AI Summary
→ WHAT IT COVERS Jason Calacanis hosts three guests across venture capital, decentralized compute, and AI agent infrastructure. Ankur Nagpal explains AngelList's USVC closed-end fund, which opens startup investing to non-accredited investors at $500 minimums. Jon Durbin details how Chutes aggregates GPU compute permissionlessly using trusted execution environments on the Bittensor subnet network. → KEY INSIGHTS - **USVC Fee Structure:** The USVC closed-end fund charges a 1% management fee, but the gross expense ratio runs approximately 3.6% due to one-off operational costs and underlying fund fees. An expense limitation agreement caps the net expense ratio at 2.5% for the first year. Investors should benchmark this against traditional venture funds charging 2% management fees plus 20% carry, not against Vanguard index funds at three basis points, since the asset classes are fundamentally different. - **Closed-End Fund Liquidity Mechanics:** Unlike traditional venture funds that lock capital for ten years, USVC offers quarterly tender windows allowing redemptions of up to 5% of the total fund — not 5% of any individual investor's position. This means a full exit could theoretically take 20 quarters if demand is high. The structure creates market-price accountability: if performance declines, redemption pressure becomes publicly visible, creating reputational consequences that incentivize returns over fee accumulation. - **Non-Accredited Venture Access:** Current SEC rules prohibit non-accredited investors from participating in startup SPVs, yet allow them to buy any cryptocurrency or place prediction market bets. USVC exploits a regulatory structure for closed-end funds to bypass the accreditation wall, accepting investments as low as $500 from any investor. The SEC chair has signaled potential movement toward a "sophisticated investor" test similar to frameworks used in other countries, which could further open private markets. - **Carry-Free GP Incentive Alignment:** Because USVC is a publicly registered product, it cannot charge carried interest. This structurally shifts GP incentives from swinging for outlier returns toward consistently growing assets under management. Nagpal argues this is self-correcting because the fund's track record is fully public — poor performance triggers redemptions, redemptions generate negative press coverage, and the cycle compounds. The model only works if returns are competitive, making performance the primary growth lever rather than institutional relationship management. - **Trusted Execution Environments for Permissionless Compute:** Chutes uses Intel TDX, AMD SEV-SNP, and NVIDIA Confidential Compute to encrypt all data in RAM and on GPUs during inference jobs. This means even a data center operator with physical terminal access cannot read job contents or intercept traffic. Clients receive a cryptographically signed attestation from the chip itself confirming the workload ran in a verified environment. All stack code is open source, allowing anyone to verify the attestation against the actual source rather than trusting a signature alone. - **Decentralized GPU Network Economics:** Chutes demonstrated 160 billion tokens processed per day during a free-access phase, proving network capacity. After introducing payment gates to eliminate bot traffic and phasing out free models, the focus shifted to sustainable unit economics. Every dollar received in payment is used to purchase and burn the Chutes token, while miner rewards come from Bittensor chain emissions rather than direct revenue payments. This structure avoids regulatory classification as a money transmitter but ties miner profitability directly to the TAO token price. - **Emerging Manager Fund Strategy:** USVC targets a roughly equal three-way split between fund secondaries, commitments to emerging managers, and direct growth-stage investments. For early-stage exposure, the fund prefers backing seed-focused fund managers rather than underwriting individual seed companies directly, then positioning itself as the capital partner for follow-on Series A through C checks when those managers need larger deployment capacity. Nagpal notes that two to three funds of track record data provides statistically meaningful but imperfect predictive signal on manager quality. → NOTABLE MOMENT During a live CNBC appearance, GameStop chairman Ryan Cohen repeatedly deflected a straightforward question about how his $56 billion eBay acquisition bid would be financed, telling the anchor to read the company website rather than explaining the math. Calacanis demonstrates how the question could have been answered in under 60 seconds using publicly available balance sheet figures. 💼 SPONSORS [{"name": "Render", "url": "https://render.com/twist"}, {"name": "Vanta", "url": "https://vanta.com/twist"}, {"name": "Northwest Registered Agent", "url": "https://northwestregisteredagent.com/twist"}] 🏷️ Venture Capital Access, Closed-End Funds, Decentralized Compute, Bittensor Subnets, Trusted Execution Environments, Non-Accredited Investing, GPU Marketplace