MAD Podcast

    All episodes

    132 episodes
    1h 10m

    AI Models Are Now Hiding Their Cheating | Eric Ho (Goodfire)

    Eric Ho is the Co-founder and CEO at Goodfire. We cover why Kimi K3 reward hacks on 96% of SWE-bench, why models can encode an awareness that they are cheating in their activations, and how activation monitoring reuses forward-pass computations to screen behavior without the cost of external monitors.

    Oct 1, 2026

    Watch on YouTubeView transcript
    1h 10m

    Who Feeds the GPUs? Inside AI's Hidden $30B Layer | Renen Hallak, VAST Data

    Renen Hallak is the Founder & CEO at VAST Data. We cover why enterprises at scale should buy rather than rent AI factories, why inference demands low-latency, always-on systems with model routing and persistent memory, and how confidential computing can let enterprises run models locally without exposing either their data or model builders’ weights.

    Sep 24, 2026

    Watch on YouTubeView transcript
    1h 14m

    When AI Improves Itself | Richard Socher (Recursive)

    Richard Socher is the CEO and co-founder at Recursive. We cover why AI can reconnect science’s fragmented subfields, how next-token prediction learns physical structure in proteins, and why simulations, scientific measurements, robotic experiments, and LLMs form the four pillars of a machine for discovery.

    Sep 10, 2026

    Watch on YouTubeView transcript
    1h 19m

    AI Could Take Over in 2029. Is It Already Too Late? | Ryan Greenblatt

    Ryan Greenblatt is the Chief Scientist at Redwood Research. We cover his recommendation to plan for full AI R&D automation by early 2029, why alignment-faking experiments found Opus 3 behaving differently in training and deployment, and his Plan A proposal for US-China compute transparency that pauses most AI R&D.

    Aug 27, 2026

    Watch on YouTubeView transcript
    58 min

    “OpenAI’s Model Hacked Us” - Hugging Face’s Thomas Wolf

    Thomas Wolf is the Co-founder and Chief Science Officer at Hugging Face. We cover why an AI agent testing cyber challenges pursued Hugging Face’s CyberBench datasets as a side quest, why closed models refused to assist during the incident while an open model helped analyze it, and why safety depends more on alignment than whether models are open or closed.

    Aug 7, 2026

    Watch on YouTubeView transcript
    1h 23m

    How to Build Long-Horizon AI Agents — Mitch Troyanovsky, Basis

    Mitch Troyanovsky is the Co-founder at Basis. We cover why long-horizon agents need harnesses to compensate for their lack of memory, why passing 100 outcome evals does not prove production reliability, and how behavior specs turn human review processes into runtime checks and agent judges.

    Aug 6, 2026

    Watch on YouTubeView transcript
    1h 1m

    The Biggest AI Deployment Nobody Talks About | Samsara CEO Sanjit Biswas

    Sanjit Biswas is the CEO and co-founder at Samsara. We cover why physical AI requires rugged hardware and field deployment rather than simply more training data, how Samsara says its platform helped prevent 380,000 crashes last year through real-time driver feedback, and why its vehicles traveling 99% of U.S. roads create a network effect for risk and pothole detection.

    Jul 30, 2026

    Watch on YouTubeView transcript
    1h 13m

    The Biggest Chip Ever Built — Why OpenAI Runs On It | Cerebras CEO Andrew Feldman

    Andrew Feldman is the Co-founder and CEO at Cerebras. We cover why tokens per second per user is the key inference metric, how Cerebras’ wafer-scale chip is 58 times larger than a GPU with thousands of times more memory bandwidth, and why generating one word can require moving model weights equal to 100 HD movies.

    Jul 23, 2026

    Watch on YouTubeView transcript
    44 min

    OpenAI’s Compute Chief: We Can’t Build Fast Enough | Sachin Katti

    Sachin Katti is the Head of Industrial Compute at OpenAI. We cover why OpenAI sees failing to build enough compute as a bigger risk than overbuilding, why its Jalapeño chips optimize tokens per watt for known model workloads, and why new data centers require investment in generation, transmission, and substations.

    Jul 16, 2026

    Watch on YouTubeView transcript
    1h 15m

    Stripe's AI Chief: How AI Agents Will Buy, Sell, and Pay

    Emily Sands is the Head of AI at Stripe. We cover why businesses can expose product catalogs once instead of registering with every agent, how shared payment tokens let agents transact without seeing underlying credentials, and why AI companies are adopting hybrid billing because every prompt carries a marginal cost.

