not a venture round. Every connection that does exist is therefore personal rather than institutional.
| Person | Augur role | Later | Verdict |
|---|---|---|---|
| Joey Krug | Co-founder and core developer, 2014–. Thiel Fellow (2016). | Pantera Capital co-CIO from 2017; departure announced to LPs 3 Feb 2023; joined Peter Thiel's Founders Fund as a partner in April 2023, where he remains. | The real link — as an investor |
| Vitalik Buterin | Adviser and early financial backer, named by CoinDesk in March 2015 and by Bloomberg in July 2018. | Participated in Polymarket's May 2024 round; its most prominent public defender, framing prediction markets as “info finance” and a “social epistemic tool”; also its public critic on drift toward sports and short-term price bets. | Adviser → investor |
| Jeremy Gardner | Co-founder, Director of Operations. | Claims on his own site to have written Polymarket's first check, via Ausum Ventures, listed at 234× realized. | Self-reported only — not independently confirmed; he is not named in Forbes's coverage of the 2020 seed or in Coplan's own disclosed investor list |
| Ron Bernstein | Adviser from 2015. | Intrade's founding CEO; built Tradesports as a real-money fantasy-sports exchange (launched April 2014, reportedly closed 30 Nov 2015); took another run at a sports-betting exchange in 2019. | He connects Intrade → Augur, not Augur → Kalshi/Polymarket |
| Paul Sztorc | Truthcoin/Hivemind designer. CoinGecko's write-up describes him as having advised the Augur developers while believing their drift from the original Truthcoin design would be the project's downfall; that characterisation is CoinGecko's reporting of his views, is several years old, and is attributed to them rather than stated here as fact. | Authored BIP-300/301 (Drivechain). Per Protos, “neither Truthcoin nor its tokens exist on any mainnets”. | No documented connection to either |
| Jack Peterson, Tom Kysar, Peronet Despeignes | Co-founder/lead scientist; head of operations at launch; board member. | — | No documented connection to either |
The strongest and best-sourced link is Krug, and it is worth stating precisely because it is startling without being a founder link. The co-founder of Augur is the Founders Fund partner who announced that firm's lead of Polymarket's $45m Series B in May 2024:
“Thrilled to announce @foundersfund's investment in @Polymarket… This marks our first consumer crypto investment and is a truly zero-to-one company.” — Joey Krug, May 2024 · x.com/joeykrug/status/1790412838072123510
And on the ICE deal in October 2025: “Earlier this year @foundersfund doubled down on Polymarket and led a $150M round from both our venture and growth funds, our first crypto growth investment since @bridge__harris and I joined.” There is a Thiel thread running through the whole thing: Krug's Augur work was funded by a Thiel Fellowship, and he now invests from Thiel's Founders Fund, which is Polymarket's largest venture backer.
It will not say Krug has a board seat or advisory role at Polymarket. One wiki lists him as a board observer or adviser; that is an unsourced wiki, Crunchbase does not list him among Polymarket's board members or advisers, and the claim is treated as unestablished. His documented relationship is investor and public advocate.
It will not say Gardner wrote Polymarket's first check. He says so; nobody else does. A looser chain exists — he also claims the first check into Polychain, which led Polymarket's 2020 seed — but that is an inference and not a documented Augur→Polymarket link.
It will not say Kalshi has any Augur connection. Its disclosed backers — Y Combinator, Sequoia, Charles Schwab, Henry Kravis, SV Angel, Neo, Paradigm, a16z, Coatue — contain nobody from the Augur era, and its named leadership and hires (Eli Levine, Rebecca Schick, Sarah Slane, regulatory lead Jeffrey Bandman, ex-CFTC) include nobody from Augur. No documented connection at any level: founders, staff, advisers or cap table.
It will not treat “both were in crypto” as a connection. That is the inference this section was commissioned to avoid.
What is a real structural inheritance, and is usually missed: Polymarket is built on Gnosis's Conditional Tokens Framework, per Gnosis's own account. So the lineage from the 2015–17 Ethereum prediction-market generation to the 2020s winner is real — it just runs through the code rather than through the people.
This section needs care rather than avoidance, so the frame comes first.
