Marketing already runs on prediction markets. They just hide behind dashboards.
An ad platform predicts which impression deserves the next dollar. A creator platform predicts which voice can move an audience. A sales team predicts which lead deserves a call.
Bittensor now has subnet teams building around all three decisions.
Tao Outsider proposes a name for that cluster. DeMkt is short for decentralized marketing.
The category is editorial. It is not an official Bittensor taxonomy, and the three projects inside it are not versions of the same product.
AdTAO SN21 asks models to predict outcomes from changes in Google Ads accounts. Bitcast SN93 coordinates campaign briefs, creator submissions and engagement verification. Leadpoet SN71 runs research and fulfillment tracks around sales leads and an AI sales agent.
Together they outline a larger possibility. Bittensor miners can compete over commercial decisions that already absorb enormous budgets outside crypto.
They also expose the hardest part of the thesis. A subnet can measure a score without proving that a marketer earned more money. DeMkt will deserve the name only when the chain’s internal competition maps to outcomes buyers trust.
Disclosure
Victor Lamenha and Tao Outsider have a professional relationship with AdTAO. That relationship is why this article separates public code, project statements, operational snapshots and editorial interpretation claim by claim. AdTAO receives no presumption of superiority here.
The market is large enough to demand better measurement
The Internet Advertising Bureau and PwC reported that United States internet advertising revenue reached nearly $300 billion in 2025.
IAB also expects United States creator advertising spend to reach $44 billion in 2026. Its research identifies measurement, standards, creator selection and operational tools as persistent problems.
Those numbers describe the broad advertising economy. They are not Bittensor revenue, a total addressable market calculation for these subnets or evidence that brands want decentralized infrastructure.
They explain why the problem attracts builders.
Marketing produces abundant data and incomplete truth. Platforms expose clicks, views, conversions and costs. Teams still argue about attribution, incrementality, fraud, audience quality and which action caused the result.
A decentralized market can add competition. It cannot remove ambiguity from the outcome.
The three DeMkt projects choose different points in that chain.
AdTAO SN21 competes over campaign outcome prediction
AdTAO’s current public repository describes an impact prediction subnet for Google Ads.
Miners ship a digest pinned container. The subnet runs that model against a fresh daily basket of real account changes. Models produce distributions over future outcomes. Validators score those predictions and convert standings into weights.
The core object is a forecast.
The phrase “AI for advertising” is too broad for this design. A model has to predict what happens after a campaign change, under the input and scoring contract defined by the subnet.
The project has recently tightened its reward audit. An August 26 commit added a public allocation audit alongside the daily vector and documented earning rules. An August 27 performance change added a daily fingerprint index around payer receipts.
Those changes improve the traceability of the internal reward process. They do not prove that an AdTAO forecast beats an agency, Google or a commercial baseline.
AdTAO has made public revenue, customer and miner payout claims. Tao Outsider covered its June recap in a separate article. Those figures remain project reported unless matched to an auditable financial source.
At the August 28 UTC TaoSwap snapshot used here, SN21 showed 15 active miners. The count confirms current subnet participation in that data source. It does not tell us whether customer campaigns improved.
The strongest future proof would connect three records. Readers should be able to follow the model prediction, the action a buyer took and the incremental business result after the action.
Bitcast SN93 turns creator campaigns into a verifiable protocol
Bitcast operates closer to media distribution.
Its public Bitcast X repository describes a protocol with several roles. Brands define briefs. Creators publish content. Miner platforms coordinate participation. Validators verify submissions and engagement before setting rewards.
The core object is a verified creator campaign contribution.
Bitcast’s protocol is more than a creator dashboard. The repository documents signed miner commitments, creator identity checks, finalized metagraph discovery and validator reconciliation. Recent commits bound submissions more tightly to creator identity, isolated campaign mutations and added retry handling for edited briefs.
Those are practical trust problems. A campaign market must know who created the post, whether it matched the brief, whether the result belongs to the claimed campaign and whether a later edit changed the terms.
Bitcast says its products are expanding beyond Web3 and has described fiat payments, new platforms, revenue and close to 1,000 creators. These are Bitcast statements. The public protocol can help verify submissions. It does not independently audit the company’s revenue, partner pipeline or creator count.
The TaoSwap snapshot showed eight active miners on SN93 and a materially larger emission share than the other two projects in this comparison. Emission can signal current market position inside Bittensor. It is not customer adoption.
Bitcast sits in the DeMkt map because it tries to make creator coordination and campaign evidence programmable. Its success depends on whether brands prefer that market to the centralized platforms and agencies they already use.
Leadpoet SN71 splits sales intelligence into research and fulfillment
Leadpoet starts later in the funnel.
Its current public repository describes two tracks.
