mikyo.one / IOTA / The Macrocosmos papers / Generative Adversarial Mining
Generative Adversarial Mining, explainedWhitepaper · Macrocosmos · January 2026
Updated 29 Sep 2026
A reader's guide to one paper · 3 of 4 · incentive track

Paying for quality no one can score

This page reads Generative Adversarial Mining on Decentralized Networks (Macrocosmos, January 2026) for people who know AI models exist but have not looked at how networks like Bittensor decide who gets paid. It is a side track in these papers: not about training a model, but about the rule that pays miners on Apex, Macrocosmos's subnet 1.

A company that builds an AI product judges quality with its own staff and tests. An open network has no staff, so every subnet needs a rule a machine can apply to decide which miner did better work. For some tasks that is easy: re-run the job, or wait and see if a prediction came true. For open-ended answers, such as a written explanation or an agent's plan, there is no rule to apply. This paper's answer is to turn judging into a guessing game borrowed from how GANs train.

statedthe paper says itinferredmy arithmetic or reading of the paperPart of the Macrocosmos papers on mikyo.one
01 · Why

Every way a subnet pays its miners assumes the work can be checked

The four families of incentive the paper names, and what each one cannot reach.

Four ways to pay a miner

Section 1.3 of the paper
Mechanism
How it works, the paper's example, and its limit
Spot-checking
Validators re-do a sample of deterministic work. Data Universe (subnet 13) miners scrape about 250 million records a day this way. Limit: miners scale a known method; they can't go past it.stated
Code attestation
Miners run a program they are given, confirmed by attestation. The paper's example is IOTA (subnet 9): "miners do not innovate, but compete by finding increasingly cheap GPUs with high network speeds". Same limit.stated
Prediction
Miners forecast; time settles the score. Vanta (subnet 8) runs equity portfolios. Limit: only works where the future reveals the answer.stated
Asymmetry
Problems easy to set and hard to solve, such as reversing a hash chain. Limit: needs a very particular problem, and closeness to an answer is hard to score.stated

None of the four can pay for a better written answer, a better plan or a more realistic image, where quality is a judgment rather than a check.stated

The idea it borrows

A generative adversarial network, or GAN (Goodfellow et al., 2014), trains two models against each other. A generator makes fake samples; a discriminator tries to tell fakes from real ones. Neither needs a hand-written definition of "realistic": the discriminator's judgment is the definition, and both improve by competing.stated

The paper asks whether the same trick works as a pay rule for people running models on a network, instead of as a training method inside one model.

The question the paper takes on: can an open network pay for quality it cannot score, by letting competition between miners stand in for the score?

02 · What

Miners are paid for fooling each other, and for not being fooled

One round of the game, and the numbers the paper reports.

A validator poses a challenge. A coin flip decides who answers: a random miner, or the validator's own reference system, which is given more time and tools. A panel of other miners then guesses who wrote the answer. Correct guessers split points; the answering miner earns whatever the panel got wrong. Over time miners learn to produce answers indistinguishable from the validator's expensive ones, at a fraction of the cost.

Generative Adversarial Mining on Decentralized Networks, by Felix Quinque, Kalei Brady and Steffen Cruz of Macrocosmos, dated 9 January 2026. This page reads the Wayback Machine copy captured on 9 January 2026.

Points per round
1strictly zero-sumstated
Who answers
50 / 50a miner or the validatorstated
Skill correlation
> 0.9after one answer per miner, 256 miners, panel of 100stated
Deployed on
2 subnetsSN1 Apex, SN34 media detectionstated
Stable cabals
0the only equilibrium, in the modelstated
Deployment results
noneno measurements publishedinferred

One round

After the paper's Figure 1
Challengefrom a validator COINFLIP Miner generatortight time limit Validator referencemore time, tools, context Answersource hidden Panel of N miners each guesses: "miner" or "validator"? 3 right, 1 wrong every miner plays both roles over time, and is paid on the total

The panel's ability to tell a cheap answer from an expensive one is the quality measure. A miner earns as a generator by making answers the panel mistakes for the validator's, and as a discriminator by spotting the difference.stated

The next section shows how the single point is split, why a cabal loses, and how quickly scores settle.

03 · How

One point per round, burned when the validator wins, and a proof that collusion backfires

The scoring rule, the cabal argument, the convergence result, the deployments, and where the claims stop.

How the point is split

Panel of 4, 3 correct
¼¼¼¼ each correct guesser: 1/N generator: 1 − c/N if the generator was a miner,the miner keeps it if it was the validator, burned one round = 1 point

Points the validator's reference earns are removed from the miners' pool, so miners as a group lose whenever the panel is fooled by the validator.stated The wrong guesser gets nothing.

