BitTensor — First Memo (Apr ’23)

BitTensor is a decentralized neural network that incentivizes machine intelligence.

By Escape Velocity

What Is BitTensor?

BitTensor is a two‑sided network of ML apps and models where miners (GPUs) run inference on models and validators (CPUs) query, rank, and reward useful responses. End‑users access apps (e.g., ChatTensor) by delegating $TAO to validators. Rewards are split between validators and miners based on usage and measured utility (“proof‑of‑intelligence”).

Why We’re Paying Attention

  • Real AI/ML DNA: Founders and core team include multiple machine‑learning engineers (ex‑Google, IBM; Instacart, Workday) with deep technical chops.
  • Breakout network scale: 100+ validators and 3,400+ miners reported; competitive dynamics have already forced rapid model scaling across the network.
  • Community & ethos: Bitcoin‑inspired issuance with no premine, 21M max supply, and halving; distribution driven by mining/delegation rather than insider allocations.
  • Ecosystem growth: After decentralizing consensus, developers can launch permissionless subnets for text, image, multimodal, search, and more.

Network Scale & Momentum

The network has surpassed 100 validators and 3,400 miners. Community competition has increased hardware and model sizes (from ~2.7B parameter LLMs early‑year to 6B+ to remain competitive). The protocol has already weathered adversarial events, including early hard‑forks and suspected DDoS. With subnets live, several services reportedly generate material annualized rewards.

Token Design & Distribution

BitTensor mirrors Bitcoin’s schedule: no premine, 21M cap, and four‑year halvings. Inflation today is elevated (similar to BTC circa 2012) as circulating supply grows toward ~10.5M at the first halving, now expected around Q4’25–Q1’26. Rewards are funded by issuance and distributed via delegation to value‑adding validators and utility‑proven miners.

Where It’s Early / Key Risks

  • Product‑market fit: Early applications (e.g., ChatTensor) were limited; subnets have broadened scope, but sustained end‑user demand remains to be proven at mainstream scale.
  • Decentralization mechanics: Delegated proof‑of‑stake concentrates influence with high‑stake validators (including the foundation); governance and funding need to remain healthy for long‑term progress.
  • Exchange access: Historically limited CEX presence; growth path is more “Bitcoin‑like” (organic hash‑power accrual) than catalyst‑driven listings.

Mining Economics (Snapshot)

Mining resembles “horse racing”: source a differentiated dataset, train a competitive model, register (fee‑based or PoW prelims), then mine for rewards. Historical estimates suggested competitive miners could reach short paybacks when renting GPUs, though difficulty has risen and retail miners without unique data/models face headwinds.

Outlook

We’re tracking technical validity of proof‑of‑intelligence and signs of durable developer traction. Given supply dynamics (high staking rates) and increasing mindshare as a leading “AI token,” we expect attractive entry points as fundamentals compound.

References

Get a PDF version of the original memo:

BitTensor — First Memo (Apr ’23) by Escape Velocity

📄 Download PDF