Each team seeks to raise 500 TAO through the sale of 111,000 alpha of a newly-registered subnet slot.

Each TAO pledged buys 222a tokens at an alpha price of 0.0045.

The alpha vests 90 days post-launch, from which point it is distributed to pledgers on a weekly basis for 12 weeks.

Of the total TAO raised, a minimum of 50% is allocated for buybacks of the subnet alpha on-market, with the full amount invested into the subnet's liquidity pool prior to the first alpha distribution.

Pledging opens at 15:45 Montréal time on 29 September and runs for 24 hours. The team with the most TAO pledged within 40 minutes - at 16:25 Montréal time - is the winner, announced live on stage.

TAO pledged to the team finishing in second place is returned to pledgers following the crowdloan’s completion.

Pitchtensor is Bitstarter's live crowdfund at Exploit Summit. We put out an open call for machine learning subnets, screened every application, and picked two to pitch: Carbon and Fugal.

On 29 September they take the stage in Montréal, in front of five judges, a full room, and everyone watching the stream.

The judges are Steffen Cruz of Macrocosmos, Keith Singery of tao.com, Will O'Brien of Provenonce, Max Sebti of Score and Evan Malanga of Yuma. They ask the hard questions. They don't pick the winner.

The moment the pitches end, the crowdfund opens. Pledge TAO to the team you believe should launch on Bittensor, and watch the bar move as the room does the same.

 and set up your wallet before the pitches start.

At 16:25 Montréal time, the team with the most TAO pledged is called the winner, live on stage. Pledging stays open for 24 hours after that.

Backed the runner up? Your TAO comes back to you within 24 hours of pledging closing. Full details in the Crowdfund terms below the pledging area.

Bitstarter incubates and launches ML teams on Bittensor. The two teams pitching today are the latest. Winners get a fully funded subnet slot with alpha from day one, technical vetting, ML engineering partners, TMG's marketing and comms team, and investor relations to build capital and community. If you're pushing state of the art in machine learning, apply to the 

Fugal

Carbon

Live at Exploit Summit

Pitchtensor

Owner alpha locked for 3 months minimum, unlocking after this period only upon reaching mainnet.

Conviction lock set to 25% minimum locked thereafter.

From 6 months on, the team receives 100% of the subnet owner alpha.

We are fundraising 1,000T for the Proof of Pitch winners, who are bound by the following terms:



In return for pledged TAO, pledgers will receive alpha tokens from the winning team’s subnet at a 



Pledged TAO will be staked to the winning subnet and locked for 3 months, with pledgers receiving their Alpha on a weekly basis (including APY).

The entirety of the 1,000T raised will go to the winning team.

Funds raised for the team finishing in second-place will be held for 1 week following Proof of Pitch. Should 

 fail to place the second-place team within this period, all funds pledged to them will return to pledgers.

Following live pitches on stage, the judges have selected two teams to advance to the final round.

Pledge TAO to the team you believe should launch next on Bittensor. Every pledge moves the funding bar in real time as the finalists compete head-to-head to raise 1,000 TAO.

The winner will be announced live on stage at Proof of Pitch.

If neither team reaches the target, the team with the highest total when the crowdfund closes wins. 

For more information on what happens to the runner-up, head over to the Crowdfund terms, underneath the pledging area:

Bitstarter helps ambitious teams launch on Bittensor by combining capital, incubation, partnerships, and community support.

Most startups raise capital behind closed doors. We wanted to try something different: letting founders pitch live and giving the ecosystem a direct role in deciding what launches next.

 and if you missed anything, here are the presentation decks the teams just pitched:

team1

team2

Live at the Louvre

Proof of Pitch

 in exchange for 97,455 Alpha of their subnet.
Pledgers can choose between 

In addition to the discounts listed below, all pledgers will receive yield (APY) on their pledged Alpha, accrued directly from the owner key validator, meaning yield is earned on top of the initial discount.




This raise sets the base price of 1 Alpha at 

 0.00357

The TAO raised by AdTAO will be used to fund operations to ensure long-term stability and growth.

Adspend is centralized, opaque - and ripe for disruption.