The Assassination Market was a real darknet site, launched in late 2013, run pseudonymously as “Kuwabatake Sanjuro”, which crowdfunded bounties on named public officials in Bitcoin. It was widely reported at the time, condemned at the time, and is a notorious episode in early crypto rather than a neutral product launch. The idea behind it comes from a 1995–96 essay, Assassination Politics, by Jim Bell.
What follows is documented history: what the site was, how it was covered, who said what on which date. It contains no operational detail — nothing that would help anyone build, fund or run such a scheme — and the quotations are selected for historical and journalistic value only. Contemporaneous news reporting quoted below does state what the 2013 site publicly claimed to have raised, because that is the fact the coverage turned on; it describes no mechanism. Passages in the archive that amount to mechanism are described rather than quoted, and that is a deliberate editorial limit, stated here so a reader knows the omission is a choice.
38 broadcasts mention assassination markets, across eight shows. The first is Thomas's own, and it arrives within days of the story breaking.
“Sinister new site assassination market enables users to contribute Bitcoins for the murder of US officials. Just in time for the holidays and fortunately too late to get discussed on yesterday's congressional hearing… We get this assassination market embarrassing to Bitcoin you bet. Something that's already been offered in cash and written about since 1992 also true.”
Two things about that passage are worth stopping on. It is the first assassination-market mention in the entire indexed corpus, on 19 November 2013. And in the same breath as calling it embarrassing, he places it in a prior intellectual lineage — “written about since 1992”. 1992 is his own figure and is roughly rather than exactly right for the Bell essay, which circulated from 1995; it is close to the Hanson/Extropy idea-futures literature of 1992. The report does not adopt the date, but the instinct — that this was an old idea with a new payment rail — is correct, and he had it on the day.
Six months later, Money & Tech's daily news read gives it a full segment, and gives the operator's own self-justification:
“A site called the assassination market has raised $125,000 in Bitcoin donations for the assassination of a public figure. Launched late last year, the site is a crowdfunding platform for political assassinations. … Site creator Sanjuro says the site is targeted at, quote, people who are outside the reach of the law because it has been subverted and corrupted, and whose victims have no other way to take revenge than to do so anonymously.”
mixed — quotable and linkable, may_serve_media: false, so no media is served here.The phrase then recurs for twelve more years as a reductio — the thing people reach for when they want to say that a prediction market has no natural floor. A sample, all verified present: WCN's Dark News and an Ethereum law-and-contracts panel in 2014; The Bitcoin Talk Show #22 in August 2014; The Bitcoin Group #108 in 2016 and #143 in 2017, where a panellist's formulation is “are we going to kill all of everyone with assassination markets? I say assassination markets already exist”; WCN's Bitcoin reading group on Rothbard in 2019; Epicenter's tenth-anniversary stream in 2023. All of those are multi-voice panels and none of the speakers is named here.
And on 2 March 2020 the LTB Network interviewed the journalist Brian Merchant about the darknet's fake murder-for-hire marketplaces, in an episode that frames the whole thing as prediction markets taken to the limit: “…of like the most extreme outcome of a prediction market, which prediction markets are, you know, they're talked about frequently.” That is the archive's last substantial engagement with the subject, and it closes the circle on an idea it first mentioned in 2013.
Thomas's recollection was “I interviewed a guy who created assasination markets”. The archive has the interview. It is not Sanjuro.
Interview with Jim Bell, Author of Assassination Politics, World Crypto
Network, 7 October 2017, video id 4KLqi1v2zSE — a single-guest
interview, 46,979 characters of transcript, roughly an hour. The host's own Twitter archive carries a
retweet of World Crypto Network's post of it, tagged #hcpp17 and mentioning
@Paralelni_polis, which places the recording at Hackers Congress
Paralelní Polis in Prague.
Jim Bell is the man who originated the concept — the author of the multi-part essay Assassination Politics, written from 1995. He is not Kuwabatake Sanjuro, who built the 2013 site. That is a real distinction and it makes the interview more historically valuable, not less: this is the primary source for the idea rather than for one implementation of it.
Sanjuro is not in the archive as an interviewee anywhere.
corpus.py probe kuwabatake returns zero across 50,432,534 words;
sanjuro returns exactly one broadcast, the Money & Tech news segment quoted above.
