Last week, @Genzy brought us a limited-time run of doubled unique trader bonuses! This is a great time to make new markets and recoup some of those costs. I went ahead and made a market I had been thinking about recently, and I’ve already received M390 in trader bonuses. And don’t forget: you can make one small market for free every week via the quest bonuses!

Genzy says he’ll probably turn it off on Monday, but the markets think he might leave it on a bit longer. You can also bet on how much extra mana will be distributed as part of this promo here:

New Sort & Filter

So many markets are basically decided already, resting at 90-95% for months before resolution. Boring! How do you find the interesting, uncertain ones that haven’t been settled yet?

Starting this week, you can use the new options for finding markets far from the extremes. The Uncertain filter option only shows binary markets with probabilities between 30-70%, which is very nice when combined with a sort like 24-hour volume. And the new Mid % sort option orders all markets by how close they are to 50%, which is amazing when paired with filters like liquidity and close date.

Block Yourself

That’s right, you can now block yourself from trading in your own market.

Why would you want to do this? Well, if you had a market about something really subjective, you might want to say, “To remain unbiased, I promise not to hold any shares in this market.” But traders may still be concerned that you’ll swoop in at the last second, place a huge bet, and misresolve the market to make a ton of profit.

Now there’s a button on the market info card that lets you publicly prevent yourself from trading on that market. You’ll see a confirmation box that explains that it can’t be undone except by a moderator or admin, and if you continue, you’ll see a nice green box on your market that tells everyone you’ve committed.

If you still have any shares or profit in the market, that’s fine; it’ll be shown in that box for users to check. I think it will generally be best practice to sell all shares before hitting this button, but I decided not to make it force a liquidation since there are still valid reasons for users to hold shares, but pledge not to buy or sell any more.

I hope this becomes a signal for users to find high-quality, trustworthy markets more easily and reduce opportunities for the bad kind of shenanigans. Feel free to request that creators use this before you bet in a market!

Colors

For multiple-choice questions, you can now select any color by using a hex code!

For a while, you’ve been able to pick colors for each answer of a multiple-choice market from a list, but this week we added a field where you can enter any hex code to use a fully-custom color. Now you can use real NVIDIA Green (#76b900) or Claude Orange (#d97757) for your markets! As a bonus, we also applied these to the special 2-answer MC/binary markets, so you can now customize the colors of those markets the same way.

Personalized Mana Offers

After your merch order ships, you’ll get a notification (and a pulsing badge on the Get Mana button) for a time-limited offer: up to 30% off your next purchase of M$5,000. The offer lives for 72 hours, shows a countdown on the offer card, and works with both card and crypto checkout. If you don’t want it, hit “Dismiss”. The chip stays around so you can un-dismiss if you change your mind; it doesn’t pause the timer.

If you have had difficulty ordering merch, try again! We have made several improvements and fixed a few bugs that were blocking some people.

Referral Payouts

Referral payouts are fixed and now split into two events. A regression from early April was silently blocking new referral payments, so referrers got nothing when their referred user signed up. Now:

• M$250 when the referred user places their first non-API bet

• M$1,000 when they complete iDenfy verification

So you get something even if your referral never finishes verification, and the extra M$1,000 when they do. The Referrals page now shows the actual amount paid per referral plus badges: “Verify pending” while the verify leg is still outstanding, “N× boost” if your supporter multiplier applied, and “?” for referrals that haven’t earned yet. A backfill script has been run for users whose referrals fell through the bug window.

We are constantly taking steps to detect abuse, but of course, if you notice anything suspicious, please let us know. Using referrals to create your own alt accounts will lead to fines or further punishment — don’t do it!

Trading Log UI

Have you been annoyed at seeing dozens (or hundreds) of tiny limit orders filling up the trade logs? Bots use these types of orders to manage their positions, but it can be difficult to sift through and understand what’s going on. This week, we pushed a few features to help you humans understand the trade log better.

First, we added a button to hide all trades made via the API. This will hide all bot trades and anything else not made directly from the official website or apps. If you would rather just filter out the tiny orders, you can use the older Amount option to see trades above, say, M$100.

As a bonus, I’ve also added those filters to the Trades tab of the live page. I love having that page open on a second monitor to just watch trades flow by like a waterfall (and definitely not so I can snipe-fade random bets).

We also simplified the text of each trade in the log a bit. Now you can easily distinguish the orders that actually executed; those trades will clearly indicate how much was actually spent versus the maximum. All orders now also include a consistent field that shows if it’s open, filled, expired, or cancelled.

Miscellaneous

• You’ll see time zones displayed in a few more places, namely the market page info box and market creation page, where you set the close time. We always used your local timezone for these, but it was never super clear. You can now also set a specific close time on the market creation page.

• The balance log got a little love this week. For those of you who place tons of bets, it was getting a little hard to navigate, so we added pagination in addition to the date filters to help you track your balance better. When you go to search, we load the next 20 pages to make sure we find the transaction you’re looking for (though we should search the server instead).

• We increased the padding of some buttons, such as the “back” button in the top left of the post page. We also made some other buttons more consistent, like the DM and settings buttons on the profile page. Consistency!

• We did some work to try to speed up the prize page loading this week. I don’t actually think this fixed it, though. Someone should make a market about this.

• I decided to change the default date market type from buckets (where you bet on individual dates or ranges) to thresholds (where you bet on if it happens before/after a date). Both have their uses, but I think thresholds are usually more intuitive for most situations. Yell at me in the comments if you think I’m wrong!

• Last time I mentioned a new API validation for bot comments, checking that replyToCommentId exists. Our database expects that this is the ID of the top-level comment since our comment threads aren’t deeply nested. The API now enforces this instead of silently eating your comment.

• Errors for “Maker has insufficient balance” should be much less likely. When you place a bet on a dependent multiple-choice market, we now check limit orders on other answers and cancel them if they would be filled, and the user doesn’t have enough mana to back it.

• We added a document outlining our responsible security disclosure procedures to the GitHub repository. The document is just a recap of our existing procedure, now in a better location.

• The stats page now uses query navigation between tabs so you can link to a particular one.

• The cool sprout hat we give to new users was falling off their heads sometimes, so we added some superglue to make sure it stays perfectly centered over their avatar.

Thanks for reading! For those of you in the US, have a nice Memorial Day weekend!

-wasabipesto

Fields Medalist Tim Gowers wrote the companion paper:

“If a human had written the paper and submitted it to the Annals of Mathematics and I had been asked for a quick opinion, I would have recommended acceptance without any hesitation.”

OpenAI got this wrong once already. Thomas Bloom, who maintains the Erdős Problems website, called an October 2025 OpenAI claim “a sharp misrepresentation.” Bloom is present alongside this result. Gowers wrote a companion paper. I think the excitement is warranted.

Markets give 79% odds that AI will solve any important mathematical conjecture before January 1st, 2030.

Whether yesterday’s result resolves that market depends on how the creator reads “important” : Erdős Problems are about as prestigious as it gets in combinatorics, and Gowers is saying Annals without hesitation.

First Proof

In February, eleven mathematicians from Harvard, Yale, Stanford, MIT, and EPFL published a challenge they called “First Proof”: ten unpublished research lemmas drawn from their own ongoing work, available to AI systems for one week before the answers were released. None of the problems had appeared in any training dataset and that was seemingly the point.

Without restrictions, GPT-5.2 Pro and Gemini 3.0 Deep Think produced confident proofs for all ten problems. Two held up under expert review. Scientific American ran the headline: “The verdict, it seems, is in: artificial intelligence is not about to replace mathematicians.”

That was February 14th. The Erdős result was announced May 20th.

I don’t think these are actually in tension. First Proof tested whether AI could reliably solve a broad, uncurated sample of research-level problems with no direction, no particular reason to expect the problems to fall within whatever the model is capable of. The Erdős result was a specific, famous, intensively studied problem surrounded by decades of mathematical literature. The model wasn’t working blind: it had encountered every approach anyone had tried across its training data. Producing convincing nonsense across a diverse range of fields is a different failure mode from finding an unexpected connection within a domain saturated with relevant signalling.

Markets give 54% odds that an AI will co-author a mathematics research paper published in a reputable journal before the end of 2026, up 10 points recently.

Whether that market is pricing AI capability or academic gatekeeping is rather unclear. I’d also be curious what that market would price if the question specified “without meaningful human scaffolding,” because OpenAI’s First Proof submission involved a weeklong sprint of human-AI collaboration with errors already found in the result. And whether that’s co-authorship or heavily assisted human authorship is a also a question the market hasn’t quite settled.

The First Proof team has announced a second batch of problems. I think this is the more interesting benchmark to watch right now, partly because it’s adversarial in the right way and partly because the second batch will presumably be designed with knowledge of how AI systems approached the first.

The IMO

Markets give 70% odds on a perfect score being achieved by an AI model at the 2026 IMO. I think this is roughly right, maybe slightly generous.

In 2025, both OpenAI and DeepMind achieved gold medals at 35/42. A perfect score is 42/42. Problems 3 and 6 are categorically harder than the rest of the paper. Fewer than 10% of human contestants solve them, and twenty-six humans scored better than either AI last year. The gap between gold and perfect is not a numerical gap. Gold means you solved five of six problems that the world’s best high school mathematicians spent 4.5 hours on. Perfect means you solved them all, including the two that stumped 90% of those students.

The bull case is that DeepMind could train a specialized system specifically targeting IMO 2026, as they did with AlphaProof and AlphaGeometry. Specialized systems have historically outperformed general models on structured competition math by substantial margins. The bear case is that Problems 3 and 6 require genuinely novel proof techniques, closer in character to First Proof than to the kind of structured reasoning current models handle well.

The open-source version sits at 44%. The labs most likely to build a specialized system for IMO 2026 are not the ones who open-source it afterward, and the 24-point gap is roughly the market's price for that constraint.

90 and 19

Two numbers sit oddly against each other in this cluster.

90% on whether AI will produce an Annals-quality paper for $100k by March 2030. Gowers just said he would send the Erdős proof to the Annals without hesitation. This market is up 24 points recently and feels directionally right, perhaps even slightly conservative.

19% on whether AI will outperform humans in all mathematical research areas by 2028,.

I don’t think these are inconsistent, though the gap is real. Producing an Annals-quality paper on a specific famous problem and outperforming human mathematicians as a practice are different things. The Erdős conjecture had been worked on for nearly eighty years - it came preloaded with context, partial results, and a research community that had already significantly narrowed the solution space. Knowing which problems are worth that kind of attention, before anyone else does, is also a significant part of the profession.

As someone who’s yet to have a strong background in mathematics, I lack the ability to truly appreciate the scale of the achievement, but I found Gowers’s comments in the full companion paper to be perhaps the most representative take on where the mathematics community truly stands (among those who haven’t got their heads too far up their own arses to seriously ponder the question).

I find myself not only trying to assess what AI has achieved in this particular case, but also thinking more generally about how such assessments can possibly be made. Can we still identify some mathematical capability that human mathematicians have and AI does not yet have? If so, what might that capability be, and how could one go about demonstrating that AI still lacks it?

Almost certainly the answer to the first question will have to be quantitative rather than qualitative. That is, we are unlikely to be able to show that there is something we can do that current AI models cannot in principle do at all, but we might be able to show that there are things we can still do much more efficiently than those models. But when a model has just solved a major open problem, it is clear that even a modest conclusion like that will not be straightforward to demonstrate, and indeed isn’t obviously true.

I think (for what little my perspective may be worth on assessing a Fields medalist on his carefully selected beliefs on the notion), that this is the right frame of thought. Too often the question gets lost in categorically useless philosophical waffle that sounds profound and gets you precisely nowhere: whether the models are "really thinking," and whether they're "truly creative." None of that matters for working out what they can actually do. What matters is whether they can match or surpass human effort on the same problems, assessed quantitatively. Gowers then goes on to conclude:

In any case, there is no doubt that the solution to the unit-distance problem is a milestone in AI mathematics... Furthermore, even if it is correct that AI cannot yet find a proof that needs a long hint sequence, such proofs are very difficult to find for humans as well, so in the unlikely event that progress in AI mathematics does suddenly stall, we have still probably entered an era where it will become very difficult for humans to compete with AI at solving mathematical problems.

The 19% market is probably best read in this light. It isn’t pricing whether AI will be able to do specific impressive things (as that question is largely settled). But more so an attempt to understand whether the full practice of mathematical research, selection included, gets displaced. Gowers is saying even a stall in AI progress would leave us in an era where human competition with AI on problem-solving is very difficult.

The chess comparison is what most reach for, and there’s a market: 12% on research-level mathematics becoming a sport akin to chess before 2035.

Engines have been superhuman at chess since Deep Blue beat Kasparov in 1997. Chess didn’t die. Players use engines for preparation, opening theory expanded, the quality of human play improved. A 2025 PNAS paper found that AlphaZero’s internal representations contained concepts grandmasters could learn, machine knowledge transferring back to humans quantifiably. Whether mathematics follows the same path is unclear. There’s no Elo rating for being a mathematician, and “akin to chess” probably requires human mathematical practice to reorganise around AI tools the way chess preparation reorganised around engines, which is a higher bar than AI producing messy proofs on famous problems. I’d price 12% a few points higher, but I understand the hesitation.

The market on whether the next Millennium Problem will be solved by an AI lab or a mathematics department has had AI lab as favourite since January.

AI lab sits at 58%, and the small recent uptick looks like an underreaction given the week. The Millennium Problems are a different category from anything above: seven problems, $1 million prize each, one solved in history. Grigori Perelman proved the Poincaré conjecture in 2003 and turned down the million dollars and the Fields Medal. If a model cracks the Riemann Hypothesis, presumably someone will be available to collect.

Happy Forecasting!

- Above the Fold

On one hand, most professional forecasters agree with mainstream health organizations that hantavirus is not as serious of a threat as the media has made it out to be. Top forecaster Peter Wildeford wrote an excellent piece on this called “Hantavirus won’t be the next COVID,” breaking down exactly that. Primarily, the epidemiology of the disease does not lend itself to pandemic potential:

…critically, the infectious window of hantavirus is short and centered on visible illness. Andes virus infectiousness peaks on the same day a patient develops fever and drops sharply within days. Unlike COVID, pre-symptomatic transmission is minimal. By the time someone is contagious enough to seed a superspreading event, they look sick with a fever — which is why isolation works at all. This is the profile of a virus public health can contain.

On the other hand, many other pundits agree with the cottage industry of blogger virologists that hantavirus is quite worrisome because of what it says about how the world is coordinating around risky outbreaks. If hantavirus had slightly higher human-to-human spread, we could be in a very different world. There still might be a couple passengers from the ship unaccounted for, and over the last week there still have been breaches in quarantine in certain nations that would be problematic in a disease with a slightly higher reproduction number.