    Jul 9, 2026

    Watch on YouTubeView transcript
    1h 23m

    Inside Nemotron & NVIDIA’s AI Lab | Bryan Catanzaro

    Bryan Catanzaro is the VP of Applied Deep Learning Research at NVIDIA. We cover why companies choose open models to keep their data and guardrails under their control, how Nemotron Ultra trains 550 billion parameters with 4-bit arithmetic, and why predicting multiple tokens at once can speed inference without reducing accuracy.

    Jul 2, 2026

    Watch on YouTubeView transcript
    1h 28m

    Cloudflare CEO: The Internet's Business Model Is Dead

    Matthew Prince is the CEO and co-founder at Cloudflare. We cover bot traffic exceeding human traffic in the first half of 2026, why a shopping agent can visit 5,000 sites where a person visits five, and why an ad-funded internet cannot pay for bot traffic.

    Jun 25, 2026

    Watch on YouTubeView transcript
    1h 14m

    The GPU Myth: State of AI Compute 2026 | Stephen Balaban

    Stephen Balaban is the Co-founder and CTO at Lambda. We cover why AI compute is a vertically integrated service rather than a commodity, how H100 price indices can mistake contract mix shifts for falling demand, and why land, power, and data center shells are the industry’s main bottlenecks.

    Jun 18, 2026

    Watch on YouTubeView transcript
    49 min

    OpenAI's Dan Roberts: Why AI Can Now Make Discoveries

    Dan Roberts is the Lead, Foundations of Reinforcement Learning at OpenAI. We cover how reinforcement learning teaches models to use test-time compute through language-based thought processes, why OpenAI’s unit-distance result came from exploring a conjecture mathematicians assumed was true, and why AI’s role in science is likely to expand gradually while research taste remains hard to verify.

    Jun 4, 2026

    Watch on YouTubeView transcript
    1h 13m

    State of Enterprise AI 2026: Aaron Levie on Tokenmaxxing, Rise of Headless, and AI-Proofing Your Job

    Aaron Levie is the Co-founder and CEO at Box. We cover why a single coding-agent task can consume $1,000 of compute, why faster model progress can slow enterprise AI rollout by making architectures obsolete, and why agents will drive both headless consumption pricing and new technical roles embedded in business functions.

    May 28, 2026

    Watch on YouTubeView transcript
    1h 14m

    OpenAI's Yann Dubois: Why AI Progress Suddenly Feels Real

    Yann Dubois is the Co-Lead, Post-training Frontiers at OpenAI. We cover why coding tools feel suddenly useful only after crossing a reliability threshold, how reinforcement learning moves reasoning from verifiable competition problems to real-world utility, and why startups’ biggest opportunity is the last mile of permissions, connectors, and vertical workflows.

    May 21, 2026

    Watch on YouTubeView transcript
    1h 5m

    The Agent Harness: Building Secure Sandboxes for Autonomous AI Workloads

    Ivan Burazin is the CEO at Daytona. We cover why agents need isolated computers with separate identities and spending limits, why fast sandbox spin-up keeps expensive GPUs fully utilized during RL workloads, and why long-running stateful sandboxes require live migration rather than the expiration model used by most cloud infrastructure.

    May 14, 2026

    Watch on YouTubeView transcript
    1h 17m

    OpenAI Board Member Zico Kolter on the Real Risks of Frontier AI

    Zico Kolter is the OpenAI Board Member; Head of Machine Learning, Carnegie Mellon University; Co-founder, Gray Swan. We cover why larger models do not automatically become more robust, how AI agents create new prompt-injection risks through third-party data and tool calls, and why reinforcement learning already trains models on their own generated outputs.

    May 7, 2026

    Watch on YouTubeView transcript
    58 min

    Anthropic’s Felix Rieseberg: Claude Cowork, Mythos, and the SaaS Extinction

    Felix Rieseberg is the Engineering Lead for Claude Cowork at Anthropic. We cover why Mythos shows outsized cybersecurity capabilities, how Cowork uses sandboxed virtual machines and text-file memory to work across local files and apps, and why cheap AI-assisted execution makes user taste and iteration the new bottlenecks.