Had the recollection been of a Sanjuro interview, this report would say the archive does not contain
it.
What Bell says about his own essay, in the opening minutes:
“Well, provocatively, I named it assassination politics. And while provocative, it was a very descriptive. Instead of having 51 percent of the people decide who is going to be a politician. Let's have a much smaller number of people decide who will never be a politician ever, ever again.”
“I wrote it in and 10 parts later. It's become the second most famous internet essay ever by Ted Tizansky's essay. Certainly, the Unibomber manifesto. I tried to read it myself when it was published in the New York Times.”
“So I wrote this up in the form of an essay after considering the numerous issues. And it became extremely controversial. And I've been hated ever since then by the federal government of the United States and probably more a few other governments around the world, although they haven't really spoken up yet.”
Unprompted, about twelve minutes in, Bell connects his essay to the DARPA episode. This is the originator of the idea describing the Policy Analysis Market as a government attempt at his own concept — and he is sharper about the mechanism of its death than most of the literature:
“He did it because the federal government of the United States had composed something that was suspiciously similar to my API idea. And it was shouted down within a day by the news media who called it terrorism futures. It was a bad case of PR problems right there. By PR I mean public relations.”
And then, near the end, the passage that is the single most valuable thing in this interview — because no secondary source carries it. Bell gives a first-person account of how the 2013 site became a global news story. He says he introduced its operator to the reporter.
“Well, if you go back to October and November of 2013, when San Juro announced his assassination market, well, within about three weeks, the price of midpoint had gone up from about $200 to $1100. Now, a lot of time that rise occurred before San Juro even contacted me, which he did with one email. … I didn't know this in the time, but I directed San Juro to a journalist named Andy Greenberg. Much to my surprise, a week later, Andy Greenberg written an article for Forbes about this idea. And I was even more surprised when about a day later, by that time, 100, at least, other journalists had either duplicated the article or had written the version of the wrong around the world. It was like an explosion of publicity.”
Per CONVENTIONS, an interview is not an endorsement, and what the host does with the material belongs in the report. He pushes back twice. The first is a historical analogy, offered to Bell's face:
“So I am starting to feel a little echoes of the French Revolution here where they lined up all of the kings and so forth and sent them to the guillotine, which was a brand new killing machine recently invented.”
The second is to name the cost out loud and make the author own it. He frames the proposal to Bell's face as “basically like a violent version of wisdom of the crowds”, and later puts the consequence into the question itself — that “the initial phases of the market may lead to a small or relatively medium sized bloodbath” — and invites Bell to agree, which he does, before the interviewer offers him “we can't make an omelet without breaking a few eggs” and Bell takes it. Both of those phrases are the interviewer's, not the guest's, and that is worth stating plainly rather than letting the attribution blur: it is the interviewer who puts the body count into the conversation and the author of the idea who accepts it as a feature of the design. The exchange is summarised here rather than quoted at length, because the guest's extended justification is the part this report has decided not to reproduce. Nothing in it is operational.
Bell was promoting a fundraising website for the idea at the time of the interview, which he names and spells out on air. This report does not link it.
Substance: yes. Person: no. He interviewed the author of the essay that invented assassination markets, not the operator of the site that implemented one. The interview exists, it is an hour long, it is on his own network, it is dated, and it is one of very few recorded long-form interviews with Bell anywhere. It is also a better interview than the one he remembered conducting, because its subject supplies the connective tissue between 1995, 2003 and 2013 in a single sitting — including one link that appears to exist in no other source.
The last finding is about absence, which is the hardest kind to report honestly, because a null result in a corpus is indistinguishable from a null result in the world unless you say which you mean. Here it is clearly the corpus.