Matt Yglesias, in his piece titled “Hantavirus is a reminder we should prepare for the next pandemic,” writes:

I checked with knowledgeable medical professionals and none of them seem especially worried about this particular hantavirus outbreak, which is great. But when I start to see headlines telling people not to panic, I get worried.

Not because you should panic about hantavirus, but because panicking is — by definition — not the correct response to any situation. If you’re facing an imminent threat, you still shouldn’t panic because panicking is never correct. So by the same token, while it’s good to hear that the objective threat level from this outbreak is by most accounts lower than I initially feared, just issuing instructions not to panic doesn’t add very much.

The implication is that everyone should just continue going about their day without changing anything. And it’s possible that’s correct advice for you and your family.

But my sense is that most American households have in fact under-reacted to the threat of novel virus outbreaks (which is perhaps why companies making well-reviewed respirators go out of business). And it’s definitely the case that the American government has under-reacted.

I think both of these takes are essentially correct. Individual American citizens almost certainly shouldn’t panic about hantavirus, but world health agencies absolutely should be jumping the gun to throw any resources they can at stopping things like the current hantavirus outbreak, especially in the first few days when its etiology and transmission dynamics were more hazy. Markets hold that there’s nearly a 1 in 3 chance of a global pandemic in the next 3.5 years! If there’s even a whiff of something that has pandemic potential, they shouldn’t hesitate to throw billions of dollars of resources at an outbreak. The base rate is roughly 10% per year, with the last two being COVID-19 and H1N1, otherwise known as swine flu.

Hantavirus will probably not be this pandemic, however. The Manifold market on a hantavirus pandemic this year has consistently been trading below 5%, several points below its Polymarket mirror.

Manifold has a lower discount rate than Polymarket since the uses of Mana off the platform are considerably lower than the use of dollars off Polymarket. Even with this, the market might be overselling the likelihood of a Pandemic, which I’d personally price at around 1% or lower. Take a look at the largest shareholders:

Brad, the largest NO holder, is 9th in profit all-time on Manifold, and users Chad, Agh, and yours truly also make the profit leaderboard (you might also notice YouTuber “Rational Animations” as the third largest YES holder, “Writer”). No one on the YES side can say the same.

Getting a market down below 3-5% with asymmetric profit incentives can be challenging. When you see headlines like “prediction markets give X (thing happening in many months) a 3/5/10% chance,” this is something to keep in mind as an interpretative artifact!

Traders think there’s only about a 1 in 2 chance that even a single non-passenger will contract hantavirus from the MV Hondius outbreak. Surprisingly, the outbreak has still been limited to passengers and crew on the ship.

I’m not sure why this market spiked recently from where it was trading at around 30%, but perhaps it has to do with the MV Hondius finally docking at port in the Netherlands.

A couple people have raised the possibility that the hantavirus outbreak on the ship didn’t necessarily exhibit any human-to-human spread, with passengers on the ship perhaps all being contaminated by excreta of a rat or rats on the ship (the disease is spread by the “colilargo,” otherwise known as the long-tailed pygmy rice rat). Traders think this thesis — or rather, that convincing evidence supporting this thesis will emerge — is unlikely, at just 12%.

It still may take a few weeks for the outbreak to run its course if it does continue to spread. A market on the last confirmed case gives >20% to a case after June 18th, for example:

Are scientists busy working on a vaccine? Possibly! After some volatile price action, it seems that traders believe there’s around a 30% chance that a hantavirus vaccine is trialed in humans by the end of next year.

An OSTP report found that it would only cost in the ballpark of $26 billion to prepare mRNA vaccines against each of the known 26 families of viruses that can infect humans. That sounds like a lot but it would be pennies on the dollar if they were effective against even a single future pandemic, possibly saving the US trillions in avoiding shutdowns and lowering healthcare expenditures alone, through expeditious vaccine development.

It seems quite plausible that health agencies or international funders take this hantavirus outbreak seriously enough to preemptively do at least a Phase 1 trial in humans.

Despite low chance of a pandemic, this hantavirus outbreak could still be bad! There’s a marginally higher chance that the WHO could label it as a Public Health Emergency of International Concern if things get worse, and it seems distinctly possible that public outcry or pressure might lead them to do this, counterintuitively, regardless of the actual epidemiological risks.

Keep in mind that there’s a base rate of about a 1 in 3 chance of any PHEIC in a given year, as these things happen fairly regularly. In fact, they happen so regularly that while writing the draft of this newsletter, one happened, kind of out of the blue!

The recent ebola outbreak may have now eclipsed hantavirus! A few rumored cases of ebola in the Democratic Republic of the Congo rapidly turned into a fully-fledged outbreak over the weekend, with the WHO scrambling to account for why this happened under the radar. Perhaps health agencies were focused on hantavirus? Perhaps the U.S. distancing themselves from global health organizations played a role? This variant of ebola is not the slightly more common Zaire strain, but rather the closely related Bundibugyo strain, which complicated early testing results, confusing global health workers and putting a slight delay in the response.

Ebola outbreaks happen every couple years, with a previous outbreak from 2021 raising the alarm of a lab leak, due to suspicious genomic similarity with a previous strain from several years prior.

Ultimately, it seems far more likely that the explanation is delayed spread from a survivor of the previous outbreak as an STD, which is known to occur, but it seems remotely possible that discarded old viral samples could have led to a reinfection.

This ebola outbreak has the potential to lead to high loss of life, which makes it quite surprising that it hasn’t captured the public’s imagination in the same way the hantavirus outbreak on a cruise ship has. It has similarly failed to capture Manifold market creators’ imaginations, thus far, and we invite users to create more markets to help the public track and understand this crisis.

Roundup

Manifold user Bandors (of FOLDED fame) has created a hantavirus tracking website which he is hoping will go viral. This market resolves YES if he gets 10k unique visitors within a month of creation. Traders think this may or may not happen. With the outbreak winding down, traders are skeptical:

And if you like longer term markets, you could trade on the possibility of a pandemic before 2030:

Happy Forecasting!

-Above the Fold

The basics: You get entries with mana. The more mana you put in, the more entries you get, with no purchase necessary. There’s also a free entry just for showing up.

• One free entry per drawing. No mana required, though you do need to have at least M1,000 invested in active predictions on the site.

• Get entries with mana. More mana = more entries! The exact rate follows a bonding curve, so early entrants get a bit of an edge to disincentivise last-minute sniping.

• Verifiable randomness. When a drawing closes, winners are picked using a Bitcoin block hash as the random seed. The outcome is publicly verifiable, and nobody (including us!) can rig it.

• Paid in USDC on the Ethereum network, sent directly to a wallet of your choice. More on payouts and eligibility below.

Prize Drawing #2 was our biggest yet: a $10,000 prize pool split across seven ranked winners.

Bayesian took the $9,000 top prize out of 510 unique entrants and 1.74M mana spent on entries. It was great to see a highly active user get rewarded, and we continue to bring prizes to many more members of the Manifold community. It really is anyone’s game.

$100 a day, every day

We’re currently running a $100 prize drawing every single day, so everyone gets more chances to win. We have already seen two repeat winners, including DottedCalculator’s $100 USDC win in drawing #8 with only 821 entries for M200 total (a 0.52% chance to win).

Before that, we also saw phenomist win the 7th place prize ($50) in the first-ever prize drawing, holding only 0.09% of total entries (1,664).

Anything can happen, so check in every day to claim your free entry.

Eligibility, Payouts, and Charity

The (small) catch: To participate and claim your prize, we all need to play by the rules. That means you need to be at least 18 years old and not based in an ineligible jurisdiction (which includes some US states and a few countries around the world). Check out the Prize Drawing FAQ for specifics on how to qualify (and how to avoid some things that may disqualify you).

Prizes are paid in USDC on Ethereum. After winning a prize, you’ll need to connect a wallet on the respective prize page where you won.

A couple of winners from Drawing #2 still haven’t submitted a wallet. We’ve held their prizes - no one’s been forfeited for being slow… yet.

If a winner is ineligible or declines their prize, it rolls into the Charity Giveaway, which has now grown to $1,550 with these additions. Traders can convert mana into entries to boost their favorite charity’s odds. When the giveaway ends, we’ll draw one lucky entry (using the same mechanism as the prize drawings) to determine the winning charity.

The Charity Giveaway also comes with extras: the Champion Trophy and the Champion’s Legacy profile background.

With giveaways, physical merch, Manifest tickets, Manifold subscriptions, digital goods, there’s never been a better time to have mana - or buy it!

For a Limited Time: Double Trader Bonuses

As we keep experimenting, one of our primary goals is to improve market visibility and delivery on the site. Content on Manifold has always been primarily community-created and driven, which is what makes the site so great.

We’d like to see a lot more markets for both popular and niche topics, and we know that means ensuring traders will see and predict on the questions that are important to you. We’ll be rolling out some updates to content delivery, so keep an eye out for more details.

While we work on those, we’re also doubling all trader bonuses for the next 10 days.

This means, for each liquidity tier:

• M100 = M6 per trader, 17 traders to break even

• M1,000 = M20 per trader, 50 traders to break even

• M10,000 = M30 per trader, 333 traders to break even

• M100,000 = M40, 2,500 traders to break even

This will be immediate and universal, so existing market creators should already see the mana flowing!

We’re also keeping an eye out for well-defined markets asking timely or important questions, and Manifold will add to the liquidity pool to incentivise traders.

→ Create a Market

The Golden Age of Mana

Still have mana burning a hole in your pocket and want something in hand to show for it? Alongside the digital benefits, we’ve launched a series of real-world benefits to reward the best creators and predictors:

• Manifest 2026 tickets — at the time of writing, there are 12 tickets left in the mana shop. Manifold subscribers get their usual shop discount, and mana can be purchased with crypto for a discount. Either option makes this the cheapest way to buy a standard Manifest ticket.

  Yes, that means you can buy the Manifold Pro subscription for M2,500 and save M3,375 on a Manifest ticket immediately.

  Already going to Manifest? Place your bets on the Manifest 2026 Prop Bets market. There’s a 78% chance of 10+ attendees with a Wikipedia page, 46% chance for Karaoke, and a 33% chance of someone in a full fursuit. (Have an idea for something not on this market? Add your own answers!)

• Manifold merch — grab some ahead of Manifest! Physical merch, delivered to your door. The merch shop launched very recently, and we’ve already had nearly 100 successful merch orders, with an average time-to-ship of under 2 days. Merch is also discounted by site subscription.

  

• Prize Drawing or Charity giveaways — as explained, win cash for yourself, or for the charity of your choice.

If there’s any other way you’d like to use your mana or you have ideas for fun contests, let us know! We want everyone to feel like Manifold is a good use of their time, so we can continue to be there when you want to see the future or bet on anything.

If you still haven’t made a mana purchase with crypto, you will receive +10% bonus mana on your first crypto purchase, and an additional +10% if you buy >$1,000.

*Don’t like crypto? No worries, we heard you: card payments are back!

Now, go take advantage of limited-time 2x trader bonuses, grab an entry (or a few!) to the daily prize drawing, support your favorite charity, browse the mana shop, and feel free to give us some suggestions and feedback along the way!

-Stef, Tod, and the Manifold Team

As closing arguments begin in Musk v. Altman, you might notice that the contours of the AI race have changed a lot in the last few years. About a decade ago, Elon Musk was the face of OpenAI and their biggest funder. Musk and Altman had organized OpenAI as a counterweight to Deepmind, after its acquisition by Google in 2015. By 2018, Musk had mostly disengaged from his involvement in the company, after a complicated internal struggle for leadership. It took another few years, though, for the animosity to reach a breaking point. After the launch of ChatGPT, and Microsoft’s subsequent investment in the company, Musk argues that he became convinced that the OpenAI leadership was out of alignment with its original goals as a non-profit. A more cynical interpretation is that Musk had already begun to launch his own private company, xAI, and was setting the groundwork for an attempt at competition.

Regardless of his motives, by early 2024, Musk had filed suit. This was re-filed, later in the year, with Microsoft added as a defendant. In November, he filed for an injunction to block OpenAI’s conversion to a public benefit corporation, but Judge Yvonne Rogers, who is currently presiding over Musk v. Altman, denied this injunction.

By late last year, the case had narrowed to two of its 26 original claims.

• Breach of charitable trust: the for-profit restructuring of OpenAI violated the trust of Musk’s original $38 million in donations.

• Unjust enrichment: OpenAI, Altman, and Brockman used Musk’s contributions for unauthorized commercial purposes which enriched them and their partners.

Last month, the trial began, with Musk seeking up to $150 billion in damages, which would go to OpenAI’s charitable arm, as well as the reversal of the conversion to a public benefit corporation and the removal of Altman and Brockman from leadership at OpenAI.

All these outcomes are still on the table.

Traders currently think there’s about a 1 in 3 chance of a full win by the defense, a 1 in 3 chance of Musk winning, within which a majority of scenarios would involve purely monetary relief of some kind, and a 1 in 3 chance of a settlement or some sort of procedural resolution.

The 12% chance of structural relief should give concern to OpenAI investors. Clawing back the conversion to a public benefit corporation would be unprecedented but would likely be quite painful for the company.

A lot of online commentary has treated a Musk win as kind of inevitable, and indeed, in 61% of scenarios, according to this market, there’s at least a settlement of some kind. But that settled amount could be quite low. Traders on a market with prop bets on the trial think that there’s only a 29% chance that the total damages awarded could exceed $10 billion, which at least to me weakly implies that were a settlement to be reached, it would likely be lower than this total.

Forecasters were caught off guard by Ilya testifying in the trial, that prop traded at around 30% until the announcement was made that he would in fact be testifying. Traders here also think that a licensing restriction, nullifying OpenAI’s exclusive licensing agreement with Microsoft, is a more likely structural remedy than undoing the for-profit conversion or ousting Altman or Brockman. So far, Musk has been remarkably restrained on X, and hasn’t been held in contempt of court for posting, but there’s always a chance, with traders still giving 12% odds to such an occurrence before the trial is complete.

OpenAI v. Everyone

OpenAI’s legal troubles, across the board, are a millstone around their neck for the foreseeable future. Anthropic’s battles with the Trump administration have sort of fallen by the wayside, with the DoD now back in engagement with the company despite their “supply chain risk” dispute.