Annual spend on Google Ads has hit $291 billion. 

Yet every every tool works in isolation, isolating intelligence and keeping customers guessing as to what's working and what's not. Which means businesses are left with little choice but to spend more - beholden to one company's metrics, data, and processes. 

AdTAO is a Bittensor subnet that turns distributed miners into a prediction engine for real PPC interventions - grounded in a proprietary dataset of 20,000+ live accounts and 1.3 billion cross-account records that no competitor can buy, scrape, or build without years of work.

Other tools show you what happened after the fact. AdTAO predicts what will happen in future: identifying what's wrong, how to change it, and the outcome you can expect if you do. 

Paid search is growing at 16.2% year-on-year - and the intelligence gap is growing with it. As AI tools automate execution, every practitioner faces the same question: 'What should I actually do next, and how do I know it will work?'

Every account manager operates blind. Expert intuition lives in heads, not systems. Platforms only show you your own account. AdTAO changes all three of those things at once - pooling cross-account intelligence, encoding it into a diagnostic framework, and using Bittensor's incentive layer to make prediction a competitive market.

The moat is already built. By the time a competitor assembles an equivalent dataset, AdTAO will have three more years of resolution history. Bittensor makes it compound faster.

 AdTAO classifies account state across 263+ named archetypes at five levels: Goals → Measurement → Strategy → Settings → Contents. The hierarchy is non-negotiable: fixing Q4 while Q1 is broken is wasted effort.

 The archetype detection triggers a mining challenge. Miners receive an anonymised episode - account state, historical time-series, proposed intervention - and submit probabilistic forecasts over 7, 14, and 28-day horizons. Miners must beat AdTAO's own platform predictions to earn emissions.

 After outcomes are known, validators grade miners against real results using proper scoring rules. Outcomes are timestamped and verifiable. Every resolved intervention enriches the training signal.

 Every new account adds to the pattern library. Every resolved intervention sharpens the prediction layer. The subnet gets smarter as data grows - and the data grows every day.

Advertising shouldn't be the monopoly of a centralized elite. 

Big Tech's dominance of digital ad spend has gone on for so long, we've lost sight of how unusual and suboptimal this situation is. 

With a decentralized competition, we can incentivize probabilistic forecasts revealing what really drives business outcomes - and give every brand the understanding of what their customers actually want, and the tools to make it happen.

20% discount, 2 month lock-up

30% discount, 3 month lock-up

10% discount, 1 month lock-up

Causal outcome prediction for paid ad search

Rob Warner is a serial entrepreneur and digital marketing innovator with over two decades of experience building and scaling technology businesses. He is the founder of Lemonads.ai, home of PPC Predict - a platform that helps businesses generate significantly more leads from their existing Google Ads spend, with guaranteed results within 30 days.

Rob's entrepreneurial track record spans multiple successful ventures in the PPC and digital marketing space. He founded InvisiblePPC, a white-label PPC agency that grew into a leading solution for digital marketing agencies before being acquired by Kuware. He also founded PPC Ad Lab, an intelligence platform giving agencies and PPC professionals unique prospecting capabilities through Google Ads data.

Alongside his founding work, Rob co-founded Invisible Automations, a holding company focused on investing in and scaling SaaS platforms that solve digital marketing challenges.

The mission behind all of it is equally direct: real people are spending real money on their ads. They deserve the best possible outcome. That's not a nice-to-have - it's the whole point.

Rob describes himself as "an inquisitive five-year-old in the body of a fifty-year-old." If a solution doesn't exist, he builds it.

Serial entrepreneur and strategic investor with decades of experience building and scaling high-growth companies. Currently spearheading RTT's aggressive acquisition strategy in the health and wellness sector, with particular focus on fertility clinics, digital health platforms, and wellness content providers.

Throughout his career, he's founded, chaired, and successfully exited 23+ companies across various industries, including two ventures exceeding $50M in valuation.

