A lucid observation of Kleros and the path to follow

yadf1994 · June 7, 2026, 6:54am

I worked at Kleros for five years during my doctoral studies. I’m currently a postdoc researcher at Stanford. I wanted to share my perspective with you, in order to start a debate that I believe is fundamental and with it hoping to bring about a radical and immediate change in Kleros’s strategy. My motivation for sharing my reflections with you in article form stems, firstly, from the current situation and, more specifically, from recent discussions within the Kleros discussion channels, which I have followed closely. Secondly, it is based on the conclusions I drew from my experience at Kleros during my PhD.

The observation

The current model of “decentralized justice” has not achieved the mass market adoption required for the long-term survival and growth of the protocol. I think Kleros has to pivot the efforts toward becoming the ultimate oracle for Prediction Market, placing those platforms such as Seer at the center of its ecosystem.

A lucid assessment of Kleros current trajectory

To build a successful future, we must look at our past and present results without complacency. The signals from our community and our usage metrics (such as PNK price, TVL, etc) clearly indicate an urgent need for change.

The loss to UMA (For Now): In terms of raw adoption and value capture as an optimistic oracle, Kleros has, for the time being, lost the battle against UMA. While Kleros functions just as well technically (or even better), we have failed to secure considerable and recurring volume. The status quo is no longer an option.
Lost lead on Prediction Markets (Gnosis - DXDAO): Kleros had a major historical and strategic advantage through its early work with DXDAO. We let this opportunity slip away, even though prediction markets are now one of the most explosive and high-momentum sectors in Web3 in particular for dispute resolution.
The Failure of Proof of Humanity (PoH): The lack of significant resources and focus poured into Proof of Humanity failed to create the intended economic leverage or utility for the Kleros protocol and the PNK token.
Among the positive points not to be overlooked: PoH reached approximately 20,000 registrations and was at one point more well-known than Kleros itself. Omen generated several million dollars in TVL and a $2.5 million lawsuit.

The failure of these projects was organizational, not technological : the protocol itself was never hacked and functions correctly from both a technical and game-theory perspective.

Diagnostic

The initial hypothesis that the market would massively adopt “decentralized justice” for dispute resolutions has proven to be wrong. Kleros possesses world-class technology and expertise, but we lack the killer app capable of making the protocol take off.

Moreover, the current runway is less than 5 years, with no significant recent revenue. A status quo will simply imply the end of Kleros.

A solution: Leveraging Prediction Market and Seer

Prediction markets are experiencing a massive take off. Futarchy (“voting on values, betting on beliefs”) could be the application of prediction market, cf https://www.lesswrong.com/w/futarchy\

Kleros must capitalize on this momentum and its historical expertise. Our best, and potentially only at the moment, opportunity for real and concrete growth lies in Prediction Market.

Notably, Seer already functions as a base layer with organic demand from institutions (EF, Gitcoin, RealT), customers acquired with very little sales effort. Second-layer projects like futarchy.fi and forks (Butter) are already expanding the ecosystem. Kleros’ involvement creates a multiplier effect: the Coop’s spending is amplified by that of Seer One, second-layer projects, and DAO clients using futarchy.

To conclude, here is a strategic pivot in few points:

Prediction Market as the priority: Most of the economic, marketing, and development efforts of the DAO should be primarily deployed to push and scale Prediction Market.
Kleros as the futarchy Oracle: Kleros should stop spreading itself thin across fragmented use cases and position itself strictly as the indispensable, specialized, ultimate oracle to resolve complex prediction markets and Prediction Market decisions. However, Kleros can continue its current profitable activities with a minimum level of resources required.

I hope my initiative will generate constructive exchanges, revitalize the community and unite the Kleros ecosystem and its community around an ambitious project for Kleros to make it!

clesaege · June 7, 2026, 9:19am

Being an oracle for prediction markets is “decentralized justice”. But maybe the branding as “subjective oracle” could be seen as more practical and less idealistic. Focusing more on what it does instead of how it does it.

I am obviously super bullish on Seer, but we should probably avoid putting something “at the center” if we want to keep attractivity to other projects.

Note that for now the strategy for Seer has been to be a base layer prediction market. This worked great to attracted other prediction markets projects on top of Kleros (futarchy.fi, DeepFunding, Will it merge?). Here prediction markets projects get Kleros arbitration as default (it would be an effort to remove it instead of it being an effort to integrate it). Note that this even brought projects not using Seer to Kleros as they were reusing elements of the codebase (cf Butter even if they do not seem active anymore).

Definitely agree on that: What can be measure can be optimized.

