TML Inkling, California Forever?, TSMC Capex | David Baszucki, Everett Randle, Eric Glyman, Jordan Black

Overview

This episode dives into the latest developments in AI, manufacturing, and gaming, featuring discussions on Thinking Machines Lab's new AI model, the challenges of scaling industrial production, and Roblox's ambitious AI-driven initiatives. Guests include Everett Randle (Benchmark), Eric Glyman (Ramp), Jordan Black (Senra Systems), and David Baszucki (Roblox).

Notable Quotes

- Imagine you're that monkey throwing darts, and you just keep hitting the triple 20 every time. It feels weird, but you just keep throwing more darts. – Everett Randle, on the current venture capital climate.

- We trained it to be a broad, balanced foundation model, strong across many domains, flexible enough to adapt. – John Coogan, summarizing Mira Marati's vision for Thinking Machines Lab's Inkling AI model.

- The world needs to play more. Play is how we've evolved as humans. – David Baszucki, on Roblox's mission to foster creativity and connection.

🧠 Thinking Machines Lab's Open Source AI Model

- Thinking Machines Lab (TML) launched Inkling, an open-weight AI model with 975 billion parameters, designed for fine-tuning and customization.

- Tyler Cosgrove highlighted its Red Hat model approach, where clients can access the model's weights but benefit from TML's expertise.

- The model aims to challenge AI giants like OpenAI and Anthropic by offering flexibility and transparency.

- Debate arose over the model's reliance on distillation techniques, with Everett Randle noting that while innovative, it may not be as groundbreaking as claimed.

🏭 Revolutionizing Manufacturing with Senra Systems

- Jordan Black shared how Senra Systems is modernizing wire harness production for aerospace and defense.

- The company integrates AI and robotics to streamline a traditionally manual process, reducing production time and improving quality.

- Senra's approach includes forward-deployed engineers who co-design with clients, ensuring faster prototyping and scalable production.

- Emerging markets like data centers and nuclear microreactors are driving demand for advanced wire harnesses.

💸 AI Cost Management and Enterprise Adoption

- Eric Glyman discussed Ramp's new tool for managing token spend, addressing the rapid rise in AI-related costs (21x growth in a year).

- Companies are encouraged to use frontier models for exploration but transition to fine-tuned open-source models for repetitive tasks.

- Ramp's internal AI spend reached 10% of payroll, showcasing the growing importance of cost control in AI adoption.

- Glyman emphasized the need for visibility and benchmarking to optimize AI investments.

🎮 Roblox's AI-Powered Future

- David Baszucki unveiled Build, a tool enabling users to create games directly within the Roblox app using AI.

- Roblox's infrastructure, optimized for cost and performance, supports its vision of democratizing game creation.

- AI is also being used for NPC development, dynamic world generation, and safety moderation.

- Baszucki teased the potential for Roblox IP to expand into movies and other media, while emphasizing the platform's role as an IP incubator.

🌍 Venture Capital and Market Trends

- Everett Randle reflected on the venture capital landscape, comparing the current AI boom to the SaaS bubble of 2021.

- He noted a shift toward open-source AI models and application-layer companies leveraging fine-tuned solutions.

- Randle highlighted the importance of defensibility in startups, favoring network effects and unique data over reliance on AI labs.

- The conversation touched on geopolitical implications, including China's role in open-source AI and the U.S.'s need to foster domestic innovation.