It’s easy to forget, but the trial between the New York Times and OpenAI/Microsoft still promises fireworks. Markets have them as favorites to win (or at least settle in their favor):

The case has dragged on in pre-trial due to a number of consolidations, but traders think it’ll be a blockbuster, with more than a 1 in 4 chance of getting up to the supreme court!

Reading through the prop bet market’s strikes provides a little window into the Overton window of 2024. It’s almost inconceivable to imagine the court “ordering GPT-4 to be deleted” whatever that would mean at this point. Also, the idea of $10 or $100 million being reasonable bars to set for “large amount of money to OpenAI” may have made sense in early 2024 but now seem laughably small.

Another trial, which briefly made the headlines, is the most prominent of several wrongful-death lawsuits filed in California against OpenAI by the Social Media Victims Law Center. The parents of 23-year old Zane Shamblin have brought multiple claims against OpenAI for the wrongful death of their son.

It would be great to have more forecasters take a look at this new market, as wrongful death suits have the potential to dramatically shift the way AI is deployed among laypeople, especially with certain AI companies positioning themselves as deploying primarily enterprise models and others developing apps for the median global citizen.

Renting Compute

This trial has also made some unusual bedfellows. As Intelligencer and others have reported, xAI (which has plenty of compute but not enough users) has loaned its Colossus 1 data center to Anthropic (which has plenty of users but never enough compute). Pertinently, Colossus 1, which has a hodgepodge of different GPU clusters within it rather than a single, unified architecture , is better for inference than training workloads (serving AI models rather than shaping them), which is precisely where Anthropic is currently straining at its buttons. But it’s hard not to see this move as a middle finger to their mutual rival despite it also making business sense on a couple levels.

Forecasters had been mixed on the odds of this happening, and in fact, earlier in the year, it looked quite unlikely, with xAI posturing as though they were going to go all-in on competing as a frontier AI lab. After their merger with SpaceX was announced in February, however, the market shifted dramatically north.

With certain markets, I actually think the creator deserves a lot of credit for writing a question well in advance of that question becoming highly relevant. Manifold user RS also has markets on whether OpenAI or Meta will rent compute to external customers by the end of the year. OpenAI is trading at 11% (I guess there’s always a chance of some sort of compute arbitrage even among the leaders) but the market on Meta, which is trading at 38%, has woefully few traders and is probably your best bet at fading Meta’s near-term chances.

If Meta starts renting out their compute before the end of year, I think it’s safe to say that that would be a negative signal towards their success at staying afloat in the race.

Roundup

Pope Leo’s first encyclical, now all-but-confirmed to be about AI, will come out on Friday. I’m excited to read the document, Pope Francis’s encyclicals were famously bangers, from Laudato si’ on consumerism and climate change to Fratelli tutti on global consensus and camaraderie.

You can trade on whether it’ll release on time, mention human extinction or deepfakes, discuss if an AI can possess a soul, and name an AI exec (I think this option is slightly underpriced at 3% but I can’t trade on my own market: free alpha here).

You can also trade on whether Pope Leo will mention prediction markets this year:

Also, Trump and Xi appeared to have shaken hands in their recent face-to-face meeting for 14.5 seconds, just under the highest bucket. Too short!

Happy Forecasting!

-Above the Fold

Market Boosts

You can now boost a market directly from the creation page! If you know that everyone on the site is going to want to see your genius market, you can go ahead and grab a boost as an add-on instantly.

Boosts are awesome because they put your market in front of everybody for a day, and since March, they now come with a cool bonus of M$5000 in liquidity! You can bet on boost sales here:

Prizes & Charity

The Charity Champion Trophy is now showcased on the Charity page - be the #1 entry holder in the charity giveaway to claim this super-cool trophy from @Eliza. Holding the trophy at any point grants permanent ownership of “Champion’s Legacy” hovercard background.

Don’t care about prize drawings or charity giveaways? Don’t sweat it, turn off the notifications here in the notification settings tab. There was definitely a bug here before, but now we’re 100% sure it works.

If you were one of the first people to buy Prize Drawing entries, you may have noticed that you paid a bit more than you expected. This was due to the way the rates were calculated on the bonding curve: the mana price of each entry increased with each additional one you purchased. This was most noticeable for the first few users in each drawing when the curve was the steepest, but as of yesterday, you will no longer be charged any more than the amount you put in the field! Your total spend will be capped at that amount, even if the bounding rate is steep or another user swipes those cheaper entries out from under you.

Thanks to @MarySmith for adding the Free State Project to the list of available charities! At the moment, the FSP has a 15% chance of winning the $1500 payout, contribute towards (or against) them on the Charity page.

Merch & Shop

In case you missed it: Card purchases with Stripe are back! We still prefer buyers to use crypto to avoid fees, but all verified or grandfathered accounts are now able to use their credit card to buy mana again.

When ordering merch, there’s a new optional email field - if there are any delivery issues, this allows the customs office or carrier to contact you to get it resolved. For Brazil and some regions of Spain, the field is mandatory for duty resolution.

Also, for users shipping merch to Brazil or South Korea, we now collect a personal tax ID, which is required to pass through customs. If you live in one of those countries and had issues with your order, try again now with the updated form! As always, DM @Gen if you have any issues with an order.

In the shop, there’s a new “Owned” tab where you can find every profile and avatar cosmetic that you’ve purchased. Build the right fashion loadout for today’s trading, all in one spot.

If there are no seasonal items available at the time, the Seasonal tab will be hidden. But don’t fear, your previously purchased seasonal items are still visible in the “owned” tab. There’s also a handy tooltip on those items to tell you why they’re hidden in the regular shop.

There’s now a link to the Manifest website on the Manifest ticket in the shop, in case you were wondering what exactly you were buying a ticket for.

There was a secret feature in the shop where you could make the item cards dance by hovering your cursor at just the right spot - the card would raise just above the cursor and then lower back down, jittering over and over again. This has been removed because dancing is illegal on Manifold.

API & Bots

Do you have a bot on Manifold? If you do, and they make over M$100 in profit in a month, they’re now eligible to receive league rewards! The top 6 bots each month will receive rewards similar to other leagues, with the top bot earning M$800.

If you don’t run a bot already, this is a great time to start. You can now mark your account as a bot easily from your Account Settings to automatically enroll in the Silicon League. Ask in the #api-and-bots channel of Discord if you need help or advice from other bot-makers!

For you API power users: All timestamps are now precise! There was an issue with how timestamps (such as a bet’s createdTime) were handled as they made their way from the database through the API server, which led to them being truncated to the second. Now, all timestamps should have proper resolution all the way down to the millisecond.

If you’re placing comments from the API, you should also be aware that we’ll now reject any comment with a replyToCommentId that doesn’t exist. Instead of that comment being lost in the void forever, we now let you know so you can reply to a comment that does exist.

Also, when you go to cycle your API key, the site makes it clear that this will break all integrations using the key instead of referring to the old Twitch integration that no one has used in years.

Miscellaneous

• The daily $100 prize drawing is now automatic! No more waiting until an admin clicks the button. Long-time league members will remember a similar situation way back when.

• Did you know that when you zoom in on a market’s price history, the y-axis shrinks to show you the relevant area? Now that also applies to numeric and date charts! A good example:

• You can now filter the browse page by date markets, helping you find all those “when will X happen” markets to bet down since nothing ever happens.

• When we introduced cosmetics in the shop, we had hardcoded in the names of each item on the stats page, which led to some items in the table being referred to by their internal key. Now everything is nice and consistent, and will update auto-magically.

• Enlarging an image posted in a market description or comment section used to have a bug where a couple of special UI elements would continue to be drawn on top, especially user avatars from the right-hand sidebar on desktop. This is fixed, so now you can watch Supernatural gifs in full screen without any distractions.

• Pro tip: when you hover over a bet in the feed, you can see a lot of detailed information, such as the timestamps, share price, and limit order details. Now we also show a couple more decimals on the mana fields to help you understand exactly how much was spent on each trade.

• On mobile, the leagues horizontal scrollbar now shows arrows indicating that you can scroll, or just press the arrows to move. On desktop, these only show when you hover over the ribbon.

• Some buttons on the sidebar used to be comically narrow, especially the App button. This bothered me way more than it should have, and now all sidebar items are the same width.

Feedback

Any comments or questions about the changes above, feel free to post them below! I’ll do my best to answer anything, and I hope to do these changelog roundups in the future at least once a month. If you see any bugs similar to the ones above, feel free to post in the #bugs channel on Discord, or DM me on Manifold.

- wasabipesto

But while spot-scoring is fun, and lets you reward accurate forecasts rather than constant updating, some people (like you, the reader) like to forecast and trade on markets all year long. And there’s even an internal Manifold leaderboard for the traders that earn the most on this question set.

All these questions resolve at the end of the year, and Scott Alexander with Metaculus has done a great job of creating a diverse question set of fairly significant questions.

Currently, the market drawing the most attention and traders is on whether an AI-created song will chart in the top 20 of the Billboard Hot 100 this year.

Traders are heavily divergent on this, with #1 on the all-time leaderboard, SemioticRivalry and one of the Manifold founders, Ian, staking out massive YES positions against the #24 all-time trader, Panfilo.

While AI-generated songs have reached as high as #1, on other charts, there is discussion as to whether Billboard would allow an AI song to chart, or selectively exclude it from the rankings. AI-generated music hasn’t yet had the kind of viral moments as text, image, or video-generation, but this is probably due to the nature of how users share this content. While images and videos can go viral from anyone, songs often require a bit of lift-off. This can happen organically through things like TikTok audio clips, but generally a platform is required, and artists are hesitant to turn over creative control to the AI to the extent that a song could meaningfully be described as “AI-created,” rather than, perhaps, utilizing AI to create a single track that might be imperceptible to the listener.

The market remains at 31%, well below the aggregated forecast on Metaculus. The divergence between the markets and the aggregated forecast is interesting here, as typically Metaculus tends to be more conservative in the direction of things-not-happening.

Another AI capabilities market, on whether an AI model will reach a 3-hour time horizon at 80% reliability this year on METR’s benchmark, seems all but certain at this point.

The remaining uncertainty centers around whether METR will change their methodology or stop adjudicating models by this benchmark due to saturation or other priorities.

There have also been some large swings this year on the market on whether OpenAI will manage to file for (not complete!) an IPO this year.

This market trades a few percentage points above similar real-money markets on whether OpenAI will complete their IPO by the end of December, as filings typically happen anywhere from 15 days to six months in advance of the actual offering.

A couple of questions recently resolved, such as NASA’s Artemis II completing its mission successfully, and SCOTUS somewhat surprisingly issuing a firm ruling that Trump’s emergency tariffs were unlawful.

There are plenty of remaining questions that have split the Manifold community. For example, Keir Starmer’s fate has traded between 40% and 60% for most of this year:

And in lieu of any public statements by Supreme Court justices Alito or Thomas, the best we can do is this prediction market on whether SCOTUS will have a retirement this year before the midterms, which Metaculus prices decently lower at 24% and real-money markets place higher, above 40%!

On some of these questions, with just a couple dozen traders, there’s room for newbies to stake out a claim for themselves. For example, my favorite question in the tournament, on Montenegro’s accession to the EU, is one of the least-traded markets in the set!

Do none of you care about EU expansion? Croatia was the last country to join back in 2013; a lot has happened in the world since then. Montenegro, which began the year with just 12 negotiating chapters closed, has set the ambitious goal to close the remainder of these chapters this year, and hopefully join the EU by 2028. This is the EU equivalent of trying to walk 100,000 steps in one day.

The world’s second-most deadly ongoing conflict, in Sudan, might finally reach a ceasefire this year, as well, with a 25% chance of at least a temporary end to the civil war. Perhaps more forecasters should devote their analytical attention to this:

There’s also a meta-question in the ACX tournament, on how well the top bot will place, relative to the top 5 humans.

Forecasters expect that the top bot will be about as good as last year’s 92% score, which some forecasters viewed as anomalously high. They give a 24% chance that the best bot does as well or better than the average of the top 5, though. Someone should have made a bot to trade on the ACX question set, so we could see how it matched up on the same question set in a different modality. Currently no bot ranks on the leaderboard:

These markets have large liquidity subsidies and many will come down to the wire, so there’s still time to win the Unofficial Manifold x ACX 2026 contest. Special thank you to Manifold user “walmart” for creating all the markets.

Get those limit orders up and happy forecasting!

-Above the Fold

And markets can provide clarity. Here’s the polling on the race:

Not super informative. With the top-two primary in six weeks, no candidate currently is polling above 18%. Two of the four candidates polling the highest are Republicans, who haven’t won statewide office in twenty years. And another is, ummm, Eric Swalwell. Until a couple weeks ago, U.S. Rep. Eric Swalwell was the frontrunner in the race. He has since suspended his campaign after serious and credible allegations against him came to light.

The second-place Democrat in the polls, Tom Steyer, has now swapped places with Swalwell in betting odds (see Swalwell in purple drop to zero as Steyer’s in yellow soars up to ~60% above). Steyer is a slightly awkward candidate, as an aging billionaire turned climate activist who has managed to ingratiate himself into three of the resurgent wings of the Democratic primary base: progressives, YIMBYs, and resistance libs. And yet, he’s still only polling around 15%. But with just a few weeks to go, it’s unclear that any other Democrat stands a good chance of making the top two.

Matt Mahan, the tech industry-backed San Jose mayor, was a prediction market darling for a month or two, but has failed to gain momentum in the polls. Katie Porter, like Swalwell a U.S. rep with a nationwide profile, thanks to her schtick of holding up whiteboards in congressional hearings, has also failed to convert that profile into enthusiastic support.

Xavier Becerra has been within striking distance of Steyer in a couple recent polls, and was HHS secretary under Biden for four years, a position now occupied by RFK Jr. Unlike RFK Jr., Becerra isn’t exactly a household name; nevertheless, he finds himself cutting into Steyer’s lead in the markets.

While polls testing the waters for a very late write-in entry, Kamala Harris, have teased her interest, the markets think that it’s unlikely that she’d enter at this stage. Instead, traders continue to think her eyes might lie on the 2028 election:

The nightmare scenario for Democrats, being locked out of the general election in the top-two primary, appears unlikely at this point. But a few % is not nothing. Conservative TV show host Steve Hilton and Riverside County Sheriff Chad Bianco are both polling above 10%, and if the Democratic primary becomes even more splintered, or an independent candidate like Elaine Culotti takes off, this is a remote possibility. For example, imagine Steyer falls in the polls due to a scandal too minor to force him to drop out, or if Mahan starts rising, eating share from Steyer and Becerra. If Bianco gains on Hilton, they could realistically each get ~16% of the vote.

