Bittensor’s Rise, Meta’s Llama Goes Cloud, & AI Now Writes Your Code | E2119

Episode Summary

Executive Summary: The episode centers on BitTensor, a Bitcoin-inspired crypto network using token incentives to build decentralized AI compute and services, with subnets like Shoots and Targon competing for TAO emissions. The panel argues this model can cut cloud costs, accelerate AI infrastructure, and create real utility—while also contrasting it with crypto’s darker use cases, the rise of AI-generated code, VC outreach best practices, and the continuing debate over return-to-office culture.

Main Topics: BitTensor, TAO, and decentralized AI compute (Priority: 5/5): Mark Jeffrey explains BitTensor as a Bitcoin-like incentive network for AI and compute, where TAO is earned by supplying GPU resources and subnets provide specific services such as AI inference hosting. Subnet economics and governance (Priority: 5/5): The conversation details how subnets like Shoots and Targon operate, how staking TAO helps launch a subnet, and how subnet market caps influence emissions and resource allocation. Crypto legitimacy, crime, and stablecoins (Priority: 4/5): The panel discusses Tether’s reputation, illicit use in crypto versus cash, and whether stablecoins are likely to become mainstream and regulated. AI infrastructure and the commoditization of compute (Priority: 4/5): The discussion connects BitTensor to broader AI infrastructure trends, including Meta’s Llama hosting partnerships and the likelihood that AI compute prices will keep falling. AI-generated code and startup productivity (Priority: 4/5): The hosts cite major companies reporting that 20-30%+ of code is now AI-generated and argue this will dramatically lower startup engineering bottlenecks. Founder outreach, cold emails, and fundraising (Priority: 3/5): The panel critiques AI-generated spam and revisits how founders should craft concise, personalized, high-signal outreach to VCs. Return-to-office and professional development (Priority: 3/5): The episode closes with a discussion of why companies are pushing in-person work, emphasizing intensity, mentorship, and access to decision-makers.

Key Arguments: BitTensor borrows Bitcoin’s incentive design but redirects it toward useful compute rather than proof-of-work busywork. The subnet structure creates a market mechanism to allocate TAO emissions toward the most useful services, reducing extractive behavior. Decentralized compute can offer major cost advantages over AWS, reportedly around 85% cheaper in some cases. The most valuable AI infrastructure will trend toward lower costs and broader availability, putting pressure on incumbents like OpenAI and traditional cloud providers. Crypto’s illicit-use problem is real, but the panel argues cash is still used for far more crime than crypto or Tether. AI is already writing a meaningful share of production code, which will significantly reduce startup engineering constraints. Effective founder fundraising still depends on targeted, human, high-context communication rather than mass automated outreach. In-person work remains important for speed, culture, mentorship, and proximity to leadership, especially for younger workers and ambitious operators.

Data Points: BitTensor age: about 4 years - Mark Jeffrey says the project is roughly four years old. TAO total supply: 21 million - BitTensor is modeled after Bitcoin’s capped supply. TAO in circulation: about 8 million admitted so far - Mark cites the number of coins already released. Number of subnets: 100 - Jeffrey says BitTensor has about 100 subnets in its ecosystem. Shoots market cap: $89 million - Shown while discussing the top subnet for decentralized AI compute. Shoots subnet age: 129 days - The subnet is described as relatively new but already substantial. Shoots coin price: $98 - Displayed as the subnet token price during the walkthrough. Shoots emissions share: 16% - Used to explain how emissions are allocated to successful subnets. Targon market cap: $44 million - Referenced as another major subnet in the ecosystem. Subnet launch stake: $400 right now - Mark says staking TAO to start a subnet currently requires about this much. TAO token price: about $377 per coin - Used to approximate the capital commitment for subnet participation. Compute savings vs AWS: about 85% less - Claim made for Shoots as a decentralized compute alternative. Crypto illicit transaction volume: $40-50 billion per year - Cited as an estimate for illicit crypto activity overall. AI-written code at Google: over 30% - Sundar Pichai reportedly said this during earnings commentary. AI-written code at Microsoft: 20-30% - Satya Nadella reportedly said this at LlamaCon. Cursor output: 1 billion lines of accepted code per day - The CEO shared this as a productivity metric. VC landscape: 400 funds formed a year / 1,500 active funds - Jason estimates the size of the venture capital target universe for founders.

Pivotal Quotes: "BitTensor is the open source project to replicate this. Tau is the coin." — Mark Jeffrey: Explaining the relationship between the BitTensor network and its native token. "It’s about 85% less than what it costs on AWS." — Mark Jeffrey: Describing the cost advantage of decentralized AI compute via Shoots. "The average criminal cryptocurrency is being used for between $40 and $50 billion of illicit transactions per year. That would be a magnitude less than US dollars." — Mark Jeffrey: Responding to concerns about Tether and illicit finance.

Implications: If the claims hold, decentralized compute networks like BitTensor could pressure cloud prices, reshape AI infrastructure, and create new token-based markets for useful work. At the same time, AI is automating more software creation, making execution, distribution, and human relationships even more important.