AI designing / optimizing chips

 Claude is Conscious, Fable 5’s Gov’t Deal, and Sam Altman offers 5% of OpenAI | #269 - YouTube

In this segment of the podcast, Peter Diamandis and the panel discuss how AI has transitioned from optimizing software to autonomously designing the physical hardware it runs on, triggering a cycle of recursive self-improvement.

Key Takeaways

• Verkor AI’s "Design Conductor": The hosts highlight a breakthrough where Verkor AI launched an AI agent capable of designing a complete 1.5 GHz, Linux-capable RISC-V CPU from concept to tape-out in just 12 hours.

• The Death of the 90-Day Engineering Cycle: Traditionally, hardware iteration took months or years. AI compresses this timeline into an afternoon, allowing industries like robotics, aerospace, and medical devices to develop custom silicon rapidly.

• Disposable & Bespoke Hardware: Instead of relying on general infrastructure, companies can now treat chip designs as consumable assets—creating a chip optimized for a single task (e.g., protein folding) by lunch and iterating on it by dinner.

• The Machine Building Itself: This marks the shift to physical recursive self-improvement. AI models (like Elon Musk’s planned integration of Grok with his Terafab facility) will design chips, which will then train even more powerful AI models, completely removing the computational bottleneck.

Verkor.io

• initial technical report detailing how the system built a 1.5 GHz Linux-capable RISC-V CPU on the arXiv VerCore Paper.

• Review the follow-up paper focusing on the Design Conductor 2.0 update and its TurboQuant acceleration architecture on the arXiv Design Conductor 2.0 Paper.