If we wake up on June 3rd and the vote tallies read…

Hilton: 17%
Bianco: 16%
Steyer: 14%
Becerra: 14%
Mahan: 13%
Porter: 12%
Culotti: 9%
Other: 5%

…would anyone be truly shocked? Why would any of these candidates drop out if they’re all polling within a couple percent of each other, with decent odds of making the general election? Betty Yee dropped out this morning, and candidates like Tony Thurmond and Antonio Villaraigosa are potentially poised to do so as well, with their poor polling numbers. But just 4 Dems and one indie candidate remaining alive could allow a comfortable shutout. A ton of people will be on the ballot and a Californian primary election isn’t exactly the place you’d expect the extreme salience of tactical voting required to drive down the “Other” box.

That being said, it’s likely that news coverage around the race, along with top-down action from the Democratic party, will push forward a favorite or two to ensure this doesn’t happen, especially because this election is far less ideologically contentious than you might expect. There’s no clear “progressive” candidate running, with Porter, Steyer, and Becerra all having claims to that mantle. Most of the candidates are more or less amenable to a broad swathe of the Democratic electorate.

This is quite different from the election we’re seeing in Los Angeles, where incumbent mayor Karen Bass is now trailing rising progressive star Nithya Raman.

This wasn’t always the case. Raman has enjoyed a steady rise in prediction market odds since entering the race on the last possible day after observing that there was no strong challenger from the progressive wing.

Karen Bass, who if readers will recall, was close to being selected as VP by Biden in the 2020 election, has not a great run of things as mayor. She has received strong criticism for her handling of the L.A. fires and their aftermath, as well as the inability to address L.A.’s housing crisis. As a result, traders give her less than a 1 in 3 chance at retaining her mayoralty.

Raman’s campaign has drawn comparisons to Mamdani’s, but she’s actually being battled on her left by community organizer Rae Huang. DSA hasn’t endorsed either yet, likely expecting that Raman has better odds of winning but slightly poorer ideological alignment with DSA.

Spencer Pratt, a reality TV star and critic of Bass’s handling of the fire, also stands an outside chance, if he can consolidate the more centrist and populist voters in the primary.

Despite what seem to me to be two of the most interesting and significant elections we’ve seen for a long time in California, Manifold traders seem way more interested in something happening farther down the ballot.

A proposal for a wealth tax on billionaires is trying to make its way onto the ballot this November. Traders think it has a ~73% chance of qualifying, and give it ~58% odds of passing, conditional on appearing on the ballot.

All told, that gives the proposal better than 1 in 3 odds of passing this year.

Billionaires have preemptively begun to flee the state and establish residency elsewhere, as the law retroactively applies to anyone in residence at the start of 2026! Traders think that even if the law passes, there’s a considerable chance that it’s overturned by the courts or fails to practically collect revenue (at least $1 billion):

If you are completely uninterested in the California elections except for as it pertains to AI, perhaps you’ll be pleased to know that traders remain unconvinced by Altman’s threat from a few months ago to move OpenAI out of California…

…despite decent odds of another AI ballot measure this fall:

And if you like mayor elections but only those that have no governor above them, well, that leaves you with just the fascinating and increasingly contentious DC mayor election, where the DSA-aligned Janeese Lewis George is leading against the center-left Kenyan McDuffie.

And finally, if you don’t care about elections at all, and you also hate California, well… perhaps you’ll be interested in this long time-horizon market.

Happy forecasting!

-Above the Fold

We’re returning to Lighthaven on June 12-14. You can buy your tickets now on the Manifest website, with Early Bird pricing through this Saturday, April 18:

Get Tickets

What: Manifest is a festival ostensibly about prediction markets, but secretly about connecting with old friends and people you admire from your favorite niche corners of the internet. It’s a gathering of nerds who want to find the thinkers and practitioners they vehemently agree/disagree with, share a meal around a cozy campfire, and come away with radically new ways of thinking. If you attend, don’t be surprised to find yourself engrossed in a late-night rabbit hole with your future cofounder, a serendipitous nerd snipe from an internet celebrity, and the joy of a crowd cheering (and betting with mana) on a wrestling match between strangers.

Who: 500-650 exceptional people, from many fields, including but not limited to: forecasting, rationality, effective altruism, AI & AI safety, economics, finance, tech, policy, writing, and novel ideas.

When: June 12-14 (Fri-Sun)

Where: Lighthaven, Berkeley, CA (map)

How:

• 🎟️ Tickets (Early Bird until Sat 4/18!)

• 🏡 Room bookings

To get updates: Join the Manifest Discord

This year, we want to make it even better, focusing on the things attendees have told us they value most: high session quality and variety, well-curated attendees, thoughtful sponsor presence, delightfully engaging activities, easy-to-use scheduling, clear community guidelines, and structured ways to meet and stay in touch with the people you’re supposed to meet.

Email team@manifest.is if you have questions, ideas, or feedback. We’d love to hear suggestions for more excellent speakers or organizations to collaborate with. Help us fill Manifest with the most stellar, thoughtful, kind humans you know!

You’ll hear updates from us as we confirm more of the speakers, sponsors, and plans we’re cooking up. We look forward to seeing you in June!

Highlights from 2025

2025 Talks included:

• A Story Telling of Kalshi’s Internal Drama & Decision Making (Noah Sternig)

• Reversible Cryopreservation (Laura Deming)

• Writing and Slop (Roon, Scott Alexander, Gwern, Alexander Wales)

• Life as a Professional Gambler (Anonymous)

• Has Trading and Gambling Gone Too Far in the US (Jeremiah Johnson, Isaac Rose-Berman, Christopher Gerlacher)

• How Tech Hiring is Changing with AI (Sholto Douglas [Anthropic], Ben Cohen [Substack])

• Schools Should Pursue Excellence (Tracing Woodgrains)

• How to be Hot (Aella, Chesed)

Check out our YouTube channel for more examples from past years!

Some of our 2025 Events:

• Nash Pit: A Prediction Market Game Show

• Night Market + Career Fair

• Live Polling on Spicy Questions

• Speed Friending

• Rare & Exotic Drama Games from Australia

• Hot Seat w/ Niacin (vitamin B3, which causes a rapid flushing response)

• Startup Pitch Competition

• Poker tournament with former pros

Get Tickets

Speakers & Events

Every year, we’re grateful for the amazing speakers and guests who come out to Manifest!

Stay tuned as we announce the speaker lineup for this year. And if you want to give a talk, or have an idea for an event that needs our input/support, email winter@manifest.is.

Also, anyone is welcome to host a community talk or event! In “un-conference” style, Manifest will release a schedule with rooms and time slots you can add your session to.

Keep an eye out if you have something you’d like to share, we’ll send this calendar to attendees closer to the event (and it can be updated on the fly throughout Manifest, so you can add your talk if the inspiration strikes!)

Sponsors

Interested in recruiting the kinds of folks who find these kinds of intellectual discussions energizing? Or showcasing your company’s work in front of the influential speakers and leaders in attendance? Manifest has you covered. Each year, attendees report having impactful interactions with sponsors and appreciate their presence at the event. They find jobs, form new collaborations, and even switch industries! Email team@manifest.is to learn more and schedule a call.

Career Fair

On Friday, June 12 (the first night of Manifest), Lighthaven will be open and free to the public for our Career Fair + Night Market! This will be an excellent opportunity for even more people to mingle, talk to hiring sponsors, and join the fun for a few hours.

We ask that everyone interested (whether or not you’re attending the rest of Manifest) please register your interest in the Career Fair here.

Get Tickets

Festimonials

Don’t just take our word for it; here’s what our past attendees have to say (emphasis ours):

Jake Seliger:

Bess and I went to Manifest, which bills itself as “A festival for forecasting and prediction markets,” a description that may technically be true but fails to capture the spirit; to my eye and experience, it’s maybe more accurately stated as “Substack and Twitter live” or “a mixture of festival-conference-party-Burning-Man for nerds with many interests to show up and enjoy each other’s company.” Bess excitedly exclaimed it to be “Nerd Camp!” with a sort of takes-one-to-know-one gleam in her eye.

Tomie:

I love Manifest. I paid the full price for the full ticket, sucker I am, and my subsidy provided for these swaying bauble lights, these warm soporific nooks, these flames and corridors, these souls brought to Earth together, eyes lighting up at their electric worlds made real.

TracingWoodgrains:

For much of my life, I have poured my attention into tough-to-explain solitary pursuits, finding myself often sitting in quiet corners on the fringes of gatherings wondering if they’re worth the effort. Not so last weekend.

“I really like that Manifest is inspiring in me the ‘Oh, I should make and bring random goofy stickers and swag’ impulse that I’ve had inculcated in me by Defcon” - Dave Kasten

“Maybe the best conference I ever went to, beating all programming conferences, EAG, singularity summit” - MQP

Venue

Just like last year, Manifest will be held at Lighthaven. We credit much of the great Manifest vibes to this beautiful and cozy space. It has fires for literal fireside chats, intimate nooks for one-on-one conversations, spaces for small workshops & big talks, and gorgeous overnight accommodations.

Festival Season! LessOnline —> Summer Camp —> Manifest

Manifest is actually the capstone to a 10-day season at Lighthaven! The weekend before will be LessOnline (June 5 - June 7): “A festival of writers who are wrong on the internet”. It is an unconference with a collaborative schedule, bringing together a “mostly-online subculture of people trying to work together to figure out how to distinguish truth from falsehood using insights from probability theory, cognitive science, and AI.”

In the days between the two conferences, there will be Summer Camp (June 8 - June 11), a peaceful yet vibrant week of coworking, collaborating, and late-night campfire chats. You might see: Hackathons (or “Forecastathons”), organized discussions, jam sessions, dance parties, tournaments, games of all kinds, camp activities (sardines, s’mores), or multi-day intensive workshops (e.g., CFAR-style).

You can come to any combination of these, but we’re offering a $350 discount off standard pricing if you buy All-Access tickets to all three events, available here.

See you at Manifest!

Amidst the recent turmoil between the federal government, Anthropic, and its rivals, I’ve seen a number of commentators point out something interesting. The private ownership of Anthropic and OpenAI means that far fewer Americans have a direct stake in their success than in other firms of their size and centrality to the economy.

If the Pentagon had attempted to label, say, Walmart or Exxon Mobil a supply chain risk, then the tens or hundreds of millions of Americans with direct stakes in the stocks of those companies would feel the effects, regardless of their personal affinities towards them.

This might change soon. Anthropic, OpenAI, and other AI companies are rushing to IPO as soon as this year. They will skyrocket into the top echelon of publicly traded companies if and when this happens. It wasn’t always obvious that they’d choose to press for a fast timeline. With their ability to attract the largest private investment rounds in history, it wasn’t clear that going public was as necessary as for other capital-intensive private companies.

But, with secondary markets clamoring for access to shares, and capital expenditures for hyperscalers rising just as precipitously as revenue, IPOs in the next 12 months are likely. Internal pressure from employees with equity and early VCs won’t hurt. Manifold traders are quite interested in the topic as well…

For a while, the race to IPO for American private AI firms looked like more of a three-horse race, but the merger of xAI and SpaceX put that to rest, with SpaceX/xAI now very likely to go public within the next few months.

While many observers might fairly think that xAI isn’t really in the same league as the other two companies, they did in fact raise money earlier this year at about half the valuation of Anthropic (about $250 billion). And now, packaged with SpaceX, they might be more valuable than both.

But, ummmm, it’s not clear how long this will hold.

Valuations

Just like the intelligence of their models, their number of employees, their revenue, and the number of protestors outside their offices, the valuations of these companies (check out MNX) seem to be doubling every few months.

Epoch has had great success drawing straight lines on log-plots, and they called Anthropic surpassing OpenAI’s revenue in early 2026.

With their latest run-rate revenue forecast of $30 billion a few days ago, this market resolved YES, sending markets on whether Anthropic would flip OpenAI skyrocketing.

The market on combined EOY revenue for America’s three private AI companies had to create some new buckets in the last week, as the top bound was around $150 billion, which now looks to be below the median trader’s expectations. Two Twitter users (at least one of whom is a top Manifold trader, I believe) made an over/under bet at $90 billion on just Anthropic’s run-rate revenue, and we now have a market on it:

Much of the recent bullishness around Anthropic has been the disclosure of their Mythos model. Enough ink has been spilled about Project Glasswing and Mythos’s emergent ability to spot zero-day vulnerabilities, but nevertheless…

It’s a story that at the same time is extremely urgent for the public and policymakers while being slightly overhyped in breathless reporting. As keen observers have by now pointed out, previous models like Opus 4.6 and GPT-5.4 likely have abilities nearing this as well, with Mythos representing a continued climb up that particular capability curve, rather than a discontinuous jump. Skeptics have noticed that the amount of cost and compute thrown at discovering individual vulnerabilities during testing ranged up into the ballpark of tens of thousands of dollars, and mused that Opus 4.6, given that amount of time and energy, could probably find a similar vulnerability.

Obviously, without public access to the models, it’s hard to say. No one seriously doubts that Mythos is a highly capable model, and traders in fact think that sooner or later, Anthropic will release it to the public in some form:

And when it does, it will likely come in at about 10x the cost of other frontier models on a per-token basis.

And even if we don’t get Mythos, we’ll likely have access to something with similar capabilities within the next few months:

Bubble

Alternatively, many online and even some on Manifold hold a very different set of beliefs about how AI will play out over this year. While it’s hard to operationalize a market on what an AI bubble “popping” would look like, one creator has given it a good try:

The criteria for this market read:

This could manifest in many different ways:
- Sudden drop of the NQINTEL or S&P500 indexes
- Halt of capital expenditure and/or VC investment in AI
- Private AI companies (such as OAI, xAI, Anthropic, etc.) raising at substantially lower valuations than previous rounds, defaulting or shutting down.
- Governments / big enterprises moving down AI in their priority lists.
These events need to be widely reported and referred to as an AI crash or similar by trusted news sources such as Bloomberg, NY Times, FT, Reuters, AP, etc.
Other reasons for a stock market crash or economic slowdown (tariffs, wars, etc.) will not count.

Alternatively, another creator has simply made a market on which AI companies will drop dramatically from their peak this year. Apart from Oracle, traders think this is unlikely for most firms that are heavily leveraged in the AI boom.

Notably, the odds for a general recession—potentially caused by, umm, overseas conflict—are much higher than a crash localized to the AI sector:

Roundup

Not just the AI companies might go public this year. You can bet on Discord, Anduril, Skims, and Jersey Mikes too!

Is Mythos really AGI if it can’t speedrun Minecraft?

And we probably won’t know what Ilya has been up to by July:

Happy Forecasting!

