Excerpts from our Q2'26 Investor Letter
The following is Unsupervised Capital's Q2'26 Letter to its Partners, with sensitive information redacted. Disclaimers below.
There’s a quote from Chris Burniske that lives in my head rent free. “Bitcoin is like a mirror, its hottest narrative reflecting what society needs most.”
From internet money to freedom money to an inflation hedge to a geopolitical hedge, Bitcoin never settled on one story. It kept morphing into whatever the moment called for.
I was chatting with an investor recently who was asking how Bittensor kept surviving like a cockroach, taking on one new narrative after another. He meant it as a critique, but it’s actually a core part of the fund’s thesis. Bittensor is becoming the mirror for AI the same way Bitcoin is for money.
The main story in AI last quarter was sovereigns meddling in access to frontier intelligence. Anthropic was forced to pull Fable and Mythos for several weeks to comply with US export controls. Rumors circulated that China may implement their own export controls on models developed by Chinese AI labs. Access to frontier intelligence turned out to be more fragile than anyone assumed. Right now, Bittensor is the geopolitical hedge for AI. It removes the fragility and uncertainty that stems from centralized development and control of AI systems.
Whatever the next narrative turns out to be, distributed training, data quality as the bottleneck, autonomous agents that finance their own ascent, or something else entirely, Bittensor will absorb it. It's the only ecosystem with the full stack, from agent frameworks at the top to the GPUs underneath. It's internet native and agent native. It has everything, so it can become anything.
Bittensor’s Reflexivity
Bitcoin got memed into a trillion dollar asset through financial reflexivity. Initially, its value was pure social belief. There was nothing underneath it but people who wanted a currency beyond the reach of any sovereign.
What turned that belief into a durable store of value was a feedback loop between price and the network’s most important measure of health: the compute securing it. Higher price meant more mining revenue, which meant more compute defending the chain, which meant a network that was harder to attack. Price going up actually made the tech better.
Musk ran a version of this play outside of crypto with Tesla. He sold the vision years before the fundamentals existed. As investors bought in, Tesla's rising valuation unlocked ever cheaper capital, and that capital brought the engineers, the factories, and the output that let the fundamentals catch up to the narrative.
Bittensor has the same reflexivity as Bitcoin, pointed at intelligence instead of security. Because every subnet token is denominated in TAO, a higher TAO price lifts the productive capacity of every subnet at once. When TAO goes up, every subnet can pay its miners more.
If TAO doubles, Targon's miner rewards double in dollar terms, so miners can profitably deploy more compute. The same is true for Metanova, where higher rewards attract stronger engineers and machine learning researchers.
I'm not saying Bittensor is a meme today the way early Bitcoin was. Subnets are already generating tens of millions in ARR and are contributing at the frontier of AI research.
But conviction can still manifest the network everyone wants Bittensor to become. As more people conclude that centralized AI needs an alternative and that TAO is the way to bet on it, subnet funding increases. Better funded subnets produce better intelligence. Better intelligence proves the original belief right, and the flywheel takes another turn.
Performance and Market Thoughts
You'd expect a manager whose entire thesis is Bittensor to constantly tell you he’s more bullish than ever. So instead of than asserting it again, let me point to the actual developments:
Const is back at the helm. The network had been slowed by democratic process it wasn't ready for. Const is effectively operating as CEO again, and upgrades are shipping faster than they have in a year. More on this here.
Recent upgrades have targeted value accrual for subnet tokens and TAO. The chain now uses emissions to buy back subnet tokens instead of only injecting liquidity. For many subnets, chain buybacks now cover all the miner emissions, which removes the single largest source of structural selling. With most discretionary selling already exhausted, I think a bottom is in on the highest quality subnets, and we're positioned for it. More on this here.
Beyond protocol changes, new subnets are also drawing outside attention back to the network.
Minos co-authored a paper with OpenAI on agentic AI in scientific computing.
Engy is challenging Chutes, running open source models like Kimi-K3 on 5090s.
GM launched a fully private inference gateway for OpenAI and Anthropic models, similar to VeniceAI.
ChronoLLM is building a frontier model that reasons only on information available at each point in time. Built by CrunchDAO, and the use case for hedge funds is obvious.
Progress on the blue chips has been just as strong.
IOTA kicked off the largest decentralized pipeline-parallel training run to date. 16.2B params, 10 stages, 18 replicas.
Metanova is beginning nanobody production with Yalotein, taking virtual drug discoveries into physical synthesis to test viability.
Actual released its beta. Cluster your local machines, run inference on them, and invite friends to use your compute.
Lium revenue jumped to $12M annualized from GPU rentals.
[Redacted]
Outlook
More reputable fund managers are now echoing what we’ve called from the jump: decentralized AI will be the largest crypto movement since Bitcoin. There are far more credible crypto-AI projects this cycle than last, but Bittensor still leads on nearly every dimension I track: most research publications, highest revenue generated, widest diversity of real businesses across the AI stack, densest technical talent network, deepest liquidity, most institutional product offerings, and the largest cult.
Best,
Sami Kassab
Managing Partner
This commentary is provided by Unsupervised Capital Management LLC for informational and educational purposes only. It does not constitute an offer to sell, a solicitation of an offer to buy, or a recommendation of any security, investment product, or investment strategy. Nothing herein constitutes investment, legal, tax, or other advice, nor should it be relied upon in making any investment decision.
This material contains forward-looking statements that are based on current expectations and assumptions and are subject to risks and uncertainties. Actual results may differ materially. References to specific digital assets, protocols, or projects are not endorsements and should not be construed as recommendations to buy or sell.
Digital assets involve substantial risk, including the potential for complete loss of principal. Past performance is not indicative of future results.
