Bittensor’s (alleged) $10M rug pull (feat. Mark Jeffrey) | E2275

Watch on YouTube ↗  |  April 14, 2026 at 00:44  |  1:17:56  |  This Week in Startups
Speakers
Mark Jeffrey — Partner at Stillcore Capital, Investor in Bittensor Subnets
Will Squires — Head of Content, CoinDesk
Jason Calacanis — Angel Investor / Founder, LAUNCH

Summary

This episode dives into the recent alleged $10M rug pull by a Bittensor subnet operator (Covenant AI's Sam) and its impact on the TAO token and ecosystem. Guests discuss governance fixes, investment strategies in TAO and subnets, and showcase promising subnets like BitMind (deepfake detection) and IOTA (distributed AI training). The conversation highlights Bittensor's potential as a decentralized AI platform and the risks and opportunities for investors.

  • Detailed breakdown of the alleged rug pull by Covenant AI's Sam, who dumped $10M in tokens and left the project.
  • Discussion on governance improvements to prevent future rug pulls, such as token lockups for subnet operators.
  • Showcase of BitMind (Subnet 34), an AI security company focused on deepfake detection with VC backing.
  • Introduction to IOTA (Subnet 9) by MacroCosmos, which aims to enable distributed training of large AI models using interruptible compute.
  • Mark Jeffrey's investment thesis: focus on subnets with product-market fit and revenue, while also being bullish on TAO as diversified exposure.
  • Comparison of investing in TAO (like a mutual fund) vs. picking individual subnets (like angel investing).
  • Debate on whether the rug pull was due to decentralization issues or simple greed.
  • Insights into how Bittensor's incentive alignment engine works and where it can fail.
Trade Ideas
Mark Jeffrey Partner at Stillcore Capital, Investor in Bittensor Subnets 3:11
Bittensor is the Linux of AI with strong early signal.
For most investors, buying TAO provides a diversified 'mutual fund' exposure to the entire Bittensor ecosystem, similar to investing in a venture fund. Picking individual subnets is more like angel investing and requires significant research. He notes that subnet investing is open to non-accredited investors, unlike traditional startup equity.
Mark Jeffrey Partner at Stillcore Capital, Investor in Bittensor Subnets 10:30
Invest in subnets with product-market fit and revenue.
Invests in specific Bittensor subnets that demonstrate product-market fit, revenue, and mature teams, similar to evaluating traditional startups. Examples include Ridges, Shoots, Hippius, Vid.io, and Score. Also invests in research subnets like NOVA (pharmaceutical molecule discovery) which are highly speculative but could be enormous if successful.
BitMind subnet detects deepfakes with self-improving system.
BitMind (Subnet 34) is an AI security company focused on deepfake detection and proof-of-human verification. It uses a decentralized, permissionless system where miners compete to develop detection models and generate data to fool them, creating a continuously learning system. The company has raised $3M from VCs and targets enterprise and consumer use cases in media, social media, and government.
Will Squires Head of Content, CoinDesk 42:00
IOTA enables distributed training of large AI models.
IOTA (Subnet 9) aims to orchestrate distributed compute worldwide to train frontier-scale AI models at significantly lower cost. It uses interruptible compute from sources like idle laptops or data center surplus, allowing training with 'fractal compute' (short, variable intervals). The project offers a train-at-home application for easy participation and targets the high cost and inefficiency of centralized training.
Up Next

This This Week in Startups video, published April 14, 2026, features Mark Jeffrey, Ken Miachi, Will Squires discussing TAO, Ridges, Shoots, Hippius, Vid.io, Score, NOVA, BitMind, IOTA. 4 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Mark Jeffrey, Ken Miachi, Will Squires  · Tickers: TAO, Ridges, Shoots, Hippius, Vid.io, Score, NOVA, BitMind, IOTA