-Above the Fold

Trump accused him of leaking classified information. He then directed Pete Hegseth (my bad - Secretary of War) to begin identifying and releasing government files related to “alien and extraterrestrial life, UAP, and UFOs,” citing “tremendous interest shown.”

Hegseth reposted the announcement with an alien emoji and a saluting emoji. Three days later, asked whether he personally believed in extraterrestrials, he said the task was not on his “bingo card” but he was “digging in” and would find out “along with all of you.” The Office of the Director of National Intelligence posted on social media that the files would “soon” be declassified.

Last week, the White House registered aliens.gov and alien.gov through CISA’s federal domain registry. Currently, neither connects to anything. The deputy press secretary responded to press inquiries with “Stay tuned!” and an alien emoji. Both domains were registered while CISA is not currently accepting new .gov requests, due to a lapse in federal funding.

The Pentagon’s All-domain Anomaly Resolution Office, AARO, was created in 2022 to investigate unidentified anomalous phenomena. It has reviewed over 2,000 cases. Every one traced back to drones, balloons, atmospheric effects, or sensor artifacts. Sean Kirkpatrick, its first director, said of the incoming file release: “Nothing would have made me happier in that job but to have discovered alien technology and rolled it out. I don’t expect to see anything new.” He described Trump’s announcement as “a distraction for the administration.”

The reason the interesting UAP evidence can’t be released is structural. The most compelling cases are compelling because classified sensors detected them: military radar, satellite tasking, signals intelligence. Releasing the footage means revealing what those systems can see, at what resolution, from what angle. Past disclosures have leaned on pilot reports and low-resolution cockpit video while withholding the metadata. The result is always redacted PDFs.

The disclosure community has been predicting imminent official confirmation since the Clinton administration. Obama, Hillary Clinton, Biden, Trump’s first term, Trump’s second term, and Pope Leo XIV have all been identified at various points as the figure about to initiate disclosure. In 2025, the Wall Street Journal reported that hundreds of Air Force personnel had been told, falsely, that a secret program to harvest alien technology existed. The disclosure community cited this as evidence of cover-up activity.

The most useful number in the Manifold alien market cluster is not any of the confirmation probabilities. “Will the US Government make a show of declassifying everything on UFOs/UAPs before 2030?” sits at 44%.

Markets around whether the US ACTUALLY confirms that aliens exist before 2027 hover around 7-8%. Traders think there’s a real chance of an elaborate public performance of transparency, but a much smaller chance the performance amounts to anything.

Traders are also forecasting specifically if the US confirms aliens in 2026, what kind will they be? The leading answer at 49% is Microbial life, with the runner up at 27% being “completely unspecified (US government confidently asserts aliens exist but provides no further information).”

There's also a market on whether the US President or UK Prime Minister will confirm non-human intelligence before end of 2026, sitting at 13% - Keir Starmer apparently included for completeness.

This sits alongside markets on whether the US President themself will confirm extra-terrestrial existence.

Moving on from the politics, we have “Will we have contact with intelligent alien life by 2030?” sits at 5%, roughly the same as the JFK assassination being linked to aliens before 2040.

“Will the first sign of extraterrestrial intelligence be identified as artificial intelligence by 2030?” is at 13%, which is a more interesting question than it gets credit for. “Alien invasion before GTA6?” is at 4%.

The most substantive alien science of the past year happened nowhere near Washington. In April 2025, a Cambridge team using the James Webb Space Telescope announced that the atmosphere of K2-18b, a water-rich planet 124 light-years away, appears to contain dimethyl sulfide, a molecule produced on Earth exclusively by marine microbial life. The detection reached 3-sigma confidence. The lead researcher called it “a revolutionary moment.” Four independent teams reanalyzed the data over the summer and found the signal was probably ethane, or red noise in the mid-infrared, or a real detection that needs more observation time. One analysis found 87.5% of retrieval approaches showed no statistically significant evidence for DMS at all.

“Will K2-18b be generally agreed to harbor alien life by 2027?” is at 5%, with a scientific consensus sitting at the same odds.

This arc is mostly familiar. The phosphine-on-Venus announcement in 2020. The 1996 Martian meteorite bacteria claim. All had similar beginnings. K2-18b received a fraction of the coverage of Obama’s podcast clip.

Harvard astronomer Avi Loeb has a public bet with skeptic Michael Shermer: $150,000 against $1,000 that alien visitation will not be confirmed by 2030. Markets broadly agree with him. The aliens.gov domain expires in 2027. So does the confirmation market. Neither currently has any content.

Happy Forecasting!
- Above the Fold

There is a particular category of political problem where the solution becomes the source of the next crisis. Britain is currently running several simultaneously.

First-past-the-post was designed to manufacture stable majorities from a fragmented electorate, compress minor parties out of relevance, and deliver a decisive two-party duopoly that Enlightenment architects (presumably) found comforting. The 2024 election gave Labour a 172-seat majority on 33.7% of the vote, which is the system working exactly as designed. The combined Labour and Conservative vote share that year was 57.4%, itself a record low. It has since fallen to 38.2%. The mechanism for producing stability is now producing a governing party held in place partly by fear of what replaces it, an opposition so fragmented it cannot agree on its own purpose, and projection models where Reform wins a majority of 112 seats on 30% of the vote because the left cannot coordinate around a single candidate. No party has polled above 30% in months.

Manifold gives 18% odds that any party consistently hits 35% this year. The system that was supposed to squeeze chaos into a majority is instead amplifying it.

The special relationship was meant to guarantee British influence in Washington: a privileged bilateral channel that would make up in diplomatic proximity what Britain had lost in post-imperial weight. It is now the leverage Trump uses against Starmer, a relationship so asymmetric that Britain ends up defending its right to be an ally rather than assuming it. “This is not Winston Churchill we are dealing with,” Trump told journalists, which lands decisively even if it describes nothing more than a specific disagreement about which bases to lend.

The bond market was meant to discipline governments that spent too freely. It has, for the past year, been the main thing keeping Keir Starmer in office, because gilt yields spike whenever it looks like he might be replaced by someone to his left. A Labour Prime Minister, historically unpopular with voters, sustained in position by the City’s preference for his fiscal conservatism over his successor’s probable alternative. If you wanted to design a scenario that made Labour’s internal politics look most absurd, it would be difficult to improve on that one.

Reform was meant to be a clean insurgency: anti-establishment, ideologically coherent, the vehicle for everyone who had decided the old parties were irredeemably broken. It has been in local government for eight months and already has 1 in 18 of its elected councillors gone through suspension, expulsion, or resignation. One shared a meme suggesting Hitler “would have been a legend” had he targeted a different group. One was charged with assault at a Pride event. One had been the council leader caught swearing at colleagues on video, which then got leaked. Six of the seven Kent councillors who defected to Rupert Lowe’s Restore Britain had already been expelled by Reform before they defected. Reform’s response: “Restore are welcome to our dregs.” The party originally elected 57 councillors to Kent County Council. It now has 48.

This is not the same observation as saying British politics is chaotic, an insight we’ve already covered in detail. It is the more specific observation that the chaos is being generated by the mechanisms supposed to prevent it.

The Teams Call

On Sunday 25 January, at 11am, Keir Starmer dialled into a Microsoft Teams meeting. He was about to speak with Zelensky. First, though, he personally led an 8-1 NEC vote to block Andy Burnham from standing as Labour’s candidate in Gorton and Denton, on the grounds that triggering a fresh mayoral election would drain resources ahead of May.

The seat itself had been triggered by Andrew Gwynne, the former health minister suspended from Labour after WhatsApp messages surfaced in which he called a constituent a “Duki Hag” and described her neighbourhood as a quote-unquote “shithole.” The seat had returned a Labour MP, in various boundary configurations, since 1931. Burnham, the most popular figure in Labour politics (net approval ±0, against Starmer’s -47), was blocked. Labour then deployed a leaflet from a tactical voting organisation called “Tactical Choice” recommending voters back Labour. No such organisation exists. Nobody appears to have checked this before the leaflet went out.

Labour still came third. Hannah Spencer, a 34-year-old plumber and Green councillor, won 40.7% on a 27.5-point swing - the largest Green by-election swing ever, and the party’s first by-election win. It was the 127th seat on their target list. The mechanism Labour used to protect Starmer from a leadership challenge produced the most embarrassing by-election result the party has suffered since 1931, in a seat the leadership explicitly chose to fight as a show of strength.

The Self-Demolition

The more analytically interesting question about Reform is not whether electoral headwinds will slow them down but whether they will manage to do it themselves first.

The governance record after eight months in local power is instructive. Having promised to eliminate “waste” and cut council tax, every Reform council is raising it. Worcestershire got special government permission to exceed the standard 5% cap, raising council tax by 9%. Kent’s own social care lead admitted “spending is down to the bare bones.” The forest of waste, when located, turned out to be a single shrub. The issue is structural: 48% of Kent’s budget goes on adult social care, which does not yield to populist promises about DEI spending. The easy observation is that Reform’s candidates did not know this when they were elected. The pointed one is that several of them had previously served under Labour or Conservative administrations and knew perfectly well.

The candidate quality issue has been persistent enough to quantify. At least 38 Reform councillors were suspended, expelled, or resigned within the first six months after the May 2025 elections, with at least four facing criminal investigations. The party ran a man as council leader in Staffordshire who had previously posted that a statue of a Black woman was a “fat a***d black woman,” and when it emerged, the internal email from the council leader defending him was accidentally sent to all 81 members of the council rather than just Reform ones.

Restore Britain now sits as third-largest group on Kent County Council. Six of its seven defectors were people Reform had already expelled. The party Rupert Lowe calls more principled than Reform has published a 133-page mass deportation policy document and lists Carl Benjamin (Sargon of Akkad) among its affiliated figures. One of Reform’s own spokespeople called Restore “neo-Nazi.” Musk called Reform “Nazis” in return. This exchange happened in public, in February, in the run-up to local elections. The right-wing vote splitting effect this produces is not primarily an electoral threat to Reform from Restore, which is probably too extreme to gain serious traction outside online ecosystems, although traders do think there is potential for them to win a seat or two.

On funding: Reform’s biggest donor is Christopher Harborne, a Thailand-based investor who has given £12 million over the past year, including the single largest living political donation in UK history at £9 million in one go. The Electoral Commission has warned it cannot verify the identity of Reform’s crypto donors because the party’s payment processor is a Polish-registered entity not regulated by the FCA. The Commission’s exact words: donations could be “unidentifiable” under UK law. This sits interestingly alongside the party’s central pitch about cleaning up British political corruption.

The Squeeze

British politics has reorganised around tactical voting in a way that is not entirely serving its intended purpose.

Labour’s theory entering the next election is straightforward: fear of Reform winning will squeeze Green and Lib Dem voters behind Labour candidates in marginal seats, replicating the 2024 mechanism that gave them their majority. YouGov’s tactical voting study this month complicated this. 57-58% of Lib Dem and Green voters would tactically back Labour to stop Reform in their seat. But only 43-44% would do so to stop the Conservatives.

Gorton was also the clearest signal yet that tactical voting is no longer reliably flowing toward Labour from the left. Three independent tactical voting organisations endorsed the Greens as best placed to beat Reform. Labour invented a fourth to endorse itself. The Greens won by 4,400 votes. The lesson the Green Party absorbed is that they can now credibly position themselves as the tactically rational anti-Reform choice in seats where they are competitive with Labour, which changes the calculus Labour built its entire seat-count modelling around.

Manifold has Reform at 40% to win the next election, and 20% to top the YouGov poll at that election but still lose.

That second number is the most structurally interesting: the scenario where tactical voting coordination across the left succeeds well enough to deny Reform the FPTP amplification they need. Electoral Calculus models this as reducing Reform from 398 seats to 335 when tactical voting is applied. The 63-seat gap is what coordination on the left is worth, which is a reasonable description of why the Gorton result was alarming to Labour in a way that transcends the headline.

Not Winston Churchill

Keir Starmer has never said a bad word about Donald Trump in public. That is not being reciprocated.

The US asked to use Diego Garcia and British bases for offensive strikes on Iran. Starmer initially said no. He later said yes for defensive purposes only. Trump called Britain “uncooperative,” said Starmer was “not Winston Churchill,” told the Financial Times the US-UK relationship was “obviously not what it was,” and added that France had been “great,” which is a sentence Trump has not historically deployed. The Financial Times called it Starmer’s “Love Actually moment,” comparing it to the 2003 film scene where a British PM pushes back on a bullying American president. Whether Starmer was playing Hugh Grant or just muddling through is genuinely unclear, but his parliamentary line had some coherence: “American planes are operating out of British bases. British jets are shooting down drones to protect American lives. Hanging on to President Trump’s latest words is not the special relationship in action.”

The practical scale of British involvement: Typhoons and F-35s intercepting Iranian missiles across the Gulf, 300 British personnel within a few hundred yards of an Iranian strike on a Bahrain base, RAF Akrotiri hit by a drone on Cyprus. Starmer has been threading a needle between voters who remember Iraq (and do not want another Middle Eastern war) and an American president who wants “completely blind loyalty” from allies. As one former national security adviser noted, Trump’s officials “crack on as usual behind the scenes” even as the president performs public frustration. The intelligence-sharing relationship continues. The trade deal does not, currently, and may not.

Reform’s position on all of this evolved at impressive speed. Richard Tice called for Britain to “play our part in degrading once and for all the appalling Iranian regime.” Farage argued the West needed to “take the gloves off.” Then petrol prices became a political story, and within days Farage had recast himself as a voice of restraint, warning Britain “cannot get involved.” Suella Braverman, now a Reform MP, accused Starmer of “destroying the special relationship” while simultaneously arguing against sending British troops. Starmer noted, with some restraint, that “those that two weeks ago were urging us to go headlong into a full blown war are beginning to have second thoughts.”

The 1931 Problem

The conventional read on May 7 is Reform sweeps English county councils, Greens pick up London boroughs, Labour suffers the kind of losses that trigger internal recriminations. Farage has pledged to spend “every single penny” of over £5 million on the campaign. Reform at 3/10 to win most seats nationally implies roughly 75% probability. All of that is probably correct.

Another structurally significant story is Wales (probably a first).

Labour has not lost a Welsh election since 1931. Polling has Plaid Cymru at 30%, Reform at 25-27%, Labour third at 18-21%. Welsh Labour’s First Minister Eluned Morgan has taken to saying “Keir Starmer is not on the ballot paper in this election,” which is accurate and desperate in equal measure. The Senedd is simultaneously moving to 96 seats under a proportional closed-list system for the first time, replacing the hybrid that previously squeezed out smaller parties. Under PR, Reform wins seats regardless of finishing position. Whatever happens on May 7, Reform will have permanent institutional representation in Cardiff Bay in a way it has never had, which is a durable structural shift independent of the headline result.

