Episodes
-
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
Runway’s founding thesis was that improving generative models would eventually require a fundamental redesign of creative tools, not simply better standalone models.
-
🔬 Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)
A genomic language model applies language-model training to DNA, with the longer-term goal of both understanding biological function and generating new functional sequences.
-
🔬 An Oscar, Two Asteroids, and the Algorithm in Your sklearn: John Platt on AI for Science
ERA turns scientific questions into scoreable coding tasks: Gemini writes and mutates notebooks, evaluates them, and searches for improvements using Monte Carlo tree search and upper-confidence-bound selection.
-
Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI
Jev is TypeSafe AI’s first “system one” or large programmable model, designed for code to consume directly rather than for human-facing chat.
-
⚡️ A Worm With 302 Neurons Inspired Their Architecture — Ramin Hasani, Liquid AI
Liquid’s original architecture research was inspired by modeling the continuous, non-spiking neurons of the 302-neuron worm C. elegans, whose small nervous system can control 95 muscle cells.
-
Underwriting Superintelligence: Backing Agents you can Sue — Rune Kvist, AIUC
The main bottleneck for AI adoption is trust, liability, and risk management, not capability.
-
Humanity’s Last Invention — Richard Socher of Recursive
AI has the potential to revolutionize scientific discovery and invention, especially through recursive self-improvement systems.
-
🔬 “We have foundation models for language, not for physics” — Anima Anandkumar, Bren Professor of Computing
AI can significantly improve weather and climate modeling by using neural operators that incorporate physical laws, enabling faster and more accurate predictions.
-
⏭️ Forward Deployed: Voice AI on what works in 2026
Production voice agents commonly use a cascaded speech-to-text, LLM, and text-to-speech pipeline because it provides more control, observability, and guardrails than current voice-to-voice systems.
-
Simulation: the new Scaling Law — Joon Sung Park, Simile AI
Simulation of human society and individual behavior is rapidly advancing, with current models achieving about 85% accuracy in replicating human responses.
-
⚡️ Exo: Harnesses should see their own code and logs — Alex Krentsel, UC Berekeley / Google Research
Not summarized yet
-
🔬 The BioAI Phase Shift - Matthew McPartlon & Neil Patil, Chai Discovery
Advances in AI-driven structure and binding prediction models are revolutionizing drug discovery, making the process faster and more deterministic.
-
The Inference Engineering Masterclass — Philip Kiely & Ali Taha, Baseten
Inference engineering has become a crucial discipline focused on optimizing the speed, reliability, and cost-effectiveness of large AI models in production.
-
Codex from 0 to 10M Users: Building ChatGPT Work — Akshay Nathan, OpenAI
AI and tools like ChatGPT and Codex are rapidly evolving, enabling new forms of productivity and collaboration.
-
Inside the Model Factory — Eiso Kant, Poolside AI
Scaling laws and continuous iteration are crucial for advancing language models effectively.
-
🔬 Causal Models Need Causal Data - Xaira’s X-Cell model for Drug Discovery (Bo Wang & Ci Chu, Chief Discovery Officer & Chief AI Scientist)
High-quality, causal data is crucial for training effective AI models in biology, especially for simulating cellular responses.
-
🔬 The Lab of the Future Should Feel Like a Data Center — Andy Beam & Rafa Gómez-Bombarelli, Lila Sciences
Lila Sciences envisions transforming laboratories into autonomous, AI-driven data centers capable of generating and reasoning over vast amounts of scientific data.
-
⚡️ Podcast Crossover: AIE, AGI, frontier lab strategy with @matthew_berman and @swyxtv
Not summarized yet
-
Why AI Infrastructure must evolve for Agent Experience — Akshat Bubna, Modal CTO
Modal originated as an optimized runtime platform for AI workloads, focusing on improving the developer experience and later transitioning towards a full AI cloud with primitives for inference, training, and agent deployment.
-
🔬 The Coolest Diffusion Research Isn't in LLMs — Evan Feinberg & Sergey Edunov, Genesis Molecular AI
Significant AI advances, especially diffusion models and high-precision structure prediction, are transforming drug discovery.
Follow this show on AudioBrevity and new episodes get summarized for you. Create a free account