Jason Calacanis
Partner
- Co-host of the All-In podcast
- Host of This Week in Startups
- 300+ angel investments, including Uber at a $5mm valuation
The story of blockchain|Our view
Bitcoin, then Ethereum, now Bittensor (TAO)
Bitcoin showed that a network with no company could get people all over the world to contribute hardware and electricity, paid in new coins. Ethereum made money programmable. Bittensor applies the idea to AI: the network mints TAO and pays whoever produces the best AI work.
From Bitcoin
The same token design
- A hard cap of 21mm tokens
- Halvings about every four years
- Launched with no premine
- Bitcoin funded about 20 GW of computing power, with no company
From Ethereum
A platform to build on
- Anyone can launch on it
- An economy of apps, here called subnets
- One currency at the center
Bittensor (TAO)
The best of BTC and ETH, applied to AI
Bitcoin pays miners to secure a ledger. Bittensor pays them to serve, train and score AI models.
Co-founded by Jacob Steeves, a former Google Brain machine learning engineer. The idea came from Bitcoin, in 2015.
The worldwide web of intelligence|Our view
We believe Bittensor (TAO) is evolving into the next generation of the internet
In June 1993 there were about 130 websites. Two years later there were 23,500. Bittensor has 128 subnets today. In 1995 you could buy Netscape or AOL, but you could not own the internet. This time you can hold the network’s currency.
The worldwide webA network of information, for people
Bittensor (TAO)A web of intelligence, for people and agents
Websites, each serving pages
Subnets: 128 today and 256 slots on the roadmap, each a market for one AI job
Moves information: documents, pages, people
Moves intelligence: models, answers, reasoning
Used by people, through browsers
Used by people and their agents: with x402, an agent pays per call from a wallet, with no account or API key
Runs on open protocols: TCP/IP and HTTP
Runs on an open protocol: Bittensor
Anyone can publish a site
Anyone can launch a subnet or mine on one
Search engines rank the pages
Validators score every miner’s work
Sites get paid by ads and subscriptions
Whoever does the work is paid in TAO, and by paying customers
A site for every topic and every business
A subnet for every use case and model: a bazaar of AI
Investors could not own the protocols
Anyone can hold TAO, the network’s currency
TAO is for the agents. The next internet’s heaviest users will be agents. Today an agent needs a human to sign it up, issue an API key and attach a card. Any agent with a wallet can buy intelligence on Bittensor, with no approval.
How we generate returns|Our view
Subnets are our portfolio companies.
Most of our portfolio is in subnets. A subnet is a startup inside Bittensor: a competitive market for one AI job, with its own token that trades against TAO. We pick the best of them the way a venture fund picks startups, hold them with the liquidity of a hedge fund, and keep TAO underneath as the reserve asset.
01 · The thesis
A basket of subnets that beats TAO
We believe a selected basket of subnets will outperform TAO itself, the way the best companies beat the market they trade in. Every subnet we hold is priced in TAO, so we start from the network’s return and aim to add to it.
02 · Liquid venture
Venture upside, hedge fund liquidity
Subnets are early-stage companies, but their tokens trade on chain around the clock. We size positions up and down as the network changes, instead of waiting a decade for an exit.
03 · Access
Bought from founders
Many positions come over the counter, directly from founders, in private deals sourced through relationships across the ecosystem.
04 · Operating partner
We help our subnets win
We work with the teams on strategy, go-to-market, partnerships, capital and token design, and put the reach of All-In, This Week in Startups and The Hash Rate behind the subnets we back.
We back subnets that can sell a product. When a subnet’s revenue grows, it buys its own token back, and on the network that means buying TAO first, so the companies we back and the reserve asset we hold pull in the same direction.
Where our subnets sit: the whole AI stack, rebuilt as open marketsExamples, not recommendations
Frontier labsAnthropic, OpenAI
→A decentralized frontier labIOTA · Affine
NeocloudsGPU clouds for AI
→A decentralized neocloudLium · Targon
Inference platformsModel APIs, pay per use
→A decentralized model APIChutes · Engy
Cloud storageAmazon S3
→Decentralized storageHippius
AI application startupsOne company per product
→Decentralized applicationsScore · Minos · RedTeam
Research
We publish the thesis, then test it against the data.
Every Stillcore report, in full and free to read: the founding thesis, the state of the subnet market and our road map to 2030.
No. 1 · March 2026 · 35 pages
The Bittensor Thesis
Our founding case: money, then finance, now intelligence. Why we believe Bittensor (TAO) is the third great crypto network.
Read it
No. 2 · July 2026 · 15 pages
2026 State of Subnets
Our mid-year review of Bittensor subnets: valuations, revenue, the subnets that are winning and what changed on chain.
Read it
No. 3 · October 2026 · X article
TAO 2030: A $1tn+ Ecosystem
Our road map for Bittensor to 2030: what 2026 showed, what the incentive builds and the arithmetic behind $1tn.
Read it
No. 4 · November 2026 · Essay
TAO: The Commanding Heights of Intelligence
Everyone is describing Bittensor (TAO). Almost nobody is pricing it. Where the toll on cheap intelligence gets collected, and how to own it.
Read itFund Overview & ResearchOur fund deck, for accredited investors. It opens on DocSend once you enter your email.
Open the fund overviewFor accredited investors only.