To give more context on that:

We got approached by Gnosis to be an Oracle for Omen. We also had a token swap (PNK - GNO, which ended very good for us as GNO increased a lot).
Omen was given the dxDAO.
Some dxDAO members wanted to develop their own oracle. I initially told them that if they were to remove Kleros from the plan, we would launch a version with Kleros as the Oracle.
I was then convinced to cut a deal with the dxDAO. It ended up with both oracles being available with Kleros as default and the coop giving some PNK for liquidity incentives.
It initially went very well with the US election. Kleros getting both its largest dispute and its PNK all time high.
The dxDAO failed to develop Omen. It was clear that some improvements needed to be made. I actively lobbied the dxDAO team to put work into Omen. It seems they would do it, but actually never did.
I think Shayne (of Polymarket who was a dxDAO member) lost confidence in the dxDAO team way sooner and went to make its own PM also using Gnosis initial work (conditional token).
I did try to get Polymarket to use Kleros, but they first went the central way and then with UMA which gave them some liquidity incentives.

Overall we were there early, but we probably were not enough aggressive.

Agree on that. UMA first had Across as a dapp they developed internally. Once you have big dapps, you get more likely to get more dapps.

Just to clarify it was a Kleros dispute (the terminology may make some people think it was a state court lawsuit ^^).

Revenue sources could be:

Further PNK grants (probably only be possible when PNK market caps gets significantly higher than coop treasury), but the recent KIP-88 hints in this direction.

When talking to crypto people, there is indeed way more interest for prediction markets (as it is now hyped and the majority of the dispute resolution space) than any other Kleros related products.

Maybe not “most” as we should still support previous projects as no one wants to use a subjective oracle used by almost no one. But I’d agree on “most of the new”.

I think you hint to that here too. Don’t break everything, but focus new efforts (and remove ressources from projects with low impact/cost ratio after some “last chance to show traction” period.).

jnptzl · June 10, 2026, 12:48am

Hey Yann, long time no see!

A few things that would help understand the case better:

UMA fee capture: how much does UMA actually earn per year from Polymarket disputes? And has it been growing or shrinking, given MOOV2 whitelisting and Chainlink + Pyth eating the price-resolvable markets? Does the equivalent to the Kleros Coop capture fees or only the holders?
Developer experience and product maturity: how do we compare to the competition (UMA, Chainlink, Pyth) on dev UX and overall product maturity? Is it on par? Right now our PM integrations run on Reality.eth + Kleros V1, with custom proxies built per additional network outside Ethereum. What are the biggest tech blockers we’d need to clear for a PM-focused push? How much time and development resources do you estimate it should take to get to a similar level of dev UX and product?

image2722×2014 280 KB
Cost to land Polymarket: how much did UMA spend in incentives to win that deal originally? Want to gauge what the cooperative might need to match for a comparable integration, since we have a finite treasury.
AI eating PM disputes: how likely do you think it is that AI-based oracles (Gensyn, UMA’s OTB, GenLayer) handle 99%+ of disputes by 2027? If that happens, what’s the residual role for Kleros as a PM oracle and how much (% of volume, value settled, etc) do you think it can capture?
Big disputes vs market cap: how do you think we should handle cases where the disputed value is larger than PNK’s market cap? Are we thinking Kleros AVS with restaked ETH, dual-token bonds, something else? How much should be put into development to accomplish that?
Voting model: Vitalik has criticized pure token voting (and even Trueo’s council model, a project he also backed). How would the pivot move Kleros away from pure PNK-weighted selection? SBT-gated jurors, reputation weighting, hybrid?
Success metrics: what should we be tracking 12 and 24 months in to know if the pivot is working or to call it off?

Thanks!

yadf1994 · June 10, 2026, 7:24am

hey Jean, thanks for your reply it serves the Kleros community!

UMA purposely doesn’t take a fee at this stage and pays voters via UMA token emission. It’s a classic attract-extract pattern where they initially privilege growth and will collect fees only when big enough. Polymarkets itself had the same pattern, not taking fees for around 5 years and only started to take those a few months ago.

UMA market cap may look small in comparison to Polymarket, but it is 6 times higher than Kleros, despite an inferior tech (only allowing full token holder votes, voted clearly wrong in the past and hand to be overridden by Polymarket). Polymarket is also talking about changing the oracle so this is also priced in. I’d expect UMA price to at least divide by 2 if Polymarket were to remove them.

Moreover, your “meme” only works if Kleros is simply a dispute resolution provider. If it participates in creating prediction markets (funding to PMs in exchange of shares, contribution to Seer leading to retroactive tokens), it will get shares of those and be able to tap into the right part of the pic.

I think the image is pretty misleading, as integrating reality doesn’t require using their default interface. When properly integrated, which is easy to do, reality is completely abstracted.
In addition, if we were to shift to prediction markets, we could assign a dev to make our own version of reality (either contract + interface or just interface). @green, i remember you said you need more Kleros that Kleros needs you, today may be the inverse!