Why a cabal loses

Suppose some miners collude and tell each other whenever one of them wrote the answer. They can now rule out part of the field, so their guesses get more accurate, and they are right more often against the validator.stated

The paper's analysis shows the catch. The cabal's better guesses "rescue" points that would otherwise be burned, and the cabal ends up behind whenever more than half of those rescued points flow to outsiders, which the paper's plots show at every setting. In every combination of cabal size and classifier accuracy it plots, non-cabal miners score higher than cabal miners.stated

So each colluder does better by leaving, and "the only state in Nash equilibrium is a completely cabal-less state". The burn is essential: if validator points were re-shared among miners instead, colluding to depress others' scores would become the winning strategy.stated

How fast scores find real skill

The paper models each generator and discriminator pairing as a match whose odds depend on the two players' skills, in the Bradley-Terry style used for ranking players from pairwise games. It then asks how closely a miner's win rate tracks its true skill as rounds accumulate.stated

Because each answer is judged by a whole panel, every round produces many matches. With 256 miners, typical subnet capacity, and a panel of 100, the correlation between score and skill passes 0.9 after each miner has generated just once.stated Ranking miners by pay tracks their skill after one turn each.inferred

Resource asymmetry sets the ceiling

A plain GAN saturates once the generator's output can't be told apart from the real thing. Here the "real thing" is the validator's reference, which is deliberately given more: extra processing time, private resources, more context.stated

Miners, held to tighter limits, must find faster and cheaper ways to match it. The paper frames the result as distillation: the network produces lightweight versions of an expensive pipeline, and the size of the resource gap becomes a dial for how hard the game is.stated

Deployment
What each side does
Subnet 1, Apex
Validators run agentic workflows with code execution, web search and long reasoning; miners must match them under tight time limits. The product is the fast agentic workflow.stated
Subnet 34
Miners build detectors for AI-generated images and video, and generators try to evade them. The product is the detector. The paper cites Bitmind for this subnet.stated

Where the claims stop

No numbers from the fieldThe paper says the framework "has been successfully deployed across multiple production subnets" but reports no measurement from either deployment: no scores, costs, latencies or quality comparisons.stated
The proofs rest on simple modelsThe cabal result assumes every miner's classifier has the same accuracy and that colluders share only who did not write an answer. The convergence result assumes skills spread evenly and split cleanly between generators and discriminators.stated Real miners may not behave that tidily.inferred
The target is the validatorMiners are rewarded for being indistinguishable from the validator's reference. That pushes them toward the validator's quality at lower cost, not past it.inferred The paper treats this as a dial set by resource asymmetry.stated
Subnet 34 is cited to Bitmind.The paper presents it as an application of the mechanism; it does not detail who built that subnet's implementation.stated

The paper's support is a game-theory analysis. The last question is why it matters next to the training line.

04 · So what

It rewards better methods, the case IOTA's reproducibility check leaves open

What the paper adds to the picture, and what it leaves open.

The training papers here pay for work that can be reproduced. IOTA checks a miner by re-running its computation, which is why miners there compete on cheap, fast hardware rather than better methods.stated Generative Adversarial Mining is the team's design for the other case: tasks where the valuable thing is a better answer, and there is no way to re-run your way to a score.inferred

If it works as argued, it is a general-purpose pay rule for open-ended AI work on a decentralized network: agent workflows, media, and the paper suggests reinforcement learning, scientific computing and multimodal reasoning next.stated The evidence so far is a game-theory argument and two deployments reported without data.

The paper ends on a next step: capturing the miners' generator and discriminator models themselves, not only their outputs, so the innovations can be sold directly. That is described as under development, for "an upcoming publication".stated The training line resumes with ResBM, three months later.

Where this paper sits

All four papers
  1. Aug 2024SN9 pretraining whitepaperMiners each train a whole model; the best one takes the reward.Training line
  2. Jul 2025IOTA technical primerOne model split across miners, each paid for their share of the work.Training line
  3. Jan 2026Generative Adversarial MiningSide track: an incentive design for Apex, subnet 1, where quality has no scoring rule.You are here
  4. Apr 2026ResBMThe handoff between machines made 128 times smaller.Training line
Glossary · 8 terms
GAN
A generative adversarial network: a generator and a discriminator trained against each other.
Generator, discriminator
Here, roles miners play: producing answers, and guessing who produced them.
Zero-sum
A game where one player's gain is exactly another's loss; here, one point per round.
Burn
Removing rewards from circulation instead of paying them to anyone.
Cabal
A group of miners who secretly share information to win more.
Nash equilibrium
A state where no player gains by changing strategy alone.
Bradley-Terry
A model for ranking players from the outcomes of pairwise matches.
Distillation
Getting a smaller or cheaper system to reproduce the output of a larger one.
Sources · 3
  1. Quinque, Brady, Cruz. Generative Adversarial Mining on Decentralized Networks, 9 January 2026, 15 pages. Linked from macrocosmos.ai as the "APEX GAN Whitepaper" at apex.macrocosmos.ai/research/apex_gan.pdf; read from the Wayback Machine copy. Every stated label on this page is to this paper.
  2. Goodfellow et al. Generative Adversarial Nets, NeurIPS 2014. The idea the mechanism borrows.
  3. Quinque et al. IOTA: A Technical Primer for Release, arXiv 2507.17766, 2025. The paper's example of code attestation. Explained here.