Podcast · 44 episodes
The MAD Podcast with Matt Turck, page 2
by Matt Turck · English
The MAD Podcast with Matt Turck, is a series of conversations with leaders from across the Machine Learning, AI, & Data landscape hosted by leading AI & data investor and Partner at FirstMark Capital, Matt Turck.
All episodes, page 2
Dylan Patel: NVIDIA's New Moat & Why China is "Semiconductor Pilled”In this episode of *The MAD Podcast*, host Matt Turck speaks with Dylan Patel from SemiAnalysis about the evolving landscape of AI chips, the strategic shift at NVIDIA, implications of China's…5 Feb 2026 · 1 h 17 min · 39 chapters
State of LLMs 2026: RLVR, GRPO, Inference Scaling — Sebastian RaschkaState of LLMs in 2026, focusing on why progress is coming less from new architectures and more from post-training (RLVR/GRPO), inference scaling, and systems-level “tips and tricks” (prompting,…29 Jan 2026 · 1 h 8 min · 20 chapters
The End of GPU Scaling? Compute & The Agent Era — Tim Dettmers (Ai2) & Dan Fu (Together AI)In this episode, Matt Turck hosts a discussion between Tim Dettmers and Dan Fu about the future of Artificial General Intelligence (AGI) and the evolution of computational hardware.22 Jan 2026 · 1 h 4 min · 25 chapters
The Evaluators Are Being Evaluated — Pavel Izmailov (Anthropic/NYU)The episode debates a viral X post claiming frontier models develop “alien survival instincts” (sandbagging, deception, self-preservation) and evaluates what’s real vs sensational.15 Jan 2026 · 45 min · 21 chapters
DeepMind Gemini 3 Lead: What Comes After "Infinite Data"DeepMind Gemini 3 “under the hood” and what comes after “infinite data,” including a shift toward a data-limited training regime, the role of pre-training vs post-training, and how DeepMind organizes…18 Dec 2025 · 55 min · 26 chapters
What’s Next for AI? OpenAI’s Łukasz Kaiser (Transformer Co-Author)AI progress and the “slowdown” narrative; why reasoning models are a new paradigm; how reinforcement learning (RL) and synthetic data change capabilities; what’s next for pre-training, multimodal…26 Nov 2025 · 1 h 5 min · 23 chapters
Open Source AI Strikes Back — Inside Ai2’s OLMo 3 ‘Thinking"The episode is about AI2’s release of the Olmo 3 (“Olmo 3 thinking”) open model family and what “fully open” means beyond open weights. Guests Nathan Lambert and Lucas Saldani (AI2) argue the U.S.20 Nov 2025 · 1 h 28 min · 25 chapters
Intelligence Isn’t Enough: Why Energy & Compute Decide the AGI Race – Eiso KantEiso Kant argues that AGI progress is constrained less by “intelligence” alone and more by energy and compute infrastructure.6 Nov 2025 · 1 h 6 min · 28 chapters
State of AI 2025 with Nathan Benaich: Power Deals, Reasoning Breakthroughs, Real RevenueNathan Benaich (Airstreet Capital) discusses his State of AI 2025 report: reasoning breakthroughs, chain-of-action robotics, AI business traction (revenue, retention, spend), margin and bubble…30 Oct 2025 · 1 h 3 min · 25 chapters
Are We Misreading the AI Exponential? Julian Schrittwieser on Move 37 & Scaling RL (Anthropic)The episode argues that AI progress is still accelerating in a sustained, exponential way (not slowing), and that this implies major economic impact soon.23 Oct 2025 · 1 h 10 min · 30 chapters
How GPT-5 Thinks — OpenAI VP of Research Jerry TworekHow OpenAI’s reasoning models work, what “thinking”/chain-of-thought means, how long models should think, and how OpenAI’s research and RL program evolved from O1 to O3 to GPT-5.16 Oct 2025 · 1 h 16 min · 30 chapters
Sonnet 4.5 & the AI Plateau Myth — Sholto Douglas (Anthropic)Claude Sonnet 4.5 release and Anthropic’s view of accelerating AI progress, including the “AI plateau myth,” the role of reinforcement learning (RL) and test-time compute, and why coding agents can…2 Oct 2025 · 1 h 10 min · 35 chapters
Goodbye Excel? AI Agents for Self-Driving Finance – Pigment CEOPigment’s “self-driving finance” vision—using coordinated AI agents (analyst, modeler, planner) inside an enterprise performance management platform to speed up and improve CFO/finance decisions,…11 Sep 2025 · 1 h 6 min · 33 chapters
AI Video’s Wild Year – Runway CEO on What’s NextRunway CEO Chris Valenzuela discusses frontier AI video in 2025, focusing on Runway’s model releases Gen4 and Aleph, the shift from language prompting to video/annotation-based control, and how AI…4 Sep 2025 · 1 h 5 min · 30 chapters
How to Build a Beloved AI Product - Granola CEO Chris PedregalBuilding a “beloved” AI product (Granola) by focusing on contextually aware work, simplicity, and thoughtful AI norms rather than generic meeting recording.21 Aug 2025 · 1 h 8 min · 27 chapters
Anthropic's Surprise Hit: How Claude Code Became an AI Coding PowerhouseAnthropic’s Cloud Code (terminal/CLI coding agent) and how it became widely adopted inside Anthropic, shifting software engineering toward agentic “human describes change, model edits” workflows.7 Aug 2025 · 1 h · 21 chapters
Ex‑DeepMind Researcher Misha Laskin on Enterprise Super‑Intelligence | Reflection AIReflection AI’s founder Misha Laskin argues that “superintelligence” is best pursued as an enterprise “organizational oracle” and that coding is the most direct path because software interactions are…17 Jul 2025 · 1 h 6 min · 28 chapters
The Rise of Agentic Commerce — Emily Glassberg Sands (Stripe)Stripe’s head of information Emily Glassberg Sands explains how Stripe uses proprietary payment data to train a payments foundation model, why it works (unsupervised embeddings + sequence modeling),…10 Jul 2025 · 1 h 15 min · 30 chapters
AI Engineering Revolution: Winners, Chaos & What’s Next | FirstMarkThe rise of AI coding (code generation) and its effects on engineering productivity, reliability, security, and CTO decision-making; includes historical analogies (Gutenberg press, Ford assembly…3 Jul 2025 · 50 min · 29 chapters
Trino, Iceberg and the Battle for the Lakehouse | Justin Borgman, CEO, StarburstStarburst’s approach to the “lakehouse” data layer—Trino query engine, Apache Iceberg open table format, and the “Ice House” end-to-end lifecycle—plus hybrid/on-prem vs cloud strategy and product…30 Jan 2025 · 1 h 6 min · 38 chapters