Furthermore, you can see that UMA voting interface is less advanced than Kleros V1. No evidence, no vote justification, only one round.

UMA did provide liquidity incentives to Polymarket at a time where it dearly needed money (Kleros did the same with Omen). Now they have a significant war chest and it is unlikely that a payment to them would be required for an integration.
The hard part of the integration would be convincing them that Kleros (or an oracle aggregator using Kleros as one of them) would be superior to alternatives (no meal is free right).

This is not specific to prediction markets. Actually prediction market disputes are less likely to be handled by AI due to the value at stake being so high that minimizing arbitration fees would not be necessary. On the contrary, disputes in other sectors such as Kleros enterprise / Curation are more likely to be handled by AIs. And i expect Conditional Market to be more sophisticated than PM.
Plus, AIs can be prompt engineered (find a specific set of evidence to give them to get the output you like). Thus, they can’t be the ultimate oracle and they do not have access to the outside world. Even if they could process the automatic application of rules, humans would still be needed to determine facts (until there are robots reporting things, but that isn’t in a near future from my understanding of the robot industry). We could also imagine the initial Kleros court handling the prediction market disputes to be AI and then humans coming up in appeals.

Well that is a common criticism of Kleros (having a marketcap too low). However in practice, even when we had ~100M$ of TVL with Unslashed, “the best” economic exploiter (Avraham) was unable to break Kleros…
You can’t just pay 51% of the current PNK price to control Kleros. As you buy PNK, the price of PNK increases, and there isn’t even 51% of the staked PNK available on the market.
You can see that currently UMA handles a TVL far greater than its marketcap (and polymarket has an override option anyways).

Vitalik does indeed criticise naive voting models, but Vitalik is actually one of the biggest proponents of SchellingCoin systems like Kleros (this very early article, the “just use Kleros” tweet). We have the capacity to conduct the appropriate research and make the necessary improvements on these research issues (we basically explored everything, just that we didn’t have a market demand to deploy things…).
If i may, those kinds of discussions are orthogonal to a focus on prediction market and about Kleros mechanisms. But we should ultimately have the forking court with v2 not being simple voting.

Very important question. I would say:

TVL in prediction markets using Kleros;
Arbitration fees in prediction markets using Kleros;
Number of prediction market applications using Kleros;
Amount of funding in prediction markets using Kleros.

I will conclude by saying that the current situation is untenable for Kleros, and a drastic change is simply necessary (it’s basic logic).

n1c01a5 · June 10, 2026, 11:26am

I agree that the failures were organizational but above all human, and a matter of vision.

The issue isn’t only which product we focus on; it’s who is doing the business development and what environment we’ve built around it:

It’s primarily a people problem. We haven’t had the right people to close deals. Strong tech with the wrong go-to-market team produces exactly the outcome described: world-class technology, low adoption. A product pivot won’t fix a business-development gap.
The ecosystem isn’t built to reach businesses. We don’t have a healthy environment for landing real commercial usage. The pivot to prediction markets doesn’t change that unless we also fix how we approach and convert customers.
The “tech-first” identity is part of the problem. Kleros projects an image that is tech-first rather than community-, business-, or revenue-driven. The tech obviously has to be good but smart-contract security and technical soundness are the baseline we’re expected to meet, not a headline achievement. Good tech is necessary; it shouldn’t be a main subject.
We’re over-indexed on research. We tour major universities, when the higher-leverage move is showing up where customers actually are: corporate events, crypto startups, the places that generate deals.
The drama is corrosive. Kleros has generated a recurring stream of conflicts (PoH, Unslashed Finance, etc.). That environment isn’t conducive to attracting partners, talent, or trust
A divided community (normal in a transition phase) with no clear shared vision makes it hard to build trust with customers

m_u · June 10, 2026, 12:55pm

Agreed.

Legally it exposes Cooperative to too much risk. Both Prediction Markets and PoH have serious regulatory exposure, which are currently flying under the radar, that being said infrigements on gambling legislation / GDPR can have serious ramifications for Coop, and to an extend, also the reputation of the court itself.

PM and other grey areas really should be spun off in whatever form imaginable i.e. in DAO or a new LLC.

n1c01a5 · June 10, 2026, 1:37pm

Question to Fable: How can Kleros become a popular protocol with many integrations?

Claude (Fable 5) thinks that:

Kleros should become an infrastructure primitive, not a product. The protocols that achieve massive adoption (Chainlink, Safe) are the ones developers can integrate in a few lines of code. Kleros should offer a dead-simple SDK and pre-packaged “dispute resolution as a service” modules for marketplaces, DeFi protocols, and freelance platforms. Today, integration requires too much understanding of the arbitration system itself.