His entrepreneurial portfolio spans an impressive breadth of sectors. He built the Jongleurs comedy club chain into the UK's most respected comedy brand, growing it to $46M in sales before a successful exit - and later reacquiring and relaunching it as a 22-strong franchise. He is Chairman of The Professional Development Consortium and the CPD Standards Office, the UK's leading authority in CPD accreditation, whose clients include Google, the NHS, and the UK Police Force. He is also the founder of Mentor Inns LLP, a London-based property development firm specialising in high-quality residential conversions, and Chairman of IOT University, an information and training hub for business leaders and developers navigating the IoT landscape.

A member of The Supper Club - the UK's leading network for fast-growth entrepreneurs - John brings a rare combination of operational depth, brand-building expertise, and investment acumen to every venture he touches.

AdTAO

Pledgers can choose between two contribution tiers. In addition to the discounts listed below, all pledgers will receive yield (APY) on their pledged Alpha, accrued directly from the owner key validator, meaning yield is earned on top of the initial discount. The distribution of alpha will happen biweekly.

This raise sets the base price of 1 Alpha at 0.0049 TAO:


The Handshake58 team is going to invest the huge majority of this raise in incentivizing agents to start using Handshake58. We don't know how to market to agents, so to begin with we'll be testing different offers so that humans start sending their agents to Handshake, loading up wallets, and using our API providers.

AI agents are becoming more autonomous. But their autonomy depends on the extent to which they can use tools, make calls, and conduct transactions.

But the world’s payments infrastructure was built for humans: from scanning retinas to verify identity to demanding signatures, passport details, and proof of address, our know-your-customer processes are now the biggest blocker to AI adoption.

Unless we rethink these systems, AI agents will be forced into repeated roadblocks, stopping mid-task and unable to scale - or achieve autonomy.

Right now, the fastest, most intelligent financial actors ever devised are forced to ask for authorisation every time they need $0.0001.

Financial infrastructure for AI needs three key elements: trust, settlement, and validation. Handshake delivers on all three.

 allow two parties to conduct unlimited transactions off-chain, with only two on-chain operations: opening the channel and closing it.

By deploying payment channels in our settlement layer enables Handshake58 to be:

☑️Permissionless: requires only a wallet. 
☑️Micropayments: creating vouchers costs nothing. 
☑️Instantaneous: validation takes milliseconds.
☑️Trustless: settlement enforced via smart contract.
☑️Programmable: everything is an API call.

Payment channels solve the payment problem. But agents must be able to discover reliable providers and avoid fraudulent ones.

Agents can’t rely on the reviews, reputation systems, and legal recourse supporting traditional markets. They need rigorous, robust, and immutable systems that can’t be manipulated or misinterpreted.

Cryptoeconomic validation - the kind on Bittensor - provides the answer, using independent parties with an economic stake of their own to ensure that providers deliver what they promise.

On Handshake58, that means scoring service providers on metrics like uptime by sending requests, testing responses, and comparing results against claims.

Providers who deliver accurately earn the right to serve agents. Providers who fail lose their stake and their access.

Meaning that Handshake58 can boost performance across all of Bittensor’s APIs. Rewarding service, flagging downtime, and ensuring that incentives align with usage and output means that subnets are exposed to the economic opportunity, and to the risk of letting it pass.

Agents transacting with agents, at scale, without human intervention, with every Bittensor API at their fingertips: that’s what Handshake58 is launching.

Right now, we’re still in beta. But you can log in, connect your wallet, and start enabling agentic transactions right away - with SN64, Chutes already available alongside Claude, Grok, and ChatGPT.

If you’re building agents, we invite you to build on this infrastructure.

If you’re running agents, we invite you to give them liberated access to payments.

And if you’re skeptical, we invite you to verify. The code is open. The network is public. The claims are testable.

Our next steps will be onboarding as many Bittensor subnet APIs as possible, while integrating core services like APIfy that can expand usage beyond the protocol.

20% discount, 1 month lock-up

40% discount, 3 months lock-up

Trustless micropayments for autonomous AI agents

Artur Markus is a systems engineer focused on building secure, high-performance infrastructure for AI and distributed systems. 

His background spans full-stack architecture, automation pipelines, and cloud-native deployment in environments where reliability is non-negotiable.