Follow the implication through: if current polling holds, all three devolved first ministers simultaneously favour departure from the United Kingdom. This is not a sentence Westminster is currently discussing with any urgency, possibly because doing so would require acknowledging that the devolution settlement Labour created in 1999 is generating outcomes it did not model.

In London, the Greens are at 28% in capital polling, with all 32 boroughs up. In Scotland, the SNP holds a clear lead but Reform has surged to second, with Labour down to 15% on the constituency vote, their worst share in any election since 1910.

Every Labour rival has reached the same conclusion: let him own the May results before moving. Rayner, Streeting, and Burnham are the names in circulation, with Burnham conspicuously blocked from the parliamentary seat he'd need to run. Polanski told the FT: "I could see potential to work with Burnham. I would rule it out with Starmer." A party that watched the Conservatives go through six leaders in a decade and lose by the largest margin since 1832 is apparently considering the same strategy. Knowing a thing and doing the opposite are not mutually exclusive in British politics, as the Tactical Choice website would suggest.

Happy Forecasting!

- Above the Fold

Warning: discusses movies from 2025, a few borderline spoilers but nothing egregious.
Check out our Oscars 2026 Dashboard Here.

GQ thinks 2025 was the best year for movies in a decade, and I’m inclined to agree. Despite the Best Picture award being a two-film race, there’s a deep bench of truly great films from last year in the Best Picture category.

While I’ve personally only seen about half of the best picture nominees, I spoke with Manifold user Ziddletwix, our community’s resident film expert. Ziddletwix has more profit in the “TV and Movies” category than the rest of the top 10 combined, and he also manages most of the Oscars markets on the platform this year.

The Big Awards

Ever since One Battle After Another came out (and despite losing about $100 million), it has been a heavy favorite for best picture. It has all the hallmarks of a winner:

• High pedigree director who is “due” an Oscar (Paul Thomas Anderson).

• Elite cast including both old favorites like Leo DiCaprio and Benicio del Toro, as well as a talented new actress in Chase Infiniti.

• Politically relevant theme in a historical (-ish) American setting.

• Intense action balanced with character development and emotional performances.

Despite all this, and despite sweeping most of the precursor awards leading up to the Oscars, it’s not a done deal. Sinners has dogged OBAA in the market for the last couple months. Sinners has the most nominations, indicating broad appeal across many of the professions that vote for the award, populist appeal, and recent momentum through its win at the Screen Actors Guild.

The other films have little hope, according to traders. My favorite film of the year, Sentimental Value, stands virtually no chance of winning, as it’s a Norwegian film about nothing much happening. Joachim Trier’s previous film about nothing much happening, Worst Person in the World, didn’t even get a best picture nomination, despite its leading actress winning in her category (edit: clearly humans can hallucinate too, not only did Reinsve not win Best Actress, but she wasn’t even nominated… she did win Best Actress at Cannes though). Guillermo del Toro’s Frankenstein stands little chance of earning him a second best picture win, after The Shape of Water surprisingly clinched the 2018 awards. Del Toro’s magnum opus from a decade prior, Pan’s Labyrinth, didn’t even receive a nomination. How fickle the Academy Awards voters can be. Marty Supreme, which earlier in the year was generating slightly more best picture buzz, has now fallen below 1%, despite this perhaps being the singleminded focus of Timothee Chalamet over the last year or so, if you believe certain media reports.

Chalamet, in fact, is no longer the favorite to even win Best Actor, after Michael B. Jordan’s performance as both twins in Sinners has surprisingly won some awards leading up to the Oscars. Chalamet was trading at 80% just a few weeks ago, but now looks to be fighting an uphill battle. Ziddletwix informed me that Jordan’s win at the Actor Awards (formerly SAG), has shifted the odds in his favor. While only a narrow subset of the SAG voters select the Best Actor at the Oscars, it’s the most informative precursor award. The actors are also a big voting block across the Academy, so conditional on Jordan winning Best Actor, the odds of Sinners jump up a bit to win Best Picture as well:

Leo DiCaprio had appeared to be mounting a strong campaign last year, but since 2026 started has not shown much life in the market. All three of these actors have something in common, in my estimation: all leaned into the cringe of their characters. Jordan’s role(s) received mixed criticism. Obviously as the betting market favorite to win best picture, he was great, but many described his performance as, well… “corny.” But that perhaps might not be a bad thing. Both Chalamet and DiCaprio also played corny characters. DiCaprio staggered around the film in a bathrobe and Chalamet was shameless not just within the frame, but in his method acting sideshow stunts off screen. Perhaps moviegoers are developing a taste for more authentic — read: cringeworthy — performances?

Or perhaps not:

Emma Stone, despite getting into all sorts of cringe, as well as physical trouble, in Greek auteur Yorgos Lanthimos’ latest hit, finds herself on the outs for Best Actress. Same with Reinsve, the 2022 winner (which did not happen, once again). Jessie Buckley, of the emotionally wrought Shakespeare biopic (that also might be a very dark horse in the Best Picture category), Hamnet, looks to have locked up her first Best Actress. Perhaps this is a factor of a favorable direct comparison, as she played alongside the less talented heartthrob Paul Mescal, who has not been nominated for an acting award this year.

The supporting categories, or perhaps the “villain” categories, are a little more open. The extremely villainous Sean Penn is edging out the only-slightly villainous Stellan Skarsgård in the actor category, with Amy Madigan, who was disturbingly villainous in indie horror flick, Weapons, a slight favorite in the actress category.

According to Ziddletwix, her lead is a bit of a surprise. Three horse races in the acting categories are also quite rare, and goofy, campy antagonists tend not to win Best Supporting Actress, but there’s always a first. Ziddletwix also told me that an interesting thing to note with the supporting acting categories is the increasing screen-time over the years. While someone used to be able to win the category with just a few minutes on-screen, the categories now reward longer performances that in the past might have snuck into the main acting categories.

Rounding out the awards that people tend to pay the most attention to is Best Director. Paul Thomas Anderson looks close to a lock to win the award, which would shockingly be his first Academy Award of any kind (and perhaps a motivator for voters).

The Other Categories

Rounding out the Oscars are the categories that only film nerds and true degenerate gamblers care about.

The screenplay awards will likely go to Sinners (original) and One Battle After Another (adapted).

KPop Demon Hunters will win Best Animated Feature (your kids are breathing a sigh of relief), and also best Original Song for ‘Golden’. I finally managed to get my girlfriend to stop humming ‘Golden’ after months of psychological torture, so I’m turning off the sound during this portion of the Oscars so that there’s no chance this gets incepted back into our household once again.

Frankenstein appears likely to sweep Production Design, Costume Design, and Makeup & Hairstyling.

Sinners looks likely to take Best Score (I’m shocked that it’s only at 91% odds, to be honest, it was a movie built around a score almost as much as a score built for a movie).

Sentimental Value is locked in a close race with The Secret Agent for Best International Feature, and It Was Just an Accident, maybe the best film to get snubbed for Best Picture, has an outside chance.

One Battle After Another is also the favorite for the Cinematography award and Best Editing, while F1, the likely runner-up in Best Editing, is favored in Best Sound. Lots of nice “vrooms,” perhaps?

The new category this year is Best Casting, and Sinners holds the edge here, with a few films in contention, somehow.

Finally, the abysmal (sorry) Avatar: Fire and Ash will take home Best Visual Effects.

As for the Shorts awards, traders aren’t that confident. Ziddletwix described them as hard to predict. There are no good precursor awards, no box office data, and no good rumor mill to glean insights from. But perhaps the market serves as a good proxy for which short you should watch if you only have time for one. In the Live Action category, that would be… Two People Exchanging Saliva, I guess.

2027

I cajoled Ziddletwix into making a Best Picture market for next year. So far, the “Other” category is a heavy favorite, despite a lot of rumored good films being added. It’s hard to predict the Best Picture a full year out, before most of the contenders have even come out with a trailer.

That being said, Christopher Nolan’s The Odyssey and Joel Coen’s Jack of Spades lead early in the running, along with Romanian-Norwegian drama Fjord, Iñarritu comedy Digger, and the much-hyped sci-fi flick starring Ryan Gosling, Project Hail Mary.

Get your bets in now and happy forecasting!

-Above the Fold

Today is the first test in the Democratic Party’s attempt to win back the Senate this fall. This is an uphill battle, by all accounts. Retaking the Senate requires not just winning all of the states they’re favored in (including flipping North Carolina and Maine, as well as holding onto tough seats in Michigan and Georgia), but also winning two more races in challenging territory. Their likeliest options here are probably Alaska, Iowa, Ohio, Nebraska (where they’ve stepped aside for independent candidate Dan Osborn, that is) and… Texas.

It’s reasonable to litigate which of these pickups are more likely than others. In fact, Manifold traders seem to think that Alaska, Ohio, and Iowa represent slightly easier pickup opportunities than Texas, but the democrats need at least two of these reach states to take the Senate. Texas, the second most populous state in the union, would be a big prize for other reasons, including down-ballot races that have the chance to burst through the republican-gerrymandered house map. Conditional on the democrats taking back the Senate, there’s probably something like a 75% chance that the final map runs through Texas (someone should make a market on this), and if the democrats do win Texas, their victory is all but guaranteed.

Texas Senate Primaries

Tonight, the polls close in a couple hours, by 9pm ET for the westernmost tip of Texas. On the dem side of the ballot, there has been a closely fought race between Rep. Jasmine Crockett and State Rep. James Talarico. Manifold traders think that Talarico’s odds are better than Crockett’s in the general election, perhaps due to some combination of his appeal to moderate voters and less contentious style of politics.

As Nate Silver and others have noted, Talarico’s odds on prediction markets have outpaced what you might infer from the raw polling data. Until a couple days ago, the polling showed, if anything, Crockett with a slight lead, although there were nits to pick about the biases of internal polling from both of the candidates’ campaigns. Certainly, it didn’t seem like the 70-30 election that prediction markets had priced it at.

Well, late-breaking polls and some amount of tea-leaf-reading on the day of the election has now thrown Crockett’s odds down to about 11%, before any ballots have been counted. Much of the voting in the Texas primary is through early balloting, it should be noted.

On the Republican side, the race seems marginally closer. The firebrand MAGA-aligned Texas attorney general, Ken Paxton, is primary-ing the senior senator out of Texas, John Cornyn, with Rep. Wesley Hunt likely to end in a distant third, but also likely to act as a spoiler to send the race to a (costly) runoff. Most of Cornyn’s probability mass in winning the election and holding onto his seat runs through that potential runoff.

Indeed, there’s a distant chance of a runoff in the Democratic primary as well, due to an “obscure third candidate,” Ahmad Hassan.

You can also forecast on the margin of victory in the dem primary, which might be the more interesting question given the steep odds at this point.

The creator of all these Texas elections questions, Manifold user Jack1, looks to have a sharp understanding of the state of the races. He made over 100k in profit last month, topping Manifold’s Masters’ League. His biggest profit on a single market came from betting on the primary race in Texas:

Looking Forward to the Fall

There has been a slight uptick in the democrats’ odds of winning the Senate election this November, given the increased likelihood of a matchup between Talarico and Paxton, the most favorable matchup for the dems on paper.

A parlay on Graham Platner and James Talarico both winning their primaries has also risen to 70%. The two candidates make for interesting bedfellows as very different rising stars in the Democratic party. Talarico is known for his poised demeanor, his campaign’s focus on his Christian observance, and legibility to members of the other party. Platner is known for his across-the-aisle appeal as well, but also his past (and present?) dalliances with anti-Semitism, from his Nazi tattoos to his appearances with conspiracy theorist podcasters.

Platner, however, is the one with better odds this fall of unseating a republican senator. Manifold traders give him a 68% chance of beating Collins if he wins his primary.

Roundup

Dem odds of winning bicameral control of the legislature this fall is up to 38%, more than twice as high as the ~15% it was trading at for the first few months of the Trump administration.

This market on whether Greg Abbott will refuse to certify the Senate election this fall in Texas seems important and needs more than the 7 traders it has, but currently sits at 24%!

Finally, Talarico is already at 1.4% in the odds to be the Democratic nominee in 2028. Remember that other young, first-term senator who became president a few years back?

Happy Forecasting!

-Above the Fold

On Friday afternoon, Trump was at a Whataburger in Corpus Christi, declaring “hamburgers for everyone” and leaving with a bag bearing order number 47. A reporter asked how close he was to making a decision on Iran. “I’d rather not tell you,” he said. “Otherwise you’d have the greatest exclusive in history.” We now know he was being literal. Nine hours later, from Mar-a-Lago, he announced the start of a major attack.

It was, notably, also from Mar-a-Lago that he’d ordered the toppling of Nicolás Maduro in January. The weekend getaway apparently comes with options.

This also wasn’t a last-minute call, like many are speculating. An Israeli security official told Reuters the operation had been planned for months, the launch date locked in weeks ago. Which makes the diplomacy running in parallel somewhat interesting. Oman’s Foreign Minister announced a “breakthrough” with Iran just the day before the strikes, describing peace as “within reach.” Iran had apparently agreed to halt uranium stockpiling and allow full IAEA verification. Twelve hours later, the Supreme Leader’s compound was rubble.

Whether Iran was negotiating in good faith while the US had already committed to attacking, or whether the US was negotiating in good faith while Israel had already committed to attacking, is a question that will occupy historians and at least one Omani diplomat for some time.

What the markets had already worked out was that the direction wasn’t going to be a deal. The strike-by-June market had been sitting at 60-70%. The strike-by-February-28 market specifically sat at 17% days before: low, but not zero, which is the honest read of how surprising this actually was.

Six Polymarket accounts correctly positioned on the specific date and walked away with around $1.2 million.

Khamenei’s Theory of the Game

To understand what happened Saturday, it helps to understand what Iran, at least thought, it was doing.

After the June 2025 ceasefire (the 12-day war that ended with Israel having struck nuclear facilities and Iran having hit a US base in Qatar), the regime’s official position was that they had won. Khamenei gave a televised address declaring victory because the Islamic Republic had survived, and because Washington sought the ceasefire first. The logic held a certain internal coherence (at least Iran thought so): survival equals victory, American restraint equals Iranian strength.

What followed was the regime’s attempt to translate that logic into strategy: rebuild missile capacity, keep the nuclear program alive, manage enough diplomatic engagement to delay the next round while buying time.

