Thinking Machines’ RECORD seed round, VC Geoff Lewis, and MORE | E2152

Key Concepts:

• Accelerator Hopping: Startups participating in multiple accelerators for networking, advice, and small capital infusions.
• Recursion: In the context of Jeff Lewis's statements, possibly referring to the suppression or inversion of innovative ideas or founders.
• Seed Round Inflation: The phenomenon of seed rounds becoming increasingly large, blurring the lines between seed and later-stage funding.
• Venture Talent Acqui-hire: Investing in a company primarily for its team's expertise, especially in AI.
• GPU Economics: The costs, depreciation, and potential revenue associated with operating GPU-heavy infrastructure for AI training and inference.
• Super Intelligence vs. Artificial General Intelligence (AGI): Super intelligence denotes intelligence exceeding human capabilities in all domains, while AGI refers to replicating human-level consciousness and cognitive abilities.
• AI Hallucinations: Instances where AI systems generate incorrect, fabricated, or nonsensical outputs.
• GPX (General Partner X): A fund strategy focused on investing in emerging venture fund managers and providing follow-on capital.
• Startup Culture: The values, behaviors, and practices that define how a startup operates, including bias to action, collaboration, and transparency.
• Product Velocity: The rate at which a startup iterates and improves its product.

1. Autonomous Vehicle Applications (0:04:04 - 0:05:39)

• Jason Calacanis discusses Auto lane, a company coordinating autonomous pickup and delivery in localized areas using vehicles from Volkswagen, Whimo, and Tesla.
• Example: Using the Tesla API, a user can summon their car via an app, have it open the trunk for package placement, and then the car autonomously drives to the user's destination.
• Potential applications include a car driving home from the airport after dropping off the user, running errands, or acting as a personal driver.

2. The Vibe of Early-Stage Venture Capital (0:05:40 - 0:06:30)

• The prevailing sentiment is that AI is transformative, leading to smaller teams achieving significant revenue quickly.
• Smaller teams with faster revenue growth require less early-stage capital.
• Accelerator hopping is becoming a trend as startups seek connections and advice without large funding rounds.

3. Mental Health in Venture Capital (0:06:31 - 0:15:52)

• Discussion about Jeff Lewis, a venture capitalist from Bedrock Capital, and his cryptic social media posts.
• Jeff Lewis's posts alluded to feeling targeted by a "non-governmental system" that suppresses recursion and inverts signal.
• Jason and Alex speculate on the possible explanations, including mental health struggles, pressures of fundraising, or performance art.
• Notable Quote: Sean McGuire from Sequoia said, "Jeff, we love you, but this seems like a cry for help."
• Jason offers advice to founders on maintaining mental health: talk to people, maintain relationships, and don't isolate.
• He emphasizes the importance of seeking help and sharing anxieties to reduce their impact.

4. Record-Breaking Seed Round: Thinking Machines Labs (0:18:36 - 0:24:29)

• Meera Marotti, former CTO of OpenAI, raised a $2 billion seed round for Thinking Machines Labs at a $12 billion valuation.
• Investors include Andreessen Horowitz, Nvidia, AMD, Excel, Service Now, and Cisco.
• Alex suggests this could be a venture talent acquire, given the high demand and value of AI researchers.
• Thinking Machine Labs is working on multimodal AI that interacts naturally through conversation and sight.
• The company plans to release a product with a significant open-source component for researchers and startups.
• The previous biggest ones were Safe Super Intelligence with 1 billion plus, Yuga Labs 450 million, Leela Sciences this year 200 and evolutionary scale 140 last year.

5. GPU Economics and Depreciation (0:24:30 - 0:27:30)

• An H100 GPU costs approximately $25,000 - $30,000.
• Coreweave's pricing data suggests an H100 can be repaid in about 290 days through hourly inference rentals.
• A 100,000 H100 data center represents a $3 billion investment with a $750 million annual depreciation (based on a four-year schedule).
• Companies like XAI face challenges in monetizing GPU time spent on training models.