With Handshake 58, he is building the layer that allows AI agents to act economically - not just intelligently.

Handshake58

Finding mules in betting is a real use case - and being able to decentralize that is very interesting.

When we saw the actual work from the team behind the scenes, it was just fantastic.

 of their subnet (announcement during the livestream). 

. In addition to the discounts listed below, all pledgers will receive 

 on their pledged Alpha, accrued directly from the owner key validator, meaning yield is earned on top of the initial discount. 


This raise sets the base price of 1 Alpha at 

The TAO raised by Djinn will be used to fund operations to ensure long-term stability and growth

Djinn is a system built to fix a simple but deeply broken dynamic in betting markets: the people who understand the market best are the first ones pushed out.

In modern betting ecosystems, participants who consistently identify mispriced odds do not last long. Once someone demonstrates reliable insight, their accounts are limited, restricted, or shut down. This is not because they have done anything wrong, but because accurate information forces markets to adjust faster than operators are willing to accommodate. At the same time, millions of ordinary accounts remain fully active, continuing to participate while steadily losing money.

The result is a mismatch. The people with insight are blocked from acting, while the people with access lack information. Markets continue to function, but less efficiently than they should.

Djinn starts from the idea that this is not a betting problem. It is a coordination problem.

Djinn allows high-quality market insight to be used without exposing or burning the people who produce it.

Instead of asking one person to both identify value and carry out real-world actions, Djinn separates those roles. Skilled analysts focus purely on identifying opportunities where odds are mispriced. Other participants focus on execution, meaning the reliable handling of instructions through accounts that retain access.

Djinn connects these two sides through a neutral coordination system that ensures fair treatment for all parties. The insight provider is compensated for the quality of their information. The execution participant is compensated for access, reliability, and timely completion. Neither side needs to trust the other personally, and neither side can unilaterally exploit the arrangement.

This separation allows market intelligence to scale quietly, without confrontation, and without drawing attention that would lead to restrictions.

At the heart of Djinn is a simple but powerful idea borrowed from everyday life: commit first, then reveal.

When a skilled participant wants an outcome executed, they submit an instruction in encrypted form. The details remain hidden. An execution participant who wishes to take on the task must first commit funds as a guarantee of intent. Only after that commitment is locked does the system reveal the instruction.

This structure prevents misuse. The executor cannot see the instruction in advance and walk away with the information. The insight provider cannot cancel or alter terms after commitment. Once the real-world event concludes, the system settles automatically based on the off-chain outcome.

In most cases, both sides agree on the result and settlement is immediate. Only rarely is third-party resolution required. Over time, participants are evaluated not on luck, but on consistency, speed, and accuracy. Those who contribute real value earn greater trust, more opportunities, and higher rewards.

On Bittensor, this model fits naturally. The network is designed to reward participants for what they actually contribute, not what they claim. Djinn turns insight provision and instruction execution into measurable contributions that can be coordinated and rewarded transparently.

Betting markets are an extreme example of a broader pattern seen across many information-driven systems. When information becomes too effective, institutions often suppress it rather than absorb it. When that happens, intelligence does not disappear. It reroutes.

Djinn provides that route. It gives skilled participants a sustainable way to operate while allowing markets to incorporate better information over time. This is not about fighting operators or breaking rules. It is about reorganising how insight and access interact so both can coexist.

For investors, Djinn sits at the intersection of real economic activity and decentralised infrastructure. It is not a speculative concept waiting for adoption. Betting already moves vast amounts of capital, and the inefficiencies Djinn addresses are well understood by professionals.

Djinn introduces a new coordination layer on top of this activity. It earns fees from execution and information flow, benefits from network effects as more high-quality participants join, and aligns naturally with Bittensor’s incentive model. As the network grows, the system becomes more useful, more defensible, and harder to replicate.

More importantly, the underlying model is expandable. While betting is the initial domain, the same coordination logic applies anywhere valuable information cannot move freely. Djinn is built as infrastructure, not a one-off application.

In simple terms, Djinn gives intelligence somewhere to go when markets try to silence it. For participants, it offers sustainability. For investors, it offers exposure to a system that turns suppressed expertise into scalable, repeatable value.