In the public research track, miners direct compute and research toward improving an AI sales agent. The project keeps parts of the sourcing model and sealed benchmark undisclosed to reduce leakage and overfitting.
In Fulfillment, miners compete on real lead requests by submitting qualified leads.
The core objects are model improvements and candidate leads.
This creates a different verification problem. A creator post can be checked against a public URL. A campaign forecast can be scored against later account data. A lead can look correct in a database and still be commercially useless because the person has no intent, budget or timing.
Leadpoet therefore has to measure quality beyond contact accuracy. The project documents signed artifacts, benchmark protection, compute reimbursement and rewards for verified model improvements. Some of the most important sourcing and evaluation details are not public.
Leadpoet has announced partnerships, pilots and customer activity, including work associated with Dropbox and a win at an OKX hackathon. These remain project statements unless the named counterpart or a verifiable customer record supports the operational claim.
The TaoSwap snapshot showed zero active miners and zero emission for SN71 at that moment. The reading is a serious operational caution. It does not prove that the company, repository or product is inactive, and it should not be hidden behind broader enthusiasm for the category.
One category, three different buying decisions
The DeMkt label becomes useful when it keeps the differences visible.
| Project | Commercial decision | Miner contribution | Validator problem | External proof still needed |
|---|---|---|---|---|
| AdTAO SN21 | Which campaign change is likely to improve an outcome? | Prediction model container | Score probabilistic forecasts against later account data | Incremental campaign performance and customer retention |
| Bitcast SN93 | Which creator contribution satisfied a brief and earned attention? | Campaign coordination and signed submissions | Verify identity, brief compliance, engagement and reward history | Brand renewal, attributable outcomes and audited platform revenue |
| Leadpoet SN71 | Which research improvement or lead is valuable to a sales team? | Research direction, compute or qualified leads | Protect benchmarks and score lead or model quality | Buyer acceptance, conversion quality and durable fulfillment demand |
Calling all three marketing subnets is accurate at a high level. Calling them direct competitors would be misleading.
AdTAO can improve a paid media decision without finding a creator. Bitcast can coordinate creators without predicting a Google Ads account change. Leadpoet can deliver a sales lead without running either campaign.
The shared layer is incentive design around measurable commercial work.
What decentralization can improve
A DeMkt subnet can open a task to more competitors than one agency or software vendor would employ. It can make scoring and payouts visible. It can reward methods that outperform the current leader. It can reduce dependence on one vendor team.
Those are real advantages when the task and evaluator are well designed.
Competition can also make a weak metric fail faster.
If an ad model is paid for predicting an easy proxy, miners will optimize the proxy. If a creator campaign rewards raw engagement, participants can chase low quality attention. If a lead market checks only contact validity, miners can flood it with people who never buy.
The network does not rescue a bad target. It industrializes the race toward that target.
Evaluation design will decide much of DeMkt’s value. Customer feedback and fraud resistance will decide the rest. Category branding cannot substitute for either.
What would make the category credible
Three standards would move DeMkt from a useful idea to a defensible market category.
Comparable commercial evidence
Projects should disclose revenue periods, customer concentration, payment status and whether any subnet funded entity is the buyer. Annualized run rate, booked contract value and collected cash are different numbers.
Reproducible internal scoring
Miners need to understand what earns rewards without receiving the answer key. Validators need a path to reproduce weights. Buyers need to know that the winning score maps to their business problem.
Outcome evidence beyond the subnet
The final proof belongs with the user. Did the campaign improve? Did the brand renew? Did the lead convert? Did the model reduce cost or produce a decision the buyer trusted?
The Bittensor layer can coordinate competition and settle rewards. It cannot manufacture that external result.
Why the category still matters
Marketing is often treated as a soft field beside model training or scientific compute. That misses the economic structure.
Advertising and sales are repeated prediction problems attached to large budgets. They generate fast feedback, imperfect labels and constant pressure to improve decisions. Those are conditions where competitive machine intelligence can become useful.
AdTAO, Bitcast and Leadpoet do not yet prove a decentralized marketing revolution. They show three serious attempts to turn commercial work into a miner task and a validator decision.
The evidence justifies drawing the map while keeping its borders provisional.
DeMkt should remain a question before it becomes a sector badge. Can an open incentive market produce a marketing result that a buyer can verify, repeat and prefer?
The answer will come from campaigns, creators and sales teams. The chain can make them compete.
Sources
Tao Outsider: The proposed DeMkt category
-
AdTAO SN21 public repository, allocation audit commit, August 27 receipt index commit and August 27 project update
-
Bitcast X public protocol, creator identity commit and August 27 project update
-
IAB 2025 Internet Advertising Revenue Report and 2025 Creator Economy Ad Spend and Strategy Report
-
TaoSwap subnet explorer for timestamped market status
Was this article useful?
One tap feedback helps us improve each post.
