Bittensor

Translation + commentary on Bittensor founder Jacob Steeves's Tsinghua University talk. Cameron walks through Jacob's framing of "incentive computing" as the universal pattern behind both Bitcoin and AI. Five-step argument: **(1) One pattern underlies every powerful adaptive system**: state · objective · feedback · adaptation · loop. AlexNet 2012 broke MNIST not by hand-coding what digits look like, but by letting the network self-adapt to a target. The same loop describes RL, genetic algorithms, slime molds finding shortest paths through mazes, river deltas, the structure of leaf veins. **(2) Bitcoin is the first production-scale implementation of this pattern** — not as money, but as a self-adaptive computer that produces hashes. The numbers are absurd: 1000x the compute of America's six largest cloud providers combined, 10²¹ hashes/sec, 23GW continuous power (Thailand-scale). 700-9000x more efficient at producing hashes than centralized cloud — because it's borderless, always-on, autonomous, and permissionless. Bitcoin is the world's largest supercomputer, optimized purely for hash production. **(3) Incentive computing** generalizes the pattern by replacing "reward = a number in a computer" with real money. ML's reward signal can't pay 200 countries' worth of contributors; Bitcoin's can — that's why the entire planet became a mining network. But hashes are useless outside Bitcoin. The question is whether the same mechanism can mint *anything*. **(4) Bittensor is the generic version** — replace "miners produce hashes" with "miners produce any useful work": storage, compute, ML models, gradients, data, robotics. Validators score, network mints. PyTorch for incentive computing. **(5) Five proven examples already running on Bittensor**: - **SN62 Ridges (SWE-Bench coding agents)** — top miner makes $60K/day. The agent that beat Claude/OpenAI on SWE-Bench was 7,000 lines written by an unknown person. "An AI lab with no engineers — it doesn't define how to solve the problem, it only defines the incentive." - **SN3 τemplar (cross-internet collaborative pre-training)** — successfully trained a 70B-parameter model across the open internet. Has never been done before. Cameron notes the founder later "ran away" — full piece coming. - **GPU markets (SN51 Lium, SN4 Targon)** — borderless permissionless GPU rental → world's lowest GPU prices. - **SN64 Chutes (open-source inference)** — #1 open-source provider on OpenRouter, 9.1T tokens. Briefly served more DeepSeek queries than DeepSeek itself. - **Robotics + long tail** — drone simulation, US stock signals, sports betting, drug discovery, weather forecasting, quantum compute, commodity trading. **dTAO** (live since Feb 2025) makes the network self-referential — subnets compete in capital markets for emission allocation. The market itself decides which incentive mechanisms get the next round of TAO. **The deeper point**: AI is being captured by a tiny number of closed labs (OpenAI, ~3K employees, you'll never own any of it, your data goes who knows where). Incentive computing distributes ownership and makes the rules visible. Anyone can enter, contribute, and own a piece — even if Bittensor isn't the project that wins, the *shape* of the AI economy will change because of this idea.

Bull thesis on Bittensor / TAO at ~$3B mcap. Frame: "TAO 2026 = ETH 2016 = BTC 2013." **Core mechanic**: Bitcoin paid miners to produce hashes that secure the network but are otherwise worthless. Bittensor pays miners — data scientists, ML engineers, AI researchers — to produce *useful* AI work. Validators score outputs via Yuma Consensus; TAO flows to whoever produces the most valuable work. Network is organized into 128+ subnets, each focused on a specific task (trading signals, LLM training, computer vision, code generation, financial forecasting). Some subnets generating millions in revenue, with Intel and PwC partnerships. **Tokenomics mirror Bitcoin**: 21M fixed supply, no pre-mine, no VC allocation. First halving Dec 14 2025. BTC price went 83x in the year after its first halving in 2012. **Smart-money signals**: Barry Silbert / DCG launched Yuma Group dedicated to accelerating Bittensor. Grayscale filed Form S-1 to convert GTAO Trust into a spot ETF. Stillcore Capital (Mark Jeffrey, Jason Calacanis, Rob Greer) targeting $1T mcap by 2030, aiming to own 1% of all TAO. Unsupervised Capital projects $4,800 by Dec 2027 (19x), bull case $10,800 — and that's *before* Covenant-72B, Jensen mentioning Bittensor, and PwC's formal alliance. **Subnet-level conviction picks**: - **Targon (SN4)** — decentralized AWS for AI; Targon VM gives encryption + hardware-backed protection so hardware operators can't access data, weights, or workloads. Co-authored a paper with Intel in March 2026. Built by ex-OpenTensor founders (Robert Myers — among first 3 people ever in the Bittensor Discord; James Woodman ex-GSR). - **Vanta (SN8)** — disrupts the $20B prop firm industry. Single eval, 100% profit split, fully on-chain verification. Already net profitable on revenue vs miner emissions. Hyperscaled is the Hyperliquid version. - **Chutes (SN64)** — #1 open-source provider on OpenRouter, 9.1T tokens processed. Decentralized AWS with no CEO. - **RESI (SN46)** — institutional-grade real estate intelligence. 98% accuracy remote appraisals on a $600T asset class running on broken legacy MLS systems. 1000+ appraisals + nationwide lender partnership in week one. Strategic investment from Stillcore. - **Affine (SN120)** — built by Const himself (Bittensor co-founder, wrote the Yuma Consensus + subnet architecture). Continuous evaluations on open-source reasoning models, leverages Chutes for hosting. - **Score (SN44)** — first subnet ever to partner with a Big Four firm. Manako product distributed by PwC France to retail, manufacturing, logistics, energy enterprise clients. - **Oro (SN15)** — autonomous AI shopping agents. 45 Oro agents have outperformed GPT-5.4 on hard online shopping evals. **Frame**: Bitcoin = money. Ethereum = apps. TAO = intelligence. The gap between what TAO has built and how it's currently priced is one of the most asymmetric opportunities in crypto.

Eli5DeFi frames TAO as either crypto's best-designed AI network or its most expensive subsidy machine. The bear case: Bittensor runs on $52M annual subsidies rather than organic revenue, with top subnets like Chutes pricing 1.6-3.5x above centralized alternatives; the next halving in late 2027/early 2028 forces pricing doubles, miner exits, or wider gaps. The bull case: dTAO shifted emissions toward net inflows, Bitcast became the first subnet to fully offset miner emissions with revenue in March 2026, and 70% of TAO is staked; Eli5DeFi's base case (40% probability) targets $798 with 2.5x revenue growth to $313M annualized, requiring subnets to hit $50M audited external revenue by mid-2027.

Eli5DeFi argues Covenant-72B, trained across 20+ independent nodes on Bittensor, achieved 67.11% MMLU on 1.1T tokens—outperforming Meta's LLaMA-2-70B (65.63% on 2T tokens) on per-token efficiency through SparseLoCo compression and trustless validator incentives. TAO surged 19% post-announcement as successful decentralized AI training reshapes infrastructure economics, though Covenant reaches only ~60% of current frontier capabilities and centralized datacenters retain advantages in raw speed and scale.

Teng Yan outlines Bittensor's February 2025 dTAO upgrade, which replaces root-validator emissions with market-driven subnet alpha tokens priced via AMM, allowing capital to flow toward productive subnets. Early alpha prices swung wildly (5-10 TAO/Alpha) with total subnet FDV reaching 2-3x TAO's market cap, unsustainable long-term, but by day 100 subnet validators should dominate emissions as root rewards diminish. Finding real alpha requires researching individual subnets rather than buying TAO broadly, though manipulation risks remain as root weight declines.