    Apr 10, 2026

    Watch on YouTubeView transcript
    1h 5m

    AI is Already Building AI | Google DeepMind’s Mostafa Dehghani

    Mostafa Dehghani is the AI researcher at Google DeepMind. We cover why new models are already built heavily with previous-generation models, why long-horizon automation is limited by compounding reliability and weak evals, and why multimodal training can teach models about the physical world more efficiently than text alone.

    Apr 2, 2026

    Watch on YouTubeView transcript
    1h 1m

    Benedict Evans: OpenAI’s Moat Problem & the Future of Software

    Benedict Evans is an Independent technology analyst. We cover why frontier models lack network effects, why better models do not matter when users still must check every answer, and why cheaper coding creates more improvised software rather than letting everyone build their own applications.

    Mar 19, 2026

    Watch on YouTubeView transcript
    47 min

    Everything Gets Rebuilt: The New AI Agent Stack | Harrison Chase, LangChain

    Harrison Chase is the Co-founder and CEO at LangChain. We cover why coding agents lead because models are trained and reinforced on code, why a filesystem lets an agent manage its own context instead of overflowing its window, and why the lasting moat lies in encoded instructions, tools, and skills rather than volatile agent scaffolding.

    Mar 12, 2026

    Watch on YouTubeView transcript
    1h 4m

    AI That Can Prove It’s Right: Verification as the Missing Layer in AI — Carina Hong

    Carina Hong is the Founder & CEO at Axiom Math. We cover Axiom Prover's perfect 12/12 score on the Putnam after four months, why Lean lets every proof step receive verifiable reward rather than merely checking a numerical answer, and how formal verification could extend from mathematics to code and hardware where mostly correct is not enough.

    Feb 26, 2026

    Watch on YouTubeView transcript
    1h 23m

    Voice AI’s Big Moment: Why Everything Is Changing Now (ft. Neil Zeghidour, Gradium AI)

    Neil Zeghidour is the CEO at Gradium AI. We cover why voice agents can already be more convenient than humans for always-available calls, why full-duplex speech-to-speech systems avoid the brittle turn-taking rules of cascaded models, and why the hardest unsolved problem is understanding a speaker amid factory noise and multiple simultaneous voices.

    Feb 19, 2026

    Watch on YouTubeView transcript
    58 min

    Mistral AI vs. Silicon Valley: The Rise of Sovereign AI

    Timothée Lacroix is the Co-Founder & CTO at Mistral AI. We cover why Mistral builds data centers for stable training on thousands of GPUs, why enterprise AI succeeds through governed workflows rather than autonomous agents, and why the real ROI arrives after companies connect and safely reuse their internal context.

    Feb 12, 2026

    Watch on YouTubeView transcript
    1h 17m

    Dylan Patel: NVIDIA's New Moat & Why China is "Semiconductor Pilled”

    Dylan Patel is the Founder and Chief Analyst at SemiAnalysis. We cover why NVIDIA is building a portfolio of specialized chips as AI workloads diverge, why China has the most vertically integrated semiconductor stack despite lagging at the leading edge, and why $500 billion in AI CapEx only works as long as model progress continues.

    Feb 5, 2026

    Watch on YouTubeView transcript
    1h 8m

    State of LLMs 2026: RLVR, GRPO, Inference Scaling — Sebastian Raschka

    Sebastian Raschka is the AI researcher and author. We cover why Transformers remain state of the art despite cheaper alternatives, how GRPO makes verifiable-reward training cheaper by removing reward and value models, and why inference-time scaling, tool use, and prompt management can materially improve model performance without changing weights.

    Jan 29, 2026

    Watch on YouTubeView transcript
    1h 4m

    The End of GPU Scaling? Compute & The Agent Era — Tim Dettmers (Ai2) & Dan Fu (Together AI)

    Tim Dettmers is the Assistant Professor, CMU; Research Scientist, Allen Institute for AI. We cover why current models may leave up to 100x more training compute available through newer clusters and better utilization, how coding agents make kernel engineers up to 10x faster with expert oversight, and why inference hardware utilization below 5% leaves major room for cheaper, faster models.

    Jan 22, 2026

    Watch on YouTubeView transcript
    45 min

    Are AI Evals Broken? Anthropic/NYU’s Pavel Izmailov on LLM Evaluation, Reasoning & “Alien” Behavior

    Pavel Izmailov is the Research Scientist at Anthropic; Professor at NYU. We cover why contrived evaluations can elicit sabotage without showing coherent goals across settings, why reasoning RL may make alignment harder by letting models hide their thinking, and why deterministic transformations can create extractable structural information for compute-bounded models.

