This episode dissects Bitensor's evolution post-DTAO, revealing how decentralized AI training on IOTA (SN9) and data monetization on DataUniverse (SN13) are forcing a new, more competitive and commercial era for the ecosystem.
DTAO's Six-Month Retrospective: A More Competitive Ecosystem
- Six months after its launch, Dynamic TAO (DTAO)—a mechanism that allows TAO holders to stake their assets to specific subnets to influence their emission rewards—has fundamentally reshaped the Bitensor landscape. Will from Macro Cosmos notes that the previous "root network" system was an "unambiguously not an effective allocator of funds," plagued by cronyism. DTAO introduces a more market-driven, competitive dynamic.
- A Three-Pillar System: Steph frames DTAO as having three core components: the technical value of what's being built, the "crypto memetics" of trading and speculation, and the functional tokenomics. He observes that the trading aspect has made the project "much more crypto than it used to be."
- Forced Efficiency: The new model forces subnet owners to operate like public companies, focusing intensely on unit economics and capital efficiency. Teams can no longer spend TAO emissions without considering the long-term health of their token and finding product-market fit.
- A Double-Edged Sword: While DTAO has improved capital allocation, Steph expresses a key concern about its fragmenting effect on the ecosystem's narrative. He states, "I worry we dilute the Bitensor message in a sea of subnet messages sometimes," highlighting the tension between individual project promotion and the collective strength of the Bitensor brand.
Strategic Implication: Investors must now analyze subnets with the rigor of public companies. The focus has shifted from pure technical promise to sustainable tokenomics, clear messaging, and a viable path to offsetting emissions with external revenue.
The Halving's Complex Impact on Subnet Tokenomics
- The upcoming TAO Halving introduces a layer of complexity for subnet token holders that is not immediately intuitive. The speakers clarify that while the Halving cuts the amount of new TAO created in half, its effect on subnet liquidity pools and holder rewards is nuanced.
- Pool Stabilization: Steph provides a clear explanation: the ratio of TAO and subnet tokens injected into the liquidity pools remains identical because both sides of the pair are halved simultaneously. This mechanism is designed to maintain price stability within the pool.
- Holder Dilution: However, the rewards paid out to subnet owners, miners, and token holders follow a separate emission schedule that does not halve in sync with the pool injections. This creates a situation where more subnet tokens are being emitted relative to the amount of TAO available in the pool to purchase them.
- Less Fuel for the Fire: As Steph puts it, this dynamic means that "over time you have less fuel from the chain for the fire." The decreasing TAO subsidy per token puts greater pressure on subnets to generate their own demand and revenue.
Strategic Implication: Investors must model the dilutive effects of the Halving on their subnet token holdings. The differing emission schedules between liquidity pools and holder rewards create a potential supply-demand imbalance that could impact token prices if not offset by project growth or external capital inflows.
IOTA (SN9): The Ambition of Decentralized, Large-Scale Model Training
- Macro Cosmos's flagship project, IOTA (Incentivized Orchestrated Training Architecture) on Subnet 9, is one of the most technically ambitious endeavors on Bitensor. It aims to coordinate a global network of consumer and enterprise-grade hardware to train state-of-the-art AI models that can compete with those from major labs.
- Core Innovation: IOTA combines two key techniques to achieve this.
- Data Parallelism: Training multiple copies of a model simultaneously on different subsets of data.
- Model Parallelism: Splitting a single large model across multiple miners. This is a critical unlock, as it dramatically lowers the hardware barrier to entry—a miner only needs to run a piece of the model, not the entire thing.
- A Collaborative Network: Unlike most subnets where miners compete, IOTA forces collaboration. Work is passed sequentially through the network of miners to train a single, unified model. Will uses an analogy to explain the high stakes: "In IOTA... it's all one football team if somebody drops a pass you're going to concede a goal."
- Roadmap to Scale: The team is currently iterating on a 1-billion-parameter model to stabilize the system. The plan is to scale to a 15B model next, with the goal of training a highly competitive 100B+ parameter model by the end of the year.
Strategic Implication: IOTA represents a high-risk, high-reward bet on solving one of the hardest problems in decentralized AI. Its success would be a monumental achievement, potentially enabling the creation of sovereign, open, and competitive large-scale models outside the control of centralized corporations.
The Bitter Lesson and the Future of AI Architectures
- The conversation touches on the "Bitter Lesson," a concept in AI research suggesting that scaling raw compute power consistently outperforms more complex, human-designed algorithms. While acknowledging the power of scale, Steph offers a more nuanced perspective on the future of AI.