It should target use cases where arbitration is a genuine pain point: P2P and OTC escrow, parametric insurance payout disputes, optimistic oracles (Kleros as a challenge layer alongside or against UMA), content moderation, registry curation, and DAO-to-DAO disputes. One or two flagship integrations with real volume are worth more than twenty announced partnerships with no usage.

It should reduce economic and UX friction: predictable arbitration costs, guaranteed short resolution times for small disputes, and abstracting PNK away from the end user — an integrator should never have to explain the token.

Finally, credibility matters: a decentralized justice protocol whose own governance is perceived as opaque or centralized has a narrative coherence problem. Serious integrators and institutions pay attention to that. Financial transparency and genuinely community-driven governance are also business development arguments.

jnptzl · June 10, 2026, 1:46pm

I don’t understand, so you are asking Kleros to pivot to something where there’s no clear path to revenue, no estimation? And saying that we might need to spend years without value capture, while also saying the treasury is finite?

I would like to see some numbers of what the actual opportunity for value capture for Kleros is.

Is that the assessment you got from prediction market developers, or your opinion as a non-technical person? If development is necessary, how many devs and for how long to get Kleros to a best in class developer experience and product?

I think you did not get my point. As you probably know, disputes resolved through reality.eth do not go to Kleros. Same with UMA, there’s a previous instance with a bond game.

What I’m saying is that even if Kleros is the final arbitrator, it’s likely that the number of actual disputes will decrease compared to the current situation. Let’s say we have 1% of markets disputed on Kleros now, in one year it will probably be 0,1%.

Based on your response, it doesn’t look like you factored that in. Also you haven’t mentioned the alternative upcoming competition that is based on this type of tech. Some actually employ adversarial AIs that compensate from prompt injection, hallucinations and so on, so it doesn’t look like you are following that.

Please formulate a proper informed response to this.

It might be the opposite. Since they are so big, and you are arguing there is such a great opportunity for value capture there (we need to investigate if that’s actually true), landing an integration with them actually might be worth a lot more than before for potential integrators that have bigger treasuries than Kleros.

In other words, Kleros competitors might be willing to put even more money than UMA did when it was a risky bet.

So being realistic, it’s unlikely that something like that could come for free, and it would be valuable to know how much was paid.

So you don’t think that’s a barrier for new prediction market integrations?

I think this is very relevant, because the most common criticism of UMA is that it is based on pure token voting. Any big dispute on prediction markets leads to drama, and that is the first thing to be criticized. That creates a ceiling on what the success of pursuing prediction markets might be. UMA is facing a lot of criticism because of that now, having articles on every major newspaper written about it, and probably bringing that to the attention of regulators and opportunistic lawyers.

For me, you failed to make your case on why Kleros pivoting to prediction markets would change anything, since it’s not clear how value would be captured by the cooperative or token holders.

The “basic logic” is not actually considering the premises critically and jumping to conclusions because the market is big.

I would like to know @yadf1994 thoughts about this.

yadf1994 · June 10, 2026, 3:50pm

Thanks for your reply n1c01a5

“A product pivot won’t fix a business-development gap” i agree, but I think Kleros’ current target market is simply wrong. When I was at Kleros, I felt surrounded by many very talented people, so I wouldn’t be surprised if the team performs better in a relevant market with the right directions and mindset.

I totally agree.

Well yes and no, i think the mindset was focus on the product and usage will follow, with the assumption on web3 is the future, Kleros is the future of justice, etc …

I understand, here also yes and no : there are very valuable things following great connection with prestigious institutions, but if the market doesn’t follow it’s like sword stroke in the water. But i agree that customers should be the real “metric/judge”, and not the number of partnership/tour with universities.

I agree with your last two points. Mistakes were made, it’s difficult to control everything, but a better approach and greater pragmatism regarding market winning could have overcome these conflicts and united the community imo.

yadf1994 · June 10, 2026, 3:56pm

Thanks for your reply m_u

The cooperative is a quite sophisticated french legal vehicule, to have dived into it i don’t think it’s an issue for Kleros to work with PM/CM, and also develop PoH. Happy to discuss it more precisely.

And again, Kleros would constitute a layer of “regulation” through action (like insurance : I’m not insuring you if you’re too sophisticated, it’s a way to regulate AI agents).

m_u · June 10, 2026, 4:27pm

For a permissionless protocol, there is no issue. For a French / European entity the reality is harsher.

Not insisting the ANJ or other regulators make sense here, from a personal perspective people should be free etc. However, the red tape for Coop seems obvious.

greenlucid · June 10, 2026, 4:42pm

Okay, there you go. Been thinking about this for a while and this motivated me to publish.

yadf1994 · June 10, 2026, 4:52pm

Did i ask anything in the quote ? I shared the strategy of UMA, period. About the numbers, please have a look on volumes from PM, marketcap of UMA (and its track over time) and compare it to Kleros.