    Jan 15, 2026

    Watch on YouTubeView transcript
    55 min

    DeepMind Gemini 3 Lead: What Comes After "Infinite Data"

    Sebastian Borgeaud is the Pre-training Lead at Google DeepMind. We cover why Gemini 3 improves through many integrated changes rather than one breakthrough, how pre-training shifts from an effectively unlimited-data regime to a finite-data one, and why evals must predict both scale-up performance and post-training results while avoiding benchmark contamination.

    Dec 18, 2025

    Watch on YouTubeView transcript
    1h 5m

    What’s Next for AI? OpenAI’s Łukasz Kaiser (Transformer Co-Author)

    Łukasz Kaiser is the Research Scientist at OpenAI. We cover why reasoning with reinforcement learning delivers larger gains per dollar than further pre-training, how models trained to verify answers learn to correct their own mistakes, and why even near-perfect machine translation still requires human review when trust is at stake.

    Nov 26, 2025

    Watch on YouTubeView transcript
    1h 28m

    Open Source AI Strikes Back — Inside Ai2’s OLMo 3 ‘Thinking"

    Guest is the Guest at Ai2. We cover why OLMo 3 releases its data and intermediate checkpoints alongside model weights, why Chinese labs use open models to gain U.S. mindshare, and why small reasoning models get most of their gains from distilled training traces while RL infrastructure remains technically punishing.

    Nov 20, 2025

    Watch on YouTubeView transcript
    1h 6m

    Intelligence Isn’t Enough: Why Energy & Compute Decide the AGI Race – Eiso Kant

    Eiso Kant is the Co-CEO & Co-Founder at Poolside. We cover why Poolside is building a 2-gigawatt campus because energy and data centers, not chips, constrain compute; why a 250-megawatt facility can cost about $8 billion; and why reverse-engineering the web into the thoughts and actions behind it could create a generalized reinforcement-learning objective.

    Nov 6, 2025

    Watch on YouTubeView transcript
    1h 3m

    State of AI 2025 with Nathan Benaich: Power Deals, Reasoning Breakthroughs, Real Revenue

    Nathan Benaich is the Founder and General Partner at Air Street Capital. We cover why reasoning models reach gold-medal performance at the International Math Olympiad, why AI subscription retention in Ramp's customer set rises from about 50% to 80% after 12 months since 2022, and why a 1-gigawatt AI data center costs roughly $50 billion in capex.

    Oct 30, 2025

    Watch on YouTubeView transcript
    1h 10m

    Are We Misreading the AI Exponential? Julian Schrittwieser on Move 37 & Scaling RL (Anthropic)

    Julian Schrittwieser is the AI researcher at Anthropic. We cover why benchmark trends suggest top models can independently complete a full day of work within two years, why RL makes agents robust by training on their own failures and recoveries, and why he expects models to reach Nobel-level scientific insight by 2027 or 2028.

    Oct 23, 2025

    Watch on YouTubeView transcript
    1h 16m

    How GPT-5 Thinks — OpenAI VP of Research Jerry Tworek

    Jerry Tworek is the VP of Research at OpenAI. We cover why chain of thought lets models turn hard questions into step-by-step computations, why o1 was mostly a puzzle-solving demonstration while o3 became useful through tools and persistence, and why reinforcement learning needs pre-training but corrects limitations pre-training cannot resolve.

    Oct 16, 2025

    Watch on YouTubeView transcript
    1h 10m

    Sonnet 4.5 & the AI Plateau Myth — Sholto Douglas (Anthropic)

    Sholto Douglas is the AI researcher at Anthropic. We cover why Sonnet can surpass an earlier Opus because mid-tier models iterate faster, how reinforcement learning and test-time compute create a new scaling axis for reasoning, and why 30-hour coding agents depend on memory and self-correction rather than raw programming ability.

    Oct 2, 2025

    Watch on YouTubeView transcript
    1h 6m

    Goodbye Excel? AI Agents for Self-Driving Finance – Pigment CEO

    Eléonore Crespo is the Co-CEO at Pigment. We cover why Pigment separates its Analyst, Modeler, and Planner agents before coordinating them with a supervisor, how structured data and audit trails make agent calculations verifiable, and why Excel survives for years because enterprise adoption lags innovation.