- Beyond Transformers: He argues that the current dominant architecture, the Transformer, is "extraordinarily inefficient." While it has been effective, visionaries like Yann LeCun are already looking for fundamentally different architectures. Steph cautions against a simplistic interpretation of the Bitter Lesson, stating, "The bitter lesson doesn't basically say that the architectures that we have today we just scale them up with more data in perpetuity. I don't personally believe that that's the case."
- The Limits of Synthetic Data: The discussion also covers the limitations of training models on synthetic data (AI-generated data). While it can work, it often creates models that are good at mimicking the specific tone of AI but fail on "out-of-distribution" data, like messy, typo-ridden human text.
Strategic Implication: Researchers and investors should monitor for architectural innovations beyond Transformers. While IOTA is built to excel with current methods, its underlying framework for distributed compute could be adapted for future breakthroughs, positioning it as a long-term play on decentralized compute itself, not just today's LLMs.
DataUniverse (SN13): Monetizing Real-Time Social Data
- DataUniverse (Subnet 13), whose product is named Gravity, provides a clear example of a Bitensor subnet with a direct path to commercialization. The subnet incentivizes miners to scrape real-time data from high-value social platforms like X, Reddit, and YouTube.
- Solving the Knowledge Cut-Off: The primary value is providing AI models with data that is current "to the second," overcoming the "knowledge cut-off" problem that plagues models trained on static datasets.
- Commercial Traction: Will highlights significant commercial progress:
- An integration with Taostats' AI assistant, "Doug," which uses Subnet 13 to pull real-time information about the ecosystem.
- Multiple five-figure enterprise deals, including one for a client building a "signal intelligence method" to predict crypto protocol hacks by analyzing social sentiment and search trends.
Strategic Implication: DataUniverse demonstrates a tangible and immediately monetizable use case for a Bitensor subnet. By providing a valuable commodity (real-time, structured data), it offers a less speculative investment thesis compared to moonshot R&D projects, showcasing a clear path to driving external revenue back into its ecosystem.
Bitensor as a New Model for Capitalism
- The discussion broadens to a philosophical examination of Bitensor's role not just in AI, but as a new economic model for fostering innovation. Mark, the host, proposes that Bitensor is a "new way of doing capitalism," and the guests largely agree.
- An Incubator for Startups: Steph describes Bitensor as an "incubator of decentralized startups." Its unique model—requiring staked TAO and providing gradual emissions—differs starkly from traditional VC funding or the boom-and-bust cycle of ICOs.
- The Next Stage of Growth: Will adds a crucial caveat. While Bitensor has been incredibly successful at bootstrapping early-stage teams, he questions if the current capital allocation mechanism is sufficient for "Series A/B" level growth. The volatility of a dual-token system makes it difficult to scale a large team, suggesting the protocol must evolve to support companies through their entire lifecycle.
Strategic Implication: Investors should view Bitensor as an experimental economic engine. Its long-term success will depend not only on its technical innovations but also on its ability to evolve its capital mechanisms to support projects from inception to mass-market scale.
Gardening the Ecosystem: The Debate on Subnet Quality Control
- The episode concludes with a critical discussion on maintaining the health and integrity of the Bitensor network, especially with the current freeze on new subnet registrations.
- Pruning Inactive Subnets: Will advocates for "gardening"—actively cleaning out inactive or low-value subnets to focus emissions on productive teams. He notes that approximately 44 of the 128 subnets currently have a 100% emission burn, suggesting they are dormant.
- Combating Grifters: The conversation takes a sharp turn towards combating malicious actors. Will explicitly calls out "grift" projects, such as a forked hash subnet that he claims is a known-to-be-flawed system designed to extract TAO emissions without providing value.
- Protecting the Protocol: He delivers a stark warning about the danger these projects pose to the network's long-term credibility and value. "That is never something that will make TAO to 10,000," Will asserts. "It might be something that makes you a buck this year, but it is not the thing that drives value."
Strategic Implication: The community's ability to self-regulate and prune low-quality or malicious projects is critical for the long-term health and capital efficiency of the network. Investors should closely monitor governance discussions around deregistration and emission allocation, as these will directly impact the protocol's integrity and investment quality.
This conversation reveals a network in transition. Bitensor is rapidly maturing from a research collective into a competitive commercial ecosystem where technical moonshots like IOTA coexist with revenue-generating utilities like DataUniverse. Investors must now assess subnets on both technological vision and capital efficiency, while researchers should track how the network balances permissionless innovation with rigorous quality control.