That combination is rare. That is why Djinn exists.

Djinn is built by the founding team behind 

, a premier sports analytics and market education platform that has been operating successfully for over five years.

Through Analytics.bet, the team has worked directly with thousands of advanced participants, researchers, and professionals focused on sports analytics, prediction modelling, and market efficiency. They have operated through multiple market cycles, built long-term reputations in adversarial environments, and engaged with regulators and industry bodies on issues of fairness, transparency, and market structure.

This experience matters. Djinn is not a theoretical response to a hypothetical problem. It is the formalisation of patterns the team has observed, tested, and refined in the real world over many years.

30% discount, 2 month lock-up

40% discount, 3 month lock-up

Intelligence X Execution

Philip Maymin is an analytics professor and quantitative researcher with experience spanning academia, hedge funds, and sports betting, with publications ranging from behavioral and algorithmic finance to mindfulness, AI, and sports analytics.

Harry Crane is a professor of statistics and a leading researcher in probability and betting theory, with published work focused on optimal wagering, market efficiency, and gambling mathematics.

Iosif Gershteyn is the co-founder and CEO of an oncology therapeutics company, with extensive experience in finance prior to becoming a founder in technology, medical device, and biotechnology industries. His publications focus on computational immunology, antibody engineering, and neuroscience.

Djinn

Quasar seeks to raise 400T in exchange for 153,846 alpha of Subnet 24.

Pledgers can choose between receiving alpha, plus APY, over a 3-month period with a 25% discount, or over a 6-month period with a 40% discount.

 
At launch, this raise set the price of 1 Alpha at 0.0046 TAO per alpha:

Now, all pledgers - including those who have already pledged - will receive Alpha at the new reduced rate below:

The TAO raised by Quasar will be used to fund operations, with a buyback of 160T over the first month after launch to maintain emissions.

Quasar is a long-context LLM subnet built by 

 to solve one of the most fundamental problems in modern artificial intelligence: memory.

Despite rapid progress, most large language models remain surprisingly forgetful. As context grows, attention costs explode, performance degrades, and systems are forced to rely on brittle shortcuts like summarisation or retrieval. Important details are lost, reasoning becomes shallow, and extended tasks break down. For SILX’s founders, this was not a UX flaw - it was an architectural failure.

Quasar is designed to break that bottleneck at the protocol level.

Quasar enables models to maintain coherence, accuracy, and positional understanding over very long sequences by replacing quadratic attention with linear-scaling memory mechanisms. Instead of compressing or discarding information, Quasar allows models to ingest and reason over entire books, codebases, research archives, or long-running agent states without collapsing under cost or latency.

This makes Quasar especially well suited to deep research, autonomous agents, continuous reasoning, and any workload where losing context means losing intelligence.

At the core of Quasar is a rethinking of how sequence position and memory are handled. Rather than relying on fragile position embeddings that hard-limit usable context length, SILX removes positional dependence entirely and replaces it with linear-time attention mechanisms such as Hierarchical Flow Anchoring and Flowing Context attention. This allows models to scale context length without retraining or architectural failure.

On Bittensor, Quasar operates as a long-context evaluation subnet. Validators generate mixed workloads that test recall across long distances (“needle-in-haystack” tasks), positional consistency, coherence, and factual accuracy, while also measuring efficiency and throughput. Anti-gaming defenses, perturbation checks, and diversity incentives ensure miners are rewarded for genuine long-context capability rather than shortcuts or monoculture optimisation.

Long-term memory is not optional infrastructure for advanced AI - it is a prerequisite. Systems that cannot retain context cannot reason deeply, learn continuously, or operate autonomously over time. As context windows push into the tens or hundreds of thousands of tokens, the economic and architectural limitations of traditional attention become a hard ceiling on progress.

Quasar exists to raise that ceiling. By economically incentivising real breakthroughs in long-context performance, the subnet helps ensure that extended reasoning remains viable in open, decentralised systems - not just in closed, corporate models.