| Term | Dated b’casts | + BU, undated | Archive-wide | First dated | Last dated |
|---|---|---|---|---|---|
| “prediction market(s)” | 237 | 23 | 260 | 2013-09-26 — LTBX, Daniel Larimer on Keyhotee | 2026-08-29 |
Augur (incl. auger) | 189 | 30 | 219 | 2014-12-16 — SFBM, Gardner introducing himself | 2026-03-28 |
| Gnosis | 121 | 3 | 124 | 2016-07-11 — EPI, Köppelmann. The tool says 2014-10-21; that is a different Gnosis — see below. | 2025-04-29 |
| “assassination market(s)” | 38 | 3 | 41 | 2013-11-19 — MB, Thomas's own news read | 2025-09-20 |
| Polymarket | 18 | 0 | 18 | 2021-01-12 — EPI host extravaganza | 2026-05-10 |
| “robin hanson” | 3 | 6 | 9 | 2015-09-28 — EPI, futarchy episode. Two thirds of this term's hits are undated. | — |
| futarchy | 5 | 1 | 6 | 2015-09-28 — EPI | 2019-07-18 |
| Intrade | 4 | 1 | 5 | 2012-11-30 — BJF Volume 26 | 2019-07-18 |
| “idea futures” | 2 | 0 | 2 | EPI 1, LTBX 1 | — |
| Kalshi | 1 | 0 | 1 | 2026-08-29 — The Bitcoin Group #502 | 2026-08-29 |
| Kuwabatake | 0 | 0 | 0 | — | — |
The 2014 “first Gnosis mention” is a different Gnosis.
corpus.py first gnosis returns WCN's Dark News #030 of 21 October 2014, and the
snippet is “a big part of that news was this post from Gnosis on October 19th this was
posted in the Bitcoin talk…” — a Bitcointalk username, nine months before
Gnosis the company existed. The project's first appearance is Epicenter, 11 July
2016, and that episode alone carries 114 mentions. This is exactly the
failure CORPUS.md §5 warns about: the tool returned a real hit and the inference from it would
have been false.
“Intrade 70” is “in trade”. The split-form probe returns 70 broadcasts and almost all of them are the ordinary English words. The real figure is four, established by reading snippets.
“Augur 235” includes drill bits. auger is a real English
word. 189 is the de-duplicated broadcast figure from the alias group with the low-confidence flag
attached, and individual hits in the live lane were read before being counted.
In 50,432,534 words across sixteen shows and fifteen years, Kalshi is named once. It is The Bitcoin Group #502, 29 August 2026, and the panel is not talking about Kalshi — they are talking about Paul Sztorc, whose Twitter alias is truthcoin, and whether he should launch his own coin. Kalshi arrives as the punchline.
“…he has been obsessed with prediction markets for quite a while. His alias on Twitter is truth coin, which to me sounds like a prediction market right there. Ironically, he's already been passed by Kalshi and other companies that are running ahead with prediction markets, not needing their own coin, not needing their own chain, having figured out some way to do it on their own. So that's interesting.”
live:MB_20260829_9kbFc0Im2dM) and carries a repetition_loop flag, which does not affect this passage or the date. No TBG mirror transcript page exists for #502 yet, so no second link is given.That is the whole thesis of section 06, spoken by somebody who was not trying to make a point about it. Eleven years of coverage of blockchain prediction markets in this archive, and the single reference to the company that won is a sentence noting that it won without any of the things the archive spent eleven years covering.
There is one more trace, and it is the closest the archive comes to a direct answer to the question that prompted this report. On 7 April 2026 a WCN clip discusses Nevada's attempt to ban prediction markets:
short:WCN_20260407_MSFgQqzDPs8), machine-titled and auto-cut from a longer broadcast. Per CONVENTIONS it is not cited as an episode: it is excluded from every count and date in this report, and appears here only as a quotation. The parent broadcast was not identified. “Kalishi” in the clip is Whisper for Kalshi; the alias probe does not catch it, so the Kalshi count of 1 above is an episode count and this is a 486-character clip outside it.“What do you think about Nevada trying to ban the prediction market, Kalishi and prediction market sports betting in general? I fail to see the real difference between a prediction market and just usual gambling, other than, as you said, maybe the odds are even worse in one of those. Yeah, it sounds like a desperate move to try to remove competition. It's been interesting to watch these prediction markets grow from kind of theoretical ideas to what my friends were building on auger.”
And one methodological note worth passing to the next session on this subject: the single most useful document for the whole 2008–2026 regulatory thread is the CFTC's own June 2026 NPRM, whose background section is a fully footnoted official history of this line with pin cites — including the agency's explicit repudiation of its own 2012 reasoning. If any part of sections 05 to 07 needs deepening, that is where to go.