The talks in Oman and Geneva weren’t necessarily capitulation more than they were time. The problem is that Trump had apparently decided, sometime in the weeks before February 28, that time was no longer on offer.

Khamenei was confirmed dead early Sunday morning, Iranian state media breaking the news that the Supreme Leader, who had ruled since 1989 and survived sanctions, wars, protest cycles, and decades of “the regime is on the verge of collapse” takes, had been killed at his compound in Tehran while “carrying out his duties.” The broadcaster announced his “martyrdom” visibly in tears. The government declared 40 days of mourning.

Seven senior defense and intelligence commanders were killed in the opening strikes. The succession question, which everyone knew was coming eventually, is now live under active bombardment (quite literally). The constitution provides for an interim three-member council (president, judiciary chief, a Guardian Council jurist) to manage affairs while the 88-member Assembly of Experts selects a new Supreme Leader. Ali Larijani, longtime Khamenei confidant and former parliament speaker, emerged as the most senior civilian official still standing and immediately vowed to “stab America in the heart,” which is roughly what passing the succession test looks like in this particular constitutional moment.

Manifold’s regime fall market has moved to 47%. In December it stood as a rounding error.

Celebrations and Condemnations

The reaction inside Iran was not unified, which is worth understanding before reaching for easy conclusions.

Celebrations broke out across Tehran, Isfahan, Karaj, Shiraz, Qazvin: cars honking, music playing, people gathering on rooftops as smoke rose over what had been the Supreme Leader’s compound.

The Iranian diaspora had been mobilising for weeks: 350,000 people in Toronto and another 350,000 in Los Angeles on February 14, Reza Pahlavi’s designated “global action day,” the largest Iranian diaspora mobilisations in history. Many had been explicitly calling for American military intervention. On Saturday they largely got what they asked for, and the celebrations in Westwood and across European cities reflected that.

Inside Iran, the picture was more complicated. Pro-regime crowds gathered in Tehran on Sunday morning. Mourners wept at the Imam Reza shrine in Mashhad. The government had killed thousands of protesters just six weeks ago, people who had taken to the streets over a collapsing rial and stayed for regime change, and the same security apparatus that carried that out remains intact. Basij forces reportedly opened fire on some celebrations. The internet blackout, near-total since January, means most Iranians are still receiving information through state channels or satellite feeds.

Internationally: Russia called the killing a “cynical murder.” China condemned an “undisguised assault on sovereignty” and noted, with some accuracy, that attacking “at the very moment diplomacy shows promise sends a dangerous message.” At least ten protesters were killed storming the US consulate in Karachi. The EU called it a “defining moment” and began working on “practical steps for de-escalation.”

The retaliation is already underway. Iran has hit US military bases across Bahrain, Qatar, Kuwait, the UAE, Jordan. Houthis resumed Red Sea operations. Larijani warned Gulf states that allowing their territory to be used for strikes made them “legitimate targets,” a notable thing to say to the countries hosting both your only mediator and, via the Strait, your only oil customer’s supply chain.

The IRGC Problem (And the Maryland One)

What is important to note is that killing Khamenei is not the same as regime change. The Council on Foreign Relations put it plainly: “The Islamic Revolutionary Guard Corps is the regime.”

The IRGC controls an estimated 30-40% of the Iranian economy: telecoms, energy, ports, significant portions of the black market. Khamenei was its ideological anchor; the mid-level commanders are its operating system. Many of them, in the Foreign Affairs framing, were less Khamenei’s subordinates than his partners. They have weapons caches, proxy networks, and no particular incentive to dissolve absent a political settlement. The most analytically likely successor scenario, according to some, isn’t a democratic transition. It’s an IRGC strongman consolidating power under the cover of wartime emergency, running a harder, more militarised version of what already existed.

Reza Pahlavi, 65, has lived in exile in Maryland for 48 years, longer than the Islamic Republic has existed (he left at 17, so make of that what you will in terms of governing preparation). He called Saturday a “humanitarian intervention” and urged Iranians to “resume protests as the Islamic Republic collapses.”

Traders now gives him a 39% chance of physically entering Iran by year’s end.

Perhaps the more revealing market is the conditional one: if the regime falls, does Pahlavi lead the new government a year later?

A supposedly 80% approval rating in diaspora polls doesn’t straightforwardly translate to governing a post-conflict state where the IRGC still controls the ports.

The succession process carries its own uncertainty. There have been exactly two Supreme Leaders in the Islamic Republic’s history: Khomeini, then Khamenei, who was notably underqualified for the role when elevated and had the constitution quietly amended to make him eligible. The Assembly of Experts must now convene under active bombardment to select a third. As Marco Rubio noted in January, “no one knows” who would take over. That assessment was made before the question became urgent.

Trump said the bombing campaign would continue “throughout the week or longer.” The stated objectives, destroying missile infrastructure, eliminating the nuclear program, producing regime change, are not all equally achievable from the air, and at least one former defense secretary has called the endgame “fuzzy.”

Whether this ends with a negotiated settlement, an IRGC-led successor state, or something genuinely new depends almost entirely on whether the security apparatus fractures. Signs of that at this point are murky. There is, however, a 47% market on whether it happens by December.

Happy Forecasting!

- Above the Fold

When pollster David Shor investigated the favorability of Anthropic late last year, he found that they had about the same name recognition as a fictional AI company, “Apex Logic.”

Well, now they’re trending towards becoming a household name, but it might not be for the reasons they were hoping.

After it came out that the Department of Defense had made use of Anthropic’s offerings during the raid on Maduro, there was understandably some internal blowback within Anthropic on potential misuse of Claude. Through a sequence of miscommunications between Anthropic and Pentagon, this escalated into a dispute over whether the US military could compel Anthropic into allowing unfettered use of their AI tools.

To those tuned into Anthropic’s approach to constitutional alignment, this might seem like an impossible ask. It’s quite plausible that even without direct action on the part of Anthropic to deny US military uses, Claude could reject certain tasks due to its internal “moral compass.” The US government isn’t used to dealing with tools that can take their own initiative to reject certain uses. On top of this, Anthropic prides itself on careful and deliberate ethical restrictions, which clashes with potential DoD use cases such as… y’know… surveillance, guided weaponry, espionage, or fomenting coups.

In any case, after a couple of meetings between Anthropic CEO Dario Amodei and DoD Secretary Pete Hegseth, the situation has come to a head. The military has given Amodei and Anthropic until 5:01pm on Friday to remove certain restrictions (which may or may not be possible to do). They have threatened several possible actions:

• Cutting ties with Anthropic and ending their contracts for the use of its products.

• Declaring Anthropic a “supply chain risk” and mandating other military contractors to abandon their use of Anthropic’s offerings.

• Invoking the Defense Production Act to compel Anthropic to produce a restriction-free version of Claude (which some have called “WarClaude”)

As Dean Ball, one of the authors of the administration’s AI action plan, has noted, these threats appear paradoxical. Either Anthropic is such a supply chain risk that no one should use it, or they’re so essential that they should be compelled to become an essential part of the military supply chain.

As a result, forecasting this situation is challenging. There aren’t clear delineations between the various actions that the Trump administration may take, and it’s possible that several actions may be taken, rescinded, or modified by the time the dust settles.

Manifold traders think the most likely outcome is something resembling “business-as-usual” after the clash: ~60% odds to Anthropic not making the requested changes. For what it’s worth, the Pentagon seems to have denied they have even made these explicit requests, contra previous reporting from Axios and others.

Despite this, traders only give about 1 in 3 odds that the Pentagon and Anthropic will reach a mutual settlement by the Friday deadline.

Traders think that some of the more escalatory outcomes are between 20 and 30% likely to occur: the Pentagon cutting ties with Anthropic, declaring them a supply chain risk, or Anthropic pursuing legal action. An invocation of the Defense Production Act, despite threats from the administration, is at only a 15% chance to occur.

The market also is skeptical that Anthropic will actually remove any safeguards (if these safeguards are something that can be removed). Suspiciously, the probability for that action is quite similar to the probability that multiple Anthropic researchers resign over this affair.

A separate market on specifically whether Anthropic will give the military “unfettered access” to Claude by their deadline is trading at just 6%:

Another market forces traders to pick the closest of three options for the Pentagon: continuing to use Anthropic’s softare, ending their contracts, and ending their contracts and additionally labeling Anthropic a supply chain risk.

Here we see about 70% odds that Anthropic continues to contract with the DoD, with traders believing that conditional on the DoD ending their contracts with Anthropic, it is likely that they also take the step of labeling them a supply chain risk, a label generally reserved for the companies of foreign adversaries!

Change Comes for Anthropic

It’s probably good to contextualize these events within the larger shifts happening at Anthropic. As the company positions itself as perhaps the leader in the AI race, with their current frontier model remaining slightly ahead of competitors by most expert assessments, they are also hoping to ensure a safe technological transition. Recent profiles on the company, including on their head of alignment, Amanda Askell, have focused on their dueling motivations. On one hand, Anthropic is racing to maximize earnings and secure sizable investments; on the other hand, they have clear goals set by their employees and executives —and indeed their very structure as a public benefit corporation — to ensure beneficial societal impacts.

Manifold users are very interested on how this will affect the near future for Anthropic, as they grow in size to rival OpenAI. One proxy could be this market on whether their head of alignment, Amanda Askell will remain in a position of leadership at the company through the end of next year.

Having goals that are perceived as orthogonal to the current US administration will put them into tension with the government, and it’s indeed possible that the publicity around Anthropic’s idiosyncratic ethical framework was what led to their current fight with Pentagon leadership.

Anthropic recently revised their “responsible scaling policy,” triggering a round of discourse on how companies should adjust their prior commitments when those commitments have become less practical or relevant. Some people have praised their transparency in amending the document, while others have called it out for no longer upholding tenets from prior versions.

Zvi Mowshowitz has recently created a market on whether Anthropic’s responsible scaling policy will be further revised over the next 3 years, with changes larger than its revisions in the current version. Traders give it 2 in 3 odds to occur.

Frontier Claude

Why is Anthropic attracting such attention from multiple sectors? Well, as even Pete Hegseth has reportedly said, they’re simply the best right now.

Despite having only ~1 in 4 chance to surpass OpenAI’s valuation this year (depending on how you look at it, that could actually be quite high!), Anthropic appears more narrowly focused than their competitors on developing highly intelligent, frontier models.

They recently released Claude Opus 4.6, which beat its benchmarks so hard that METR time horizon questions on successor models, that had buckets made before its release, now look like this:

Opus 4.6’s METR time of over 14 hours blew away projections to the point that it might have been a trigger for the recently viral Citrini article on AGI economics (personal speculation here), and was certainly a trigger for endless exponential curve projections on Twitter.

With top labs keeping up a blistering pace, traders expect Claude 5 (perhaps both Opus and Sonnet) to be released between April and July.

Happy Forecasting!

-Above the Fold

That qualifier might sound dramatic, but markets give 52% odds that FIFA cancels matches in Jalisco due to cartel violence, and 18% that at least one US ally boycotts the American portion of the tournament.

Whether all three host nations actually stage their assigned games is a legitimate question rather than the usual pre-tournament anxiety about hotel availability.

The football itself should be good. Spain leads betting at roughly 17%, England at 13%, France at 12%. The expansion from 32 to 48 teams means more matches and arguably more uncertainty, though markets give an 80% chance a previous winner takes it. Either history matters or traders seemingly lack imagination :(

Japan qualified first among non-host nations and sits at 1.2% despite beating both Germany and Spain in 2022 (while their odds of playing in the quarterfinals are more respectable, at 15%).

Markets price a non-European or South American finalist at 14%.

The US benefits from home advantage and drew a favourable group. Markets give them a stronger 75% chance to advance from the group stage but essentially zero chance of winning the tournament.

Mexico sits at similar odds, with Canada being the longest shot out of the three.

Golden boot odds favour Mbappé, followed by Kane. There’s an 90% chance someone scores 6 or more goals, 65% for 7 or more.

The scoring race matters beyond tournament MVP considerations now. It’s become the primary metric for Ballon d’Or voting in a world where nobody quite dominates the way Messi and Ronaldo did for nearly two decades.

FIFA is also introducing blue and purple cards. Blue for tactical fouls, purple for dissent. Players get 10 minutes off for either. Markets are sceptical about consistent enforcement, which seems reasonable given FIFA’s track record on following its own rules.

How to Give a Peace Prize

In December, FIFA President Gianni Infantino created a peace prize specifically to give it to Donald Trump at the World Cup draw (allegedly, of course, although markets might have been a bit more straightforward with their thinking).

Interestingly enough (not really), traders agree that FIFA loved Trump so much that they will likely honour him by not handing the prize to a new winner this year. Although, hopefully, we will get to see Trump win it again (a possibility standing at 3%)!

The trophy was gold-plated, roughly the size of the actual World Cup trophy. Trump’s team allegedly insisted on this sizing. The certificate praised Trump for “bringing peace” to the world. Trump called it “one of the great honours of my life.”

Three weeks later, US forces seized Venezuelan President Nicolás Maduro. Trump has since threatened to annex Greenland, imposed travel bans on qualified World Cup teams including Senegal, Ivory Coast, Iran, and Haiti, and overseen immigration crackdowns that sparked mass protests in Minneapolis and other US cities.

Former FIFA President Sepp Blatter backed calls for fans to boycott US-hosted matches. “For the fans, there’s only one piece of advice: stay away from the USA!” Mark Pieth, the Swiss attorney who chaired FIFA reform oversight a decade ago, said the same thing.

The peace prize sparked formal ethics complaints. A London-based rights group argued Infantino violated FIFA’s political neutrality mandate. FIFA’s response was to set up an office in Trump Tower in July and have Infantino show up to Trump’s Board of Peace meeting last week wearing a red “USA” hat.

Trump has kept the actual FIFA Club World Cup trophy as an Oval Office decoration since Infantino gave it to him in March. Chelsea had to hoist a replica when they won.

Infantino defended the decision, saying that Trump “objectively deserves it” for being “instrumental in resolving conflicts and saving thousands of lives.” He added that FIFA should “never ban any country from playing football because of the acts of their political leaders,” which is an interesting position given Russia’s ongoing suspension for invading Ukraine.

The tournament runs June 11 to July 19. The final is at MetLife Stadium in New Jersey. Ticket prices sparked outrage: hundreds of dollars per match. Blatter called it a “slot machine” operation focused on “money, money, money,” which is rich coming from the man who ran FIFA during its peak corruption era but isn’t wrong.