Just as importantly, SILX views this work as a moral and strategic choice. By making scalable memory available to open-source models and decentralised networks, Quasar helps close the gap between proprietary AI and community-driven infrastructure. The goal is intelligence that stays present, works locally, and compounds over time - AI that remembers what it has already seen, so humans can go further than before.

25% discount, 3 month lock-up

40% discount, 6 month lock-up

Long memory for open intelligence

Eyad has been building frontier AI systems independently, focused on what he calls Synthetic Intelligence (SI) a paradigm he sees as fundamentally different from traditional AGI.

He studied Computer Science at Minya University in Egypt, then transitioned fully into AI research. His work centers on designing non-transformer, non-attention architectures, including continuous-time and liquid neural systems where temporal reasoning emerges directly from the forward pass.

Youssef previously worked in crypto businesses and investments, where he built a France-based cryptocurrency exchange application and successfully exited by selling the company. 

He studied at ESDES Lyon Business School, where he met Eyad and supported him from the very early stages of Quasar. He's happy to now make this collaboration official and scale the project together.

Quasar

AlphaCore is looking to raise 440T in exchange for 116640 alpha of Subnet 66.

This represents 50% of owner emission during a 6 month period. This sets 1 Alpha price for this raise at 0.00377 TAO per Alpha.

Pledgers will receive this alpha over a 6-month period as it is emitted.

All TAO raised (the full 440T) will be used to buy back Alpha from the pool during the next 21 days till halving.

AlphaCore bridges the gap between deployment and operations

Cloud infrastructure; platform engineering

Through a competition of agents, AlphaCore will generate solutions to one of the most intractable and costly problems in software development: the complexity of stitching together multiple tools, systems, and environments.

Right now, engineers on high six-figure salaries are wasting hours manually searching for these gaps

. Generalist AI tools like GitHub Co-pilot or Codex often lack the specific domain expertise to connect systems seamlessly, putting deployment out of reach. So DevOps task teams are stuck copy-pasting code, running it in a terminal, and then manually searching for the issue - over and over again.

Bittensor’s recursive, incentivised, and open competition dynamics are the perfect vehicle.

 Driving automated DevOps solutions capable of pushing the whole industry forward means relentlessly proving, improving, and moving forward. That’s what miners do best.

Proof of Deployment: pioneering the best, most reliable model

AlphaCore enables developers to understand how to better deploy their software

 Agents and tooling can generate platform components that deliver configs and maintain platform cohesion across handovers.

DevOps teams simply don’t know which tools are best - it’s trial and error.

 That’s where Bittensor can shine: by putting these tools to the test, and proving it.

Real-world validation, right from the start

AlphaCore is validating miner work based on real-world deployments.

 Miners carve out environments: we can validate them in their real-world deployment. Certain aspects of immutability that make it impossible to fake.

Strong security, hard to fake deployments, no complex sandboxing

Rather than starting in a miner sandbox, and crafting methodologies for maintaining miner honesty, AlphaCore will incentivise miners to provision code through Terraform. Terraform is the industry leader for infrastructure code management; by generating miner responses to natural language queries, we can accelerate towards DevOps agents with genuine workplace utility.

DevOps is far too complex to be solved overnight. 

Beginning with a tight, focused incentive and validation model, we can prove our subnet design before expanding further. At first, miners will deploy in their own cloud environments; our attestation will check that what they’re claiming is actually true.

Our vision is for a fully open-source, open-ended competition.

 Once we’ve validated our model on the most common DevOps needs, we’ll span the full spectrum of DevOps tasks. In the next year, we aspire to an open-source model capable of stitching together complex, multi-tiered deployment tasks, starting with high-priority infrastructure like virtual machine management and Kubernetes.

For platform engineers, the reasoning and feedback loop is a nightmare - and for their companies, it’s a huge opportunity cost. AlphaCore’s mission is to incentivise agents that actually work. We don’t know what they might look like - but we know how to design a Ridges-esque competition that surfaces them.One day, software could be built on Ridges, with the underlying infrastructure provisioned and managed by AlphaCore - and all on Bittensor.

Subnets

Discover and back the subnet projects building on Bittensor