Stephen Grugett is cofounder of Manifold Markets, a platform where anyone can create and participate in user-created prediction markets using a play money currency called Manifold Dollars. The company raised a $2 million seed round and received an ACX grant from Scott Alexander. The core idea is that prediction markets aggregate dispersed knowledge into calibrated probabilities, and Manifold aims to make this accessible to everyone, not just financial professionals.
How Manifold Works
Users receive 1,000 Manifold Dollars for free upon signup and can bet on any market or create their own on any question they care about.
The platform uses a dynamic parimutual market mechanism (based on a quadratic cost function, similar in spirit to Hanson’s Log Market Scoring Rule) to determine prices and payouts.
Markets are user-created and user-resolved: the person who creates a market also acts as the judge of its outcome, which is a deliberate design choice that trades off some fraud risk for massive scalability.
A small fee is charged on trades: 4% goes to the market creator as an incentive to create and resolve markets, and 1% is burned (a liquidity fee that subsidizes the market).
Users can buy additional Manifold Dollars with real money, though the platform currently operates entirely with play money.
Why Play Money Can Work
Grugett’s hypothesis is that people are driven more by status, competitiveness, and reputation than by greed. The leaderboard and public track record of predictions serve as proof of expertise.
Top predictors on the platform are genuinely skilled — the person at the top of the leaderboard made a large profit betting on the Russian invasion of Ukraine in February 2022.
There is significant overlap between top predictors and financial professionals who manage money or trade for a living, countering the criticism that prediction markets are just a hobby.
The platform plans to host tournaments with real cash prizes and has a real-money crypto product on the back burner, but separated from the main play-money experience.
Why Companies Don’t Use Internal Prediction Markets
Many major companies (Google, GM, CIA) have tried internal prediction markets and ultimately abandoned them.
The primary reason is managerial resistance: managers often don’t want objective feedback that casts doubt on their chosen direction, even if the information would be net beneficial. The mere existence of a market expressing skepticism can lower the probability of a mission’s success.
Prediction markets are more clearly valuable in corporate settings for market research and informational purposes — surveying consumer behavior, competitor actions — where they don’t directly challenge management authority.
Usability is another barrier: it’s hard to build a product simple enough for all employees to use without much thought while still yielding reasonable results. Manifold’s broader mission is to lower this bar.
Long-Term and Existential Questions
For questions that resolve far in the future (e.g., catastrophic AI by 2030), prediction markets face two problems: temporal discounting (people care less about distant outcomes) and the apocalypse problem (you can’t collect winnings if the world ends).
One workaround is to break long-term questions into short-term proxy variables (e.g., benchmark results for AI, polling numbers for political candidates) that are more actionable for speculators.
Brian Kaplan and Eliezer Yudkowsky have structured a real bet on AI catastrophe where Kaplan pays Yudkowsky now, and Yudkowsky pays Kaplan more in 2030 if catastrophe hasn’t occurred — sidestepping the collectability problem.
Insider Trading
Grugett holds Matt Levine’s view: insider trading should be understood as taking from shareholders rather than as a fairness issue.
While insider trading does improve price efficiency, society has other interests — shareholders profiting from non-public information is undesirable.
For congressional insider trading, Grugett is firmly against it: elected representatives should be compensated through transparent salaries, not through roundabout corruption via trading on privileged information.
The “Worse is Better” Design Philosophy
Manifold’s key innovation is embracing user-resolved markets, which sounds risky (opening the door to fraud) but in practice works because users choose which markets to participate in and mostly make good decisions.
This mirrors the “worse is better” philosophy from Unix and Bitcoin: simple, usable mechanisms that handle the core case well are more valuable than theoretically perfect systems that are too complex to adopt.
Grugett and his cofounders (including his brother James) came to this from their previous startup Throne, a subscription group chat app for online creators, where they learned the importance of simplicity in both business model and user interface.
Concerns About Manipulation and Whales
A potential concern is bot abuse: someone could create many accounts, make random trades, and promote the one account with the best track record to the top of the leaderboard. Manifold has anti-abuse measures in place but is not committed to the free signup bonus indefinitely.
As the platform grows, domain-specific leaderboards (e.g., geopolitical predictors, economic thinkers) become more meaningful than a global ranking, since many markets are personal or between friends.
For whales buying large amounts of Manifold Dollars: Grugett is pro-efficient-markets and argues that if a whale bets heavily on the wrong side, they simply subsidize the people who are correct — the market mechanism naturally solves this problem.
He has toyed with demurrage (negative interest rates on uninvested cash balances, e.g., 20% per year) to encourage active participation, but notes users hate losing money for any reason, making it impractical.
Manifold’s Own Use of Prediction Markets
The team actively dogfoods their own platform, creating internal markets on company milestones (e.g., whether they’d complete their fundraising round by end of April — the market stayed above 85%).
Their first major case of a market substantively informing a business decision was on whether to sell play money for real money. The market thought they should and would keep the feature, which factored into their decision to proceed.
Exotic Uses Users Have Discovered
Users have hacked the platform for purposes beyond prediction: lotteries, games (e.g., Manifold Plays Wordle, where markets were created on which word to guess next), and research assistance.
Free response markets allow open-ended questions where users submit answers alongside bets, and the market creator picks the winning answer — enabling things like crowdsourced research conclusions with supporting arguments in the comments.
Background and Hiring
Grugett studied computer science, worked at Susquehanna International Group (an options trading firm), and then at a robo-advisor startup writing portfolio optimization software. All three cofounders are technical and have prior entrepreneurial experience.
Manifold is hiring for full-stack developers (React experience preferred), a community manager (Discord, blog, social media, events), and a head of growth (experience scaling startups to ~$1M ARR or ~50K monthly active users). Contact: [email protected] or [email protected].
Vision for the Future
Grugett envisions a world where prediction markets are embedded in news articles, blog posts, and broadcasts, immediately grounding public discourse in calibrated probabilities rather than pure speculation or ideologically biased thinking.
He acknowledges the pessimistic take that many people consume politics not to understand the world but for other reasons, but believes at least a portion of news consumption is genuinely aimed at understanding, and prediction markets can serve that portion.