Oh whouah that prospective ? I think Kleros survived thanks to exchange of shares. Please dive into to understand what can bring CM : https://seer-pm.medium.com/conditional-markets-for-futarchy-decision-markets-info-finance-b1b317df7705
Is it less prospective that the current direction of Kleros (that you defend) ? I hope no …

I’m not a developer, but I code, and I have done front-end development for few experiments. Moreover, my experience with Claude’s code has led me to this conclusion. Also, i think you don’t understand what mean abstraction.
I’m not here to discuss (now) number of developers, and i said “could”. It’s not the point, and to me it’s irrelevant to point this aspect which is marginal (i’m not talking about the development of a frontier model…). This doesn’t diminish the relevance of the strategic issues in any way.
If I may, you should be the one justifying the use of resources.

I disagree, please read about Induced demand mechanisms, that can apply to PM and CM. Moreover, I see value into PM and CM because it’s a way to inform, and information has always be one of the most valuable thing. When you have JP Morgan that clearly say we’re going into PM, then it catches even more my attention. Also, UMA has a better situation than us by a lot, and so if we win this market then we can objectively target at least their current marketcap.

It’s a non argument, when you go to a market it’s part of the pack … And competition can lead to a growing market, etc.

I’m trying to follow it as everyone i guess, please share with me the litterature about it.

More generally, I believe that PM and CM are currently our best strategy for Kleros’ survival right now. This does not mean, however, that Kleros should focus solely on this path for the rest of its existence.

I disagree, their current interest is to have the best technology, they don’t need money, they have plenty of it.

I was simply pointing out the ineffectiveness of your argument for a pivot. It’s about risk, but right now it’s not an issue at all for Kleros as i said.

You think PM and CM could be handle by very long judicial process ? You’re way off base, man, PM and CM need fast dispute resolution if they want to preserve the dynamics of their market, they cannot allow long trial … And there will always be criticism, because there will always be losers.

And me I’d like to know what Kleros has accomplished in terms of market conquest, and also what are your suggestions for Kleros to win.

jnptzl · June 10, 2026, 5:51pm

That’s why I said I don’t understand your point. Instead of making a business case for Kleros adopting your strategy, you mentioned an example of a project that managed a lot of TVL and volume, and still lost 2 billion dollars of market cap in the last couple of years.

So I’m assuming you have no idea about what actual value capture looks like. Please share your estimation, % of volume of PMs, etc. that go to the oracle.

Let me remind you that you are the one proposing a pivot, which would imply less resources for other initiatives.

sir I’m the one making the communications for Kleros Foresight, you are the one that should retweet our materials and help us spread the word. Start here:

Well, I think experience with Claude Code is not enough, you need to ask actual developers if you are making a serious proposition. For example:

I don’t think you got the point. My point is that AI resolution will make more markets (in %) be resolved at the optimistic oracle part, meaning less disputes for Kleros, meaning less value captured by the PNK holders vs the current status quo.

Let me remind you you are the one proposing a pivot.

No, I’m saying that pure token voting might put a ceiling on the potential of the market you are arguing we should go after. I was asking if you had any ideas on what else we should pursue, and it looks like you haven’t put much thought into it and seems to be OK with the current status quo.

yadf1994 · June 11, 2026, 8:21am

They topped at 2 billions marketcap so they did not loose 2 billions. Before the UMA scandals (i mark them in 2025, cf after) from 2022 to 2024 their market cap was between 100 and 200 millions $ (after a bear market that everyone experienced, including Kleros), while Kleros’ market cap was between 10 and 15 millions $.
Then came 2025, with stories like this: Polymarket News: Power User Slams UMA as $200M Zelenskyy Suit Dispute Unfolds and others, where the decline began from a market capitalization of 100 millions to 35 millions $ (which is still way better than us).
You understand that these are massive market capitalizations (literally Kleros’s previous ATH before their fall) and this is a market that concern us directly (it’s literally what we do). Whether you like it or not, PM & CM disputes will have to be resolved quickly, which is a necessary condition for these markets (information is continuous over time and evolves at every moment; it doesn’t have time for lawsuits…). Moreover, UMA is for only one PM. They are more now, with concrete needs and challenges, opportunities for Kleros.

Frankly, that sounds a bit disingenuous, I wouldn’t go so far as to talk about a position based on / a privileged position at this stage.