Two kinds of material went into this report and they are kept apart here, because the separation is the report's spine. Archive is the local transcript corpus and the host's own records — sections 08 to 12. Outside sources are primary documents and press — sections 02 to 07, plus section 10's answer to the connection question. Nothing from the second pile supports a claim about what was said on air; nothing from the first supports a claim about law.
Link status. No URL below was constructed or guessed, and anything that could not be resolved says so in its row.
verified fetched during this compilation and resolved. cited the URL is the one a source document itself carries, taken over unchanged but not independently fetched here. dead checked and found unavailable; kept so a reader can see what is missing. local only the archive holds no URL for this item and none could be established.
On the Epicenter and Bitcoin Game rows, because “local archive only” would have been
the wrong claim. Both shows publish publicly and both are podcast-feed shows, so their episodes
have no YouTube id for corpus.py to emit. Epicenter's source_id encodes a
slug — EPI_20160809_httpsepicentertv143, i.e. epicenter.tv/143 —
so a URL looked recoverable. It is not. epicenter.tv/143,
/episodes/143 and /podcast/143 all 404, and
epicenter.tv/episode/143 returns HTTP 200 with the title “Episode Not
Found” — a soft 404, which is exactly the trap CONVENTIONS §4 warns about when it
says an HTTP status is not a liveness test. The same soft 404 comes back for 97, 98, 139 and 642.
So no URL was constructed, and these rows say “no episode URL established” rather
than asserting that none exists, because the episodes are almost certainly reachable somewhere and
this compilation could not find where.
What the check actually returned, 23 October 2026. All 222 external URLs on this page were fetched sequentially with a browser user-agent and a 25-second timeout. 171 returned 200. 49 returned 401, 403 or 429 — NYT, WSJ, Bloomberg, Forbes, The Block, congress.gov and others that refuse automated requests, plus cftc.gov and kalshi.com rate-limiting a run of hits from one address; all six rate-limited cftc.gov URLs returned 200 on a spaced re-check, which is why they are marked verified and the remainder are marked cited. Per CONVENTIONS §4, an HTTP status is not a liveness test on these hosts and a 403 from nytimes.com is not a dead link. Three URLs are genuinely unavailable and are marked in place: two 404s (Cointelegraph, Jones Day) and one that will not connect at all (the Washington Post's copy of the Poindexter letter). Nothing was silently dropped and no replacement URL was constructed.
mixed, quote and link only, media not servedgranted-reported (Ryan Taylor, via Thomas)unknown, quotable, media not servedgranted-reported (Adam B. Levine, via Thomas)These three Bitcoin Uncensored items are quoted or referenced above and carry no date
(§01). They are the most on-topic interviews in the archive and they appear in no chronology in
this report. Rights: unknown, quotable under the archive owner's 15 Oct 2026 editorial
decision, may_serve_media: false.
#hcpp17, mentioning @WorldCryptoNet and @Paralelni_polis. This is what establishes where the interview was filmed.facebook_reaction — 53 likes on Jeremy Gardner's posts and photos, 47 on Joey Krug's, 11 on Augur's page, 0 on Robin Hanson's. Local archive only; Facebook reaction records have no public permalinks. local only~/brain/brain.db — 4.2 GB, opened read-only. Measured 2026-10-23: 50,432,534 words, 7,047 distinct broadcasts, 5,032 hours, 16 of 18 registered shows searchable. local only_scripts/corpus/corpus.py — first, spread, probe, search, stats, rights. All commands used are listed in §01. local only_scripts/corpus/aliases.json — the Augur group (augur / auger) is what makes 189 broadcasts findable rather than 41. local onlyreports/CONVENTIONS.md, reports/CORPUS.md and reports/PRIORITY-METHOD.md — the house rules this report follows. Markdown in /reports/ is excluded from deploy.sh rsync and does not ship, so these are deliberately not linked: a link would 404. local onlyPAM/website.zip) is where the background screens and pam_trade_info.htm survive/images/ autoindex verified — emptied 29-Jul-2003 12:33Per CONVENTIONS §2 these are recorded rather than omitted or guessed at. §12 lists the archive-side gaps; these are the outside ones.