Unsupervised Capital’s Q2 Letter on Bittensor and the AI Race

Sami Kassab, managing partner at Unsupervised Capital, published the fund’s Q2 2026 letter, offering a thoughtful analysis of Bittensor’s long-term investment case.

Drawing on Chris Burniske’s idea that Bitcoin reflects what society needs most, the letter argues that Bittensor is becoming the equivalent mirror for artificial intelligence.

It points to recent US export controls that temporarily forced Anthropic to withdraw Fable and Mythos, highlighting the geopolitical risks surrounding frontier AI.

Against that backdrop, the letter presents Bittensor as a decentralized hedge while exploring the ecosystem’s recent progress and the reflexive forces driving its growth.

Why Bittensor Keeps Absorbing the Next Narrative

The letter’s central argument is that Bittensor rotates through narratives the same way Bitcoin did, absorbing whatever the moment demands.

1. Frontier access proved fragile: Anthropic pulling Fable and Mythos under US export controls was the specific event that surfaced the risk. Rumors of parallel Chinese restrictions kept circulating alongside it.

The US Export Control on Fable 5 and Mythos Just Became a Bittensor Bull Case

2. Decentralized development sidesteps that risk: No single jurisdiction can gate access to a network no one owns.

3. The full stack sits under one ecosystem: Agent frameworks at the top, GPUs at the bottom, everything between. Distributed training, data quality, autonomous agents financing themselves; whichever narrative wins next, the pieces are already there.

How the Reflexivity Works

Kassab draws a direct parallel between Bitcoin’s reflexivity and Bittensor’s, both aimed at different resources.

1. Bitcoin’s flywheel pointed at security: Higher price meant more mining revenue, more compute defending the chain, and a network harder to attack. Price appreciation directly improved the underlying tech.

2. Bittensor’s flywheel points at intelligence: Every subnet token is denominated in $TAO, so a higher $TAO price lifts productive capacity across every subnet at once.

3. The compounding is symmetric: $TAO doubling means Targon (SN4)‘s miner rewards double in dollar terms and more compute get deployed, and NOVA (SN68) rewards climb enough to attract stronger ML talent.

4. Conviction manifests the network: As more capital concludes centralized AI needs an alternative, subnets get better funded, produce better intelligence, and validate the original belief.

The Moment Is Finally Here: Conviction Goes Fully Live on Bittensor

Subnets are already generating tens of millions in  Annual Recurring Revenue (ARR), so the reflexivity is running on real fundamentals now, not just narrative.

What Shipped Last Quarter

The movement clusters into three distinct patterns: protocol-level changes, new subnets pulling in outside attention, and the blue chips compounding.

On the protocol side, two shifts stand out.

1. Const is effectively operating as CEO again: The network hit a wall on democratic process it wasn’t built for, and the return to focused leadership has upgraded shipping faster than at any point in the past year.

2. Chain buybacks now cover miner emissions on many subnets: Emissions redirect toward buying back subnet tokens rather than only feeding liquidity pools, which removes the largest source of structural selling the network was carrying.

The new subnet wave is what’s pulling outside attention back to the ecosystem.

1. Minos (SN107) co-authored a paper with OpenAI on agentic AI in scientific computing, delivering the kind of academic co-sign critics said the network would never earn.

The Most Complex Tool in OpenAI’s Genomics Report Was Built by a Bittensor Team (Minos)

2. Engy (SN53) is going directly at Chutes (SN64)’ inference lane, running open-source models like Kimi-K3 on 5090s.

Bittensor’s Engy Subnet Brings Earth’s Lowest-Cost Access to Kimi K3

3. GM (SN28) shipped a fully private inference gateway for OpenAI and Anthropic models, playing a similar hand to VeniceAI.

4. ChronoLLM (SN38) is training a frontier model that reason only on information available at each point in time, built by CrunchDAO and aimed squarely at hedge funds.

ChronoLLM (SN38) Launches Vintage AI Models To Eliminate Look-Ahead Bias In Historical Analysis

The blue chips are compounding at the same rate.

1. IOTA (SN9) kicked off the largest decentralized pipeline-parallel training run to date with 16.2B parameters, 10 stages, and 18 replicas.

2. Metanova (SN68) crossed from virtual to physical drug discovery, beginning nanobody production with Yalotein.

3. Actual (SN95) released a beta for clustering local machines and sharing inference capacity with friends.

Everything Worth Knowing About Actual Computer (SN95)

4. Lium (SN51)‘s GPU rental revenue reached $12M annualized, giving the fund one of its cleanest portfolio-wide revenue signals.

The Largest Cult on the Network

Kassab closes by arguing that Bittensor remains the strongest contender in the emerging decentralized AI sector.

The letter notes that more respected fund managers are now echoing Unsupervised Capital’s early conviction that decentralized AI could become crypto’s biggest movement since Bitcoin.

While the space has become more competitive, Bittensor continues to lead in research output, revenue, business diversity, technical talent, liquidity, and institutional adoption.

The conclusion is that Bittensor already possesses the foundation to capitalize on whatever major AI narrative emerges next.