Seriously, when we consider a market and the needs of that market with real, concrete, and effective indicators (Volumes are real, interest is real, and disputes are real), isn’t a fair approach to say let’s get into this market because it’s what we do while Kleros is running out of time. Over 2k disputes have been handle by UMA last year, i expect PM & CM to grow, and probably i see in a near future (i.e before 2027) 5 big competitors for PM/CM (you already have Kalshi & PolyMarket, Binance & Coinbase are launching PM, i expect one institution such as JP Morgan to do so), so i would say at a very least 10k disputes per year (again 2k was in 2025 and volumes are skyrocketing, cf tokenterminal - markets prediction markets section) : it did 3 billions volume in 2025, and from January 2026 to today, PM have already done 3 billions volume, so i expect at a very least 2x more volume but very ikely at least 3x cause it’s not even yet half of the year + the new PMs, thus 3x5x2k = 30k disputes starting 2027, with 3 billions$ volume lead to 2k dispute per competitor) but it should increase exponentially as CM is about forecasting and not gambling, and actors that will demand CM are not expert on how to assess their demand so likely to lead to more dispute than current state of the art with PM.

My proposition is about a strategic pivot, not developers HR, stop deviating from the real topic please.

You’re asking for details, etc. Please do it for that, tell me precisely with metrics, literature and other relevants materials why we should focus on “less dispute for Kleros” ? Google searches have increased since generative AI (including the deployment of Gemini). From what i’ve read about multi ai agent system, it’s still not easy at all for ai system to deal with process and chain of thoughts.

Kleros is in a pitiful situation, one wonders what the team is planning …

We explored many ideas, from reputation management to SBT, sophisticated voting methods, randomization, sanctions… In my opinion, Kleros has more than enough technical capabilities to go far beyond that.
The situation clearly shows us that Kleros’s problem is not technical but economic, and I propose an economic pivot (because the problem is economic), i’m trying to be really educational in case that wasn’t clear.
I’m very happy to hear any other ideas of markets that Kleros can for real reach and conquest.

greenlucid · June 11, 2026, 9:11am

You posted some screenshots in which I am advocating for the need of an Optimistic Oracle, but I don’t see how that is disproving any of @yadf1994 points. Prediction Markets is an important goal to pursuit, and the moves of Polymarket and Kalshi towards legality and institutionalization, are quickly evolving into banning markets from existing. This creates a sharper need for permissionless Prediction Markets to exist, and Kleros will play a major role on this if the timing is correct.

If anything, building the Optimistic Oracle like I propose would be the opening gate for more prediction markets to easily rely on Kleros, across any chain, which supports Yann’s path to follow. I think the two on-chain killer apps of Kleros are Prediction Markets (futarchy, hedging, trading, and task-completion) and Curation.

clesaege · June 11, 2026, 9:35am

Hey, lots of talks and I’ll try to answer most of it.
My TL;DR would be that:

Prediction market is definitely the fastest growing sector on decentralized dispute resolution / subjective oracle.
Being cautious about whether we could get a big share of this market is a reasonable position. But saying that even if we were to get the majority of this market it wouldn’t be a big win doesn’t seem to be a reasonable position. I am not saying this to be your position @jnptzl (as I see you doing things for prediction markets to use Kleros) but I think it may be interpreted like it.
It’s a bear market and everyone is more likely to be grumpy, but we should make sure to keep the environment positive and collaborative. It’s particularly important to be attractive (both for team members and community members) when the price action may not be.

Competition

Here only comparing with UMA is relevant. Integrating Kleros via reality is super easy. And you don’t even need to use the reality front in your dapp. But you can do so initially (so you can launch sooner) and then you work on integrating in your dapp UI.

If we compare Kleros to UMA, I’d say Kleros looks better from a UI perspective.
This is how disputes look like in UMA (no evidence, just link to discord):

But Kleros may look more complicated (UMA is basically a simplified version of Kleros where everyone votes on everything). So we may need to wonder if making even more complicated versions of Kleros would help competitively. See this video thinking as positioning in our case being non physical.

It may still be a good idea to make an internal version of reality (Guangmian was proposing that before he left).

Cost

Polymarket financial situation now is very different than the situation when they went with UMA. For having talked with them about Kleros VS UMA, they didn’t hint at all that they would want money for it, but are interested in getting the best system (according to their definition of best which may be different than ours).

It’s more of a question of whether it’s doable than how much it cost (and for that we need good com/BD).

AI

Well AI may eat the first rounds of all kinds of disputes. This is not AI specific. But we may also end up seing more disputes as we’ll see disputes being AI agents. Systems cannot be AI only but need to at least be able to appeal to humans in the end as AI are vulnerable to prompt engineering (I can remember this AI game where the goal was to get a carpet from a carpet seller at the lowest price and I managed to get it at 0 by pretending to be Gary Gensler and arguing the carpet to be a security).

AIs are more likely to eat low value disputes first (consumers, TCRs) than prediction markets where the value at stake means that they would have low price sensitivity.

Economic security

Old chicken and the egg (I remember the CTO of Aragon saying that Kleros marketcap was so low that it was insecure). In practice SchellingCoin systems have been shown to be able to handle way more value than their own marketcap. Not relying on “it is not economical to attack” but in “it is not practical to attack”. In practice a huge share of people are honest, most of the PNKs are not for sale, so even with a tons of money, an attack would pump the price due to slippage and would be unlikely to succeed.
And then there is always the possibility of override (which is possible in Polymarket, not in Seer).