The Boycott Question

Over 150,000 people in the Netherlands signed a petition calling for their team to withdraw. Germany’s football federation vice president said it’s time to “seriously consider” boycotting, comparing the situation to Olympic boycotts during the Cold War.

Markets give 9% odds that a NATO member boycotts the World Cup.

The actual federations have been more cautious. Germany, France, and the Netherlands all said they’re not planning to boycott.

UK MPs called for the US team itself to be excluded, citing similar reasoning to Russia’s ban: undermining the rules-based international order. FIFA’s contract allows them to “cancel, reschedule or relocate one or more matches (or the entire FIFA World Cup 26) for any reason” including “health, safety or security concerns.” Markets give 30% odds that at least one US WC game is moved because of Trump-adjacent politics, fair odds considering the fact that a clear contractual pathway exists.

Iran threatened to boycott the December draw after four of nine visa applications for their delegation were rejected. They eventually sent a smaller group. Teams from Haiti face similar restrictions. Markets are thin on whether Iran gets formally banned (or whether they withdraw themselves), but there’s precedent.

February in the Premier League

Arsenal are 5 points clear at the top of the Premier League. Manchester City have a game in hand. It’s mid-February.

Markets give Arsenal 54% odds of winning the title, City 42%. That 12-point gap feels smaller than 5 points on the table would suggest. Traders have seen this before. Twice.

In 2022-23, Arsenal were 5 points clear in February with 12 games left. They dropped an astonishing 21 points from that position. City won the title. In 2023-24, Arsenal led through February before City’s late surge. Liverpool actually won last year, but Arsenal and City both finished ahead of them in the “February onwards” table that seems to matter.

The pattern, which Arsenal fans seem acquainted with at this point, is specific. Arsenal build leads through autumn when City are rotating through Champions League fixtures. Come February, City traditionally click into gear while Arsenal start dropping points to teams who’ve figured out how to sit deep and counter. Brentford last week: 1-1 after taking the lead. Wolves on Wednesday: 2-2 after being 2-0 up.

City’s recent form tells the story. Comeback win at Liverpool: 2-1, Haaland penalty in stoppage time after going behind to a Szoboszlai free kick that could have made it 9 points to Arsenal. Then 3-0 against Fulham.

The fixture markets have long been pricing as decisive pricing as decisive: April 18 at the Etihad. Arsenal travel to Manchester with presumably a smaller lead. City with momentum and home advantage. City with the knowledge that Arsenal have historically crumbled in this exact situation.

Haaland’s drought is the subplot everyone’s watching. Only 3 goals in 13 appearances in 2026. Just penalties from open play. His overall season numbers remain strong: 25 goals in his first 23 games, but the February barren spell is new.

This matters for the Ballon d’Or odds more than you’d think.

Haaland sits at 4%, down from a high of 20%, and Mbappé at 21%. The Messi-Ronaldo era set expectations so absurdly high that everyone’s still adjusting to what normal excellence looks like. Those two scored 50+ goals per season like it was routine. They won 13 Ballon d’Ors between them across 15 years. The award was basically a binary choice between two players having seasons that would have been all-time great for anyone else.

Mbappé’s 21% is largely contingent on France having a deep World Cup run.

Elsewhere: Tottenham are 16th, five points above relegation, having just appointed Igor Tudor as interim manager after sacking Thomas Frank eight months into his tenure. Eight managerial changes this season so far. Manchester United sit 4th under interim boss Michael Carrick after Ruben Amorim’s departure. The managerial carousel spins while Arsenal and City play out their annual ritual.

Five points with a game in hand. Fourteen fixtures remaining. Arsenal have been here before. Markets remember even if the table doesn’t show it yet.

Happy Forecasting!

- Above the Fold

MNX is a decentralized futures exchange targeting sophisticated traders and focused on the AI economy. The site will have prediction markets, but the focus will be on futures and perps.

MNX is a separate company from Manifold entirely. Manifold is not going away or changing - we love the site and the community, and we’re going to continue building it.

Manifold will remain the place for fun, user-created markets, and a casual, social atmosphere. A place where you can ask the questions that are on your mind, raise the stakes on your personal goals, find outside-the-box uses for markets, and join a community of traders with common curiosities.

MNX is the venue for more serious trading on a smaller, curated subset of markets. The two are complementary: On Manifold, you can put your Mana on whether your friend will land a job at Anthropic. On MNX, you can bet $100k on Anthropic’s valuation.

We’ve spent four years learning what prediction markets are good at and where they fall short. MNX is our attempt to take what we’ve learned and apply it to serious financial infrastructure for the AI economy.

We’re incredibly excited about it.

Check out our launch thread on Twitter for the full story, and follow us there for more updates.

— @StephenGrugett & @realpalebone

The theory exists because the alternative explanations are being met with…skepticism. Epstein was, by all accounts, shite at the job he claimed to do. He presented himself as a financier managing billions, but former colleagues and financial analysts who’ve examined his trades describe someone who couldn’t execute basic strategies competently.

The handful of known investments he made were unremarkable. His claimed methods for wealth generation: tax arbitrage, currency trading, or whatever story he was telling that week lack much holding.

The money came from somewhere. His lifestyle required high, sustained access to capital that his documented business activities fail to explain.

The files mention Russia 5,876 times and Putin 1,055 times. Most of these references show Epstein positioning himself as an intermediary who could connect Moscow to Western power brokers. He told people he could arrange meetings with Putin. There’s no evidence any such meeting occurred. But the pattern of behavior, the persistent attempts to insert himself into political networks, the collection of kompromat-style material on associates, the unexplained wealth, reads differently depending on whether you think he was acting alone or with institutional backing.

Markets give 12% odds he was specifically a Russian asset, 22% that Putin was involved in his operation.

The Dossier Center uncovered his close contacts with Sergei Belyakov, former Deputy Minister of Economic Development and FSB Academy graduate. Maria Drokova, former press secretary for Kremlin-backed youth group Nashi, appears 1,627 times in the files, mostly coordinating Epstein’s media contacts.

The Kremlin laughed it off. Spokesman Dmitry Peskov said he was “tempted to make a lot of jokes about that theory” but wouldn’t “waste our time.” Which is roughly what you’d expect whether the theory is nonsense or uncomfortably accurate.

The more interesting question is what he was trying to accomplish with all those connections. The emails show him offering to arrange meetings between people who had no obvious reason to want his help. Modi and Steve Bannon hours after India’s 2019 election. Trump access for various associates. He discussed foreign policy with Indian billionaire Anil Ambani, lobbying heavily for India to embrace Israel as an ally. After Modi’s 2017 visit to Israel, Epstein wrote to someone that “The Indian Prime minister modi took advice.”

Whether he was a spy or just an unusually well-connected criminal with delusions of influence is uncertain/ The no-intelligence-asset scenario requires believing he was simply that good at ingratiating himself with powerful people while being catastrophically bad at the day job he used to explain his presence. It’s possible.

The Appointment

The United Kingdom is discovering that hiring people with interesting Rolodexes carries certain risks.

Peter Mandelson spent decades as Labour’s behind-the-scenes operator, earning the nickname “Prince of Darkness” for his ability to manage difficult situations.

Last December, Prime Minister Keir Starmer appointed him UK ambassador to Washington despite knowing about his friendship with Epstein. Starmer’s logic was pretty straightforward: Mandelson had the contacts, the experience, the ability to work with Trump’s team. Epstein merely….served as a middleman.

The latest file release showed those contacts included roughly $75,000 in payments from Epstein in 2003-2004 and what appears to be advance notice of a 500 billion euro bank bailout in 2010. Police raided two of Mandelson’s properties last week. He’s under criminal investigation for misconduct in public office.

Starmer fired him in September, but the damage was done. This week, his chief of staff Morgan McSweeney resigned Sunday, followed by communications director Tim Allan on Monday. Anas Sarwar, leader of Scottish Labour, called for Starmer’s resignation Monday afternoon. By Monday evening, Starmer faced a packed room of Labour MPs deciding whether to keep him in office.

He survived, but the mechanism of survival tells you everything about British politics in 2026. Starmer held on not because MPs believed in him, but because gilt markets couldn’t handle the alternative.

Parliament rallied behind him to avoid spooking investors, which is perhaps not the strongest foundation for political legitimacy.

Prince Andrew, stripped of his titles and evicted from Royal Lodge after previous Epstein revelations, now faces investigation under the Official Secrets Act for allegedly sharing trade documents with Epstein during his time as UK trade envoy. The files show he forwarded visit reports from trips to Singapore, Hong Kong, and Vietnam, seeking Epstein’s “comments, views or ideas” on investment opportunities.

Traders are still saying that this isn’t devastating for the future of the monarch as a whole, but it’s sparking discussion at the least.

The 2029 election looms over all of this. Markets already priced in elevated uncertainty about whether Labour can hold its majority, but the Mandelson scandal compounds existing pressures. Starmer’s approval ratings were historically low before this crisis.

“Barely Knew the Guy”

Representatives Ro Khanna and Thomas Massie spent two hours at the Department of Justice on Monday reviewing unredacted Epstein files. They found six men whose names had been redacted in the public release with no clear legal justification.

One was Leslie Wexner, the Victoria’s Secret billionaire. The FBI labeled him a “co-conspirator” in a 2019 document. He was never charged. When Khanna and Massie pointed this out, the DOJ unredacted Wexner’s name along with five others: Salvatore Nuara, Zurab Mikeladze, Leonic Leonov, Nicola Caputo, and Sultan Ahmed bin Sulayem (CEO of Dubai Ports World).

“If we found six men that they were hiding in two hours,” Khanna said Tuesday on the House floor, “imagine how many men they are covering up for in those three million files.”

The DOJ’s response was that they “just uploaded whatever the FBI sent.” The FBI had already scrubbed the files before sending them to Justice. Seventy to eighty percent of the documents remain redacted even in the “unredacted” version members of Congress can now access.

Commerce Secretary Howard Lutnick is learning that proximity has costs, though apparently not career-ending ones. Last year on a podcast, Lutnick described visiting Epstein’s home in 2005, seeing a massage table, and deciding to “never be in the room with that disgusting person ever again.”

The files show Lutnick sought drinks with Epstein in May 2011. In December 2012, Lutnick visited Epstein’s private island with his wife, four children, and nannies. They had lunch. Both men invested in the same business that month.

Tuesday’s Senate hearing was something to watch. Lutnick told senators he “barely had anything to do with that person.” He acknowledged the 2011 meeting and 2012 island visit but maintained nothing “untoward” happened. When pressed on why he’d take his family to visit someone he’d called disgusting, Lutnick said it was part of a family vacation and they were just stopping by.

Representatives from both parties called for his resignation. The White House said Lutnick “remains a very important member of President Trump’s team” and that “the president fully supports the secretary.”

The Clinton depositions are scheduled for February 26-27, ending six months of resistance to congressional subpoenas. Bill Clinton flew on Epstein’s plane at least 26 times across multiple trips. The files include photos of him with Ghislaine Maxwell and in a hot tub with someone DOJ describes as an Epstein victim.

The Clintons fought the subpoenas for months, claiming they were politically motivated. As contempt proceedings moved forward, they caved. Now they’re demanding public hearings instead of depositions. Hillary Clinton tweeted: “If you want this fight, @RepJamesComer, let’s have it — in public. You love to talk about transparency. There’s nothing more transparent than a public hearing.”

Markets give her 37% odds of getting that televised hearing, along with interest in the spectacle than in whether either Clinton invokes the fifth amendment or executive privilege during testimony.

Prison time in 2026 sits at 6%, which tells you everything about how seriously traders take the prospect of consequences.

As for Trump, Representative Jamie Raskin noted he’s mentioned “more than a million times” in the unredacted files. Markets give low odds he faces charges related to Epstein.

The internet, naturally, focuses on the more important stuff.

There’s a market asking whether Trump and Clinton had sex with each other, currently trading at 1.5%.

“This is bringing down the British government,” Khanna told reporters. “It may bring down the monarchy. What are we doing here in the US to stand up to the Epstein class?”

The files are, after all, about American networks. They document American crimes on American soil. Yet somehow the sharpest consequences are landing elsewhere. Three million pages, mostly about the US, causing crises everywhere else.

Happy Forecasting!

- Above the Fold

February 1, 2015 birthed a lot of Monday Morning Quarterbacks, and as a Seattle native, I’ll just say it’s not my favorite NFL memory.

The Seattle Seahawks faced the New England Patriots in Glendale, AZ. With 27 seconds on the clock, the Hawks trailed by four points and found themselves with the ball at the one-yard line. Seems like a lock, right?

As Seattle fans started preemptively celebrating their second Super Bowl win in a row (they beat the Denver Broncos in 2014), QB Russell Wilson threw on a goal-line pass play that was… intercepted by the Patriots. The win gave the Pats their first Super Bowl ring since 2005 and left Seattle in a shambles.

*Annie Duke’s 2018 book Thinking in Bets opens with a chapter about this infamous play, the vitriol coach Pete Carroll received as a result, and serves as a canonical example of Resulting. It’s a worthwhile read if you’re interested.

Okay, this is Above the Fold, not ESPN (I haven’t forgotten), so why am I telling you all of this? Because a few hours from now, these two teams face each other again at Super Bowl LX in Santa Clara, CA.

Bowl Bets

Here’s a quick rundown of a few of the markets on the Manifold Super Bowl XL Dashboard:

For now, the Pats have a 31% chance of claiming another victory over the Hawks

There are a couple of other markets to arbitrage here and here.

If you’re into Prop Bets, you have more than 100 options to bet on and The Official will be resolving live

In it for the ads? Bayesian has a market for that! There have been rumors, teasers, and reveals already, but who will show up on the day?

Speaking of showing up, there’s a 33% chance TSwift is at the game

And a 69% chance Trump whines like a pathetic little bitch about halftime headliner Bad Bunny during the game

I, for one, assume the odds should rise if Bad Bunny wears a dress during the Halftime performance

Or if Bad Bunny makes a political statement

Bet on other things like:

• Will Super Bowl LX set a new all-time record for US television viewership? (31% chance)

• Super Bowl 2026: Color of Liquid Poured on Winning Head Coach (Green/Yellow is in the lead)

• Will there be a punt returned for a touchdown in Super Bowl LX? (only a 6% chance - if it happens, it would be the first time in Super Bowl history)

• Will the Super Bowl Final score be a Scorigami? (Scorigami is a final score that has never been recorded in NFL history, a 3% chance of that happening)

And a lot more.

Now, if you’ll excuse me, it’s a big day here in Seattle, I’m betting on a Redemption Arc, and I need to go grab a comfortable front row seat to lose my voice from.

May the Hawks odds be ever in your favor!

-Stef