Skip to main content
This Week in Startups

Bittensor Drama! TAO down 15%! | E2274

81 min episode · 3 min read
·

Episode

81 min

Read time

3 min

AI-Generated Summary

Key Takeaways

  • BitTensor Subnet Governance Risk: When a subnet owner controls wallet access, they can unilaterally sell community tokens and exit — as allegedly occurred with Covenant AI's three subnets (339, 81). The proposed fix is requiring subnets to stake TAO as collateral, functioning like a franchise deposit, while giving token holders governance rights that prevent any single operator from liquidating community assets unilaterally.
  • TAO as Crypto ETF Strategy: Buying TAO directly on Coinbase functions like purchasing an index ETF across all 128 BitTensor subnets rather than picking individual winners. Subnet tokens themselves are not freely traded on exchanges yet — acquiring them requires direct deals with subnet owners. This structure lets investors gain broad exposure to decentralized AI infrastructure without identifying which specific subnet will dominate.
  • Video Upscaling Market Sizing: The AI video processing market is projected to grow from $175M in 2025 to $1.1B by 2032. Subnet 85 (Vidaio) targets enterprise archive holders — broadcasters, national archives, Getty Images — offering 60% file size reduction at equivalent perceptual quality, plus colorization, frame rate correction, metadata tagging, and subtitling, all priced competitively against AWS offerings through decentralized miner competition.
  • Permissionless Global Talent Access: BitTensor subnets attract skilled ML engineers from Vietnam, Southeast Asia, and emerging markets who cannot access Google or Coinbase hiring pipelines. These contributors participate anonymously, receive crypto payments without visa or banking friction, and compete on proof-of-work output alone. A four-person Vietnamese university team actively mined Subnet 85, demonstrating that decentralized networks unlock talent pools entirely inaccessible to traditional startup hiring.
  • Multi-Persona LLM Council Framework: Andrej Karpathy's LLM council concept — sending one question to multiple models, anonymizing responses, then running peer review before a chairman synthesizes a final verdict — can be replicated inside Claude Opus using five personas: contrarian, expansionist, first-principles thinker, executor, and outsider. Anonymizing responses before peer review eliminates model bias, producing more balanced outputs than single-model queries on high-stakes decisions.

What It Covers

BitTensor's TAO token drops 15% after Covenant AI's Sam Dair allegedly sold subnet tokens and abandoned the project, accusing cofounder Jacob Steves of blocking operations. The episode also covers subnet 85's video upscaling technology, a Claude-based multi-persona advisory council tool, and the global permissionless workforce powering decentralized AI networks.

Key Questions Answered

  • BitTensor Subnet Governance Risk: When a subnet owner controls wallet access, they can unilaterally sell community tokens and exit — as allegedly occurred with Covenant AI's three subnets (339, 81). The proposed fix is requiring subnets to stake TAO as collateral, functioning like a franchise deposit, while giving token holders governance rights that prevent any single operator from liquidating community assets unilaterally.
  • TAO as Crypto ETF Strategy: Buying TAO directly on Coinbase functions like purchasing an index ETF across all 128 BitTensor subnets rather than picking individual winners. Subnet tokens themselves are not freely traded on exchanges yet — acquiring them requires direct deals with subnet owners. This structure lets investors gain broad exposure to decentralized AI infrastructure without identifying which specific subnet will dominate.
  • Video Upscaling Market Sizing: The AI video processing market is projected to grow from $175M in 2025 to $1.1B by 2032. Subnet 85 (Vidaio) targets enterprise archive holders — broadcasters, national archives, Getty Images — offering 60% file size reduction at equivalent perceptual quality, plus colorization, frame rate correction, metadata tagging, and subtitling, all priced competitively against AWS offerings through decentralized miner competition.
  • Permissionless Global Talent Access: BitTensor subnets attract skilled ML engineers from Vietnam, Southeast Asia, and emerging markets who cannot access Google or Coinbase hiring pipelines. These contributors participate anonymously, receive crypto payments without visa or banking friction, and compete on proof-of-work output alone. A four-person Vietnamese university team actively mined Subnet 85, demonstrating that decentralized networks unlock talent pools entirely inaccessible to traditional startup hiring.
  • Multi-Persona LLM Council Framework: Andrej Karpathy's LLM council concept — sending one question to multiple models, anonymizing responses, then running peer review before a chairman synthesizes a final verdict — can be replicated inside Claude Opus using five personas: contrarian, expansionist, first-principles thinker, executor, and outsider. Anonymizing responses before peer review eliminates model bias, producing more balanced outputs than single-model queries on high-stakes decisions.
  • Equity Structuring for Early VP Hires: For a seed-stage startup with $3M raised and a solo founder holding 85% equity, the LLM council recommended offering a VP of Engineering 0.75–1% equity, structured as ISOs with a four-year vest and one-year cliff, with double-trigger acceleration. Critically, the council flagged creating a formal 10–15% option pool before the negotiation, so the offer reflects the actual post-pool cap table rather than a number that shifts at formalization.

Notable Moment

A four-person team of Vietnamese university students anonymously mined BitTensor's video subnet from Vietnam, earning cryptocurrency without visas, bank accounts, or employer approval. This surfaced a counterintuitive reality: decentralized crypto networks are already redistributing ML engineering work globally in ways that traditional startup hiring pipelines structurally cannot replicate.

Know someone who'd find this useful?

You just read a 3-minute summary of a 78-minute episode.

Get This Week in Startups summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

Keep Reading

More from This Week in Startups

We summarize every new episode. Want them in your inbox?

Similar Episodes

Related episodes from other podcasts

This podcast is featured in Best Startup Podcasts (2026) — ranked and reviewed with AI summaries.

You're clearly into This Week in Startups.

Every Monday, we deliver AI summaries of the latest episodes from This Week in Startups and 192+ other podcasts. Free for up to 3 shows.

Start My Monday Digest

No credit card · Unsubscribe anytime