Voting model

After all those years I can say that Kleros does work. It may not be perfect (and I think we should work more on having a culture avoiding nitpicking and trying to get best outcome for disputes, not abuse of policies in a way not planned by its writers), but it works.
At this point we can get more improvements by culture and policies than we could get by voting system improvements.

I think we should start tracking metrics for all verticals. What can be measured can be improved. Track:

TVS (total value secured)
Arbitration fees
Investment multiplier (for 1$ spent by the coop, how much is spent in total for dapps using Kleros)

Way to get deals

I agree with Nico to a certain extent. In the early days, the crypto ecosystem was too small such that getting more BD wouldn’t have helped much. We could basically prospect everyone worth prospecting.
This has evolved and we now have more ressources affected to BD.

Drama

Kleros by its nature is solving disputes. So this means we are in a drama environment. And the level of drama may actually be linked to success. Here PoH having a large number of users and being a famous project. Unslashed having a TVL so high it attracts the professional exploiter (and pedocriminal if you’ve followed his legal cases) Avraham.
Interestingly prediction markets has been the only successful usecase (at the time of Omen) which didn’t lead to big drama (there were a few Trump supporters but it was a minority and most people supported the Kleros ruling that Trump has lost the 2020 US election).

UMA also gets this kind of drama, sometimes it deserves it, sometimes not (as it just applies bad rules / clarifications). So we are not gonna stop the drama, but can make sure that it is contained and not amplified in our own channel with better moderation rules.

Legal

We cannot put all prediction markets in the same basket. Sure the coop should not directly participate in something looking like sport gambling (ex: Prode). But markets which are highly targeted at information and have low variance (Deepfunding / Foresight) are very low risk (significantly lower than doing a token sale or have people trade PNK which would be higher variance).

Yeah, I think we should have a model where the coop supports projects and give them ressources (either in kind or in term of funding) in the early days and then spin them off while keeping some weaker link (in term of funding, in term of resources which are better mutualized such as BD and com). Basically the Gnosis model.
The coop would keep some stake (share / tokens) of those and either use that to extend its runaway (which I remind has been so high thanks to both the Uniswap airdrop and the Gnosis token swap) and/or airdrop it to PNK holders.

Value capture

They privilege growth and it is classic to do that (Polymarket and Uniswap only putting fees recently, even BTC used to require transaction fees in the early days).
Their marketcap is indeed quite low compared to their power in Polymarket, but this is due to people disliking UMA resolutions and the risk that Polymarket would remove them.

I think Polymarket removing UMA would lead to a 50% price drop of their token. And polymarket using Kleros a 50-100% price increase of PNK.
We can see what it did historically with Omen.

Happy to trade PNK and UMA conditionally to those if I can find a counterparty.

KlerosJuror · June 11, 2026, 3:29pm

Great discussion.

In terms of AI-based competitors, I don’t think we have much to worry about, but we do need to release a court that enables extremely fast dispute resolution. I’m thinking in terms of hours, up to one day, for the full process: evidence submission, voting, and a final outcome.

I’m releasing a tool that should as a PoC, allow agents to make purchases and submit evidence directly into disputes. A court that can operate at this high speed would put us in a strong position to capture a significant share of the AI dispute resolution market, in a way that competitors should get consumed into our product.

Do note, the concern about AI dispute resolution hurting Kleros is something that keeps me up at night. My main worry is whether we can move quickly enough to launch an AI-native court. But I do think the core protocol can engulf the disruption if we move fast.

That said, if necessary, I think I could build additional tools that allow AI agents to submit evidence, create disputes, and fully participate in the process, which would help us remain competitive. (After my learning with the other tool I’m working on)

Also, @clesaege, I appreciate that you recognize the inherent connection between philosophy, controversy, and law. Good old Schopenhauer.

One conceptual point regarding V2: I’ve never been able to fully understand why token voting combined with game theory is considered insufficient. I think the soulbound-token approach will be very important for more conservative businesses and institutions, but for experimental markets, I’ve never been convinced that token voting plus game-theoretic incentives isn’t already enough to address many of the challenges we face today.

Going to re read this thread. Thanks!

EDIT: @yadf1994 comments about AI dispute resolution maps to my own understanding of the situation as well. Human jurors should always be needed for an objectively correct system. But I still do worry about in-house experiments such as those by Stripe which use AI for dispute resolution. We do need to worry about those competitors since AI is such a hot topic and any decrease in dispute efficiency that can be sold to others is something for us to worry about.