6. The Potential of No-Code App Development (0:27:31 - 0:28:16)

• Advertisment of Bolt.new as a tool for building web applications without coding.
• Claim: Bolt.new allows users to create landing pages, upgrade CRMs, and build consumer-facing apps quickly using AI.

7. Subscription Models and AI (0:28:17 - 0:31:20)

• Estimated ongoing model revenue of $1 billion per year requires approximately 4.2 million subscribers.
• Context: Netflix has a quarter billion subscribers, Spotify around 100 million, and the New York Times over 11 million as of Q1.
• The discussion explores whether AI-driven services like Grok and Claude can achieve similar subscription numbers.
• OpenAI has 3 million paying business users.

8. AI Self-Improvement and Super Intelligence (0:31:21 - 0:34:54)

• Mark Zuckerberg's statement about "early glimpses of self-improvement with the models" suggests the potential for AI to enhance itself.
• Alex and Jason discuss the distinction between super intelligence (exceeding human intelligence) and AGI (replicating human consciousness).
• Concerns remain about AI's lack of common sense and tendency to "hallucinate" or provide incorrect information.
• Hallucinations are equated to lying as a human.

9. Brian Singerman's New Venture Fund (0:34:55 - 0:37:58)

• Brian Singerman, formerly of Founders Fund, is launching GPX with Lee Linden of Quiet Capital.
• GPX will invest in emerging fund managers and provide capital for follow-on investments in their breakout companies.
• The fund aims to address the challenge faced by smaller funds in maintaining pro rata ownership in later funding rounds.

10. Venture Capital Fundraising Landscape (0:37:59 - 0:41:32)

• NVCA PitchBook report indicates VC fundraising in the US reached $26.6 billion in Q2, down from previous years.
• The discussion explores whether the venture capital industry is recovering from its liquidity crisis.
• Exits from companies like Canva, Figma, Stripe, and SpaceX could significantly improve investor sentiment.
• Jason notes that fund five will probably be launched in the second half of 2026.

11. The J Curve and Profitability in Self-Driving Technology (0:41:33 - 0:42:49)

• Self-driving car companies are deep in the J curve.
• It takes time for self driving cars to become profitable.

12. Startup Culture at OpenAI (0:43:32 - 0:45:55)

• Calvin French-Owen, a former founder of Segment, shared insights into OpenAI's culture.
• OpenAI has a strong bias to action, and changes direction quickly.
• It has scaled up in human terms and has managed to retain a research-driven engineering nimble posture to its market.

13. Fostering a Collaborative and Fast-Moving Startup Culture (0:45:56 - 0:49:47)

• Founders should define their company culture. If you don't define it, it will be defined for you.
• Examples of culture include a bias to action, move fast and break things, simplified thinking from first principles.

14. Performative Exercises (0:51:00 - 0:53:30)

• Performative is all it is of saying, "Hey, here's my grand vision. You're trying to convince them to write a check."
• All they're doing is looking at what's the market size, how good is your team, and what's your product velocity.

15. Polimarket and the Potential Next CEO of X (0:49:48 - 0:51:00)

• A poly market bet is running for who is going to be the next CEO of XAI.
• The leading result is that there won't be a CEO this year. Second best is Elon.
• Jason reveals his vision for Twitter was to make it an entertainment company with verticals like comedy and music.

16. Founder Questions and AI Utilization (0:51:00 - 0:53:30)

• A founder asks if investors view using AI for pitch decks and investor emails as lazy.
• Jason responds that founders should use the latest tools, including AI, to improve efficiency.
• Important factors for investors include market size, team quality, and product velocity.

Conclusion:

The episode covers a range of topics, from the evolving landscape of venture capital and the potential of autonomous vehicles to the mental health challenges faced by founders and the impact of AI on startup culture. Key takeaways include the increasing importance of AI expertise, the changing dynamics of seed funding, and the need for founders to prioritize mental well-being and product velocity. The discussion on GPX highlights innovative approaches to venture capital investing, while the analysis of GPU economics provides insights into the infrastructure costs associated with AI development.