EDIT2: Also seeing that Vitalik says “Use Kleros,” which I take to mean the current game theoretic approach is powerful enough for many use cases today

EDIT3: Agree with @jnptzl that the premise has not been defended in regard to prediction markets being the major thing Kleros should focus on. But I do think integrations are important and therefore prediction markets SHOULD be one area where we can easily get an integration

EDIT4: Okay. I think I understand now. OP @yadf1994 is saying we should target PMs as our primary near-term integration. That may be true. Do you think there is anyway for us to get Polymarket to perhaps send a percentage of their disputes to us? That integration is probably most fruitful due to the entrenched nature of the platform.

EDIT5: @clesaege I think there is an issue with policy writing. I’ve learnt it’s very important to state the intent of a policy before getting to the rules. It should help better voting outcomes. I can give some examples if you want

Making an AI agent wrapper tool to allow agents to do the voting, evidence submission etc is going to be easier than the PoC I’m currently working on (if you want to try and subsume AI dispute resolution disruption process.

jnptzl · June 11, 2026, 5:21pm

Losing 2 billion in market cap must be interpreted properly, because it is a strong signal. I think the most reasonable assumption is that the 2 billion was based on speculation. So what comes after pure speculation? IMHO actual VALUE CAPTURE.

Crypto projects used to be valued mostly on speculation. Now they are valued more like companies in any other industry. That’s why Ethereum, while securing hundreds of billions in assets and having the biggest integrations in the industry, has such a terrible price action for example. Value secured and volume alone are NOT SUFFICIENT for the markets anymore.

I think you are assuming projects will just be valued again mostly on speculation when the bear market ends. I don’t think that’s the case, but even if that happens, I don’t think we should value serious BUSINESS DECISIONS assuming speculation will come back. I think we need an actual business case.

UMA makes the same point even more clearly. Their usage is at all-time highs right now: 41,431 assertions in June alone per their own Dune dashboard, TVS near its ATH, a new multi-year deal with Predict.fun in March. And the token sits at minus 99%, at a 35M cap. The market is looking at peak adoption of the biggest client in the category and still repricing the token down. This should raise RED FLAGS to anyone.

For me, the drama alone doesn’t explain that, nor the risk of losing polymarket (since they actually landed a multi year deal with a high-growth competitor (predict.fun).

So what does VALUE CAPTURE actually look like for Kleros as a subjective oracle? Fee revenue is what comes to mind. How would it look? Give numbers, estimates: which platforms, what volume, what dispute rate, what fee per dispute, how much gets paid to PNK holders/jurors?

Plus the cost of landing the integrations and the breakeven date against our runway.

I’m asking for a business case with numbers, not a narrative. Narratives shift. As you know, DAOs were supposed to be a thriving business, and here we are.

Awesome, this is exactly what I wanted to discuss. A dispute count is not yet a BUSINESS CASE though. Two factors are missing: what share of those disputes reaches a subjective oracle at all, and what each dispute pays.

On the first factor, look at how the five competitors you mentioned actually resolve. Kalshi is in-house, and as a CFTC-regulated DCM it is legally required to be the final adjudicator. Coinbase routes to Kalshi. Binance Wallet routes to Polymarket. A JP Morgan prediction market would probably be in-house by the same regulatory logic. That leaves Polymarket International, where UMA is the incumbent and the addressable share is shrinking: Chainlink resolves the high-frequency crypto markets with no dispute path at all, Pyth resolves equities, and the Markets Team resolves all of Polymarket US.

On the second factor, the dispute rate is not a constant you can multiply. Disputes are usually considered a UX failure for the platforms and they engineer them down: UMA’s MOOV2 upgrade cut disputes by 68 percent and they published it as a win. How many million dollars of volume per subjective oracle dispute are you expecting?

Is the ratio rising or declining because since these volume are flowing through price feeds and possibly AI resolution at an earlier stage?

So, to complete your calculation: what percentage of those 30K disputes gets secured by subjective oracles rather than price feeds, in-house teams, previous instances with AI resolution? What average fee per dispute do you expect? And what share of that fee would reach PNK holders/jurors?

Let me remind you, you are the one proposing a pivot and leaving other initiatives. I would like to know the results of your research about that.

If we are not sure on how efficient AI dispute resolution might be at the optimistic oracle stage (even with final arbitrators configured), and we have no estimation of that, I would oppose such a pivot.

As mentioned above, for this to be a proper economic pivot, it needs to be a clear business case.

I hope my comments steer the discussion in the right direction.

jnptzl · June 11, 2026, 5:28pm

This was my original question for Yann:

If integrations are being blocked by something in the product, maybe that doesn’t require a pivot for Kleros as a whole, but rather removing these blockers, for example.

Let’s just say: would hiring 1 dev fix that and get on par with the competition in 6 months? If that’s the case, maybe it’s easier to just hire 1 dev, and the main problem is solved.

I would love to hear the opinion of more developers and integrators about that.