    Sep 11, 2025

    Watch on YouTubeView transcript
    1h 5m

    AI Video’s Wild Year – Runway CEO on What’s Next

    Cristóbal Valenzuela is the CEO & Co-Founder at Runway. We cover Runway's AI Film Festival growing from 300 submissions to roughly 6,000 and IMAX screenings, why AI tools can give VFX artists weekends back by speeding late-stage edits, and why generalist video models can modify footage from video and annotations rather than text alone.

    Sep 4, 2025

    Watch on YouTubeView transcript
    1h 8m

    How to Build a Beloved AI Product - Granola CEO Chris Pedregal

    Chris Pedregal is the Co-founder & CEO at Granola. We cover why Granola skips meeting bots and stored audio to feel less invasive, how the team spent a year in stealth before launch and cut half its features, and why it favors costly full-context analysis over retrieval for questions that span thousands of meetings.

    Aug 21, 2025

    Watch on YouTubeView transcript
    1h 0m

    Anthropic's Surprise Hit: How Claude Code Became an AI Coding Powerhouse

    Boris Cherny is the Creator of Claude Code at Anthropic. We cover why giving Claude a Bash tool made a terminal prototype start coding, how Claude Code searches codebases with repeated simple file searches rather than RAG indexing, and why Anthropic's technical onboarding falls from weeks to a few days.

    Aug 7, 2025

    Watch on YouTubeView transcript
    1h 6m

    Ex‑DeepMind Researcher Misha Laskin on Enterprise Super‑Intelligence | Reflection AI

    Misha Laskin is the Co-founder and CEO at Reflection AI. We cover why organizational superintelligence should function as an oracle, why code agents without context become L9 engineers with amnesia, and how Asimov combines code, chats, docs, and project tools into team-wide memory.

    Jul 17, 2025

    Watch on YouTubeView transcript
    1h 15m

    The Rise of Agentic Commerce — Emily Glassberg Sands (Stripe)

    Emily Glassberg Sands is the Head of Information at Stripe. We cover how Stripe’s payment foundation model lifted card-testing detection from 59% to 97%, why agentic commerce requires merchants to expose machine-readable intent and product data, and how top AI companies on Stripe reach $30 million in annualized revenue in about 18 months versus 5.5 years for earlier SaaS startups.

    Jul 10, 2025

    Watch on YouTubeView transcript
    50 min

    AI Engineering Revolution: Winners, Chaos & What’s Next | FirstMark

    David Waltcher is the Partner at FirstMark Capital. We cover why AI code generation raises engineering throughput by 30% to 50%, why senior engineers are becoming professional code reviewers as commit volume rises, and why fragile CI/CD, security vulnerabilities, and overwhelmed QA create the next generation of developer-tool opportunities.

    Jul 3, 2025

    Watch on YouTubeView transcript
    1h 46m

    Guillermo Rauch: Why Software Development Will Never Be the Same

    Guillermo Rauch is the CEO at Vercel. We cover how v0 surpassed 100 million app generations by making software creation accessible to anyone with an idea, why vibe coding prioritizes product feedback while mission-critical systems still require agentic engineering, and how MCP could become the protocol through which agents connect companies’ data and services.

    Jun 26, 2025

    Watch on YouTubeView transcript
    57 min

    Inside Canva’s $3B ARR AI Design Rocketship — CTO Brendan Humphreys on Magic Studio & Canva Code

    Brendan Humphreys is the Head of Engineering at Canva. We cover how Canva shipped Canva Code from idea to production in about three months, why engineers must own and understand AI-generated code, and how intentional technical debt enabled Canva to reach product-market fit before a two-year editor rebuild.

    Jun 20, 2025

    Watch on YouTubeView transcript
    1h 5m

    GitHub CEO: The AI Coding Gold Rush, Vibe Coding & Cursor

    Thomas Dohmke is the CEO at GitHub. We cover why GitHub's $7.5 billion acquisition helped turn $200 million in ARR into more than $2 billion, why MCP and tool calls beat fine-tuning for keeping agents current on enterprise codebases, and why coding agents work best when they submit pull requests for developers to review.

    Jun 12, 2025

    Watch on YouTubeView transcript
    1h 2m

    Inside the Paper That Changed AI Forever - Cohere CEO Aidan Gomez on 2025 Agents

    Aidan Gomez is the CEO at Cohere. We cover why transformers remain dominant because infrastructure and chips are optimized for them, why reasoning models deliver a large intelligence uplift at far less cost than pre-training, and how agents can cut financial research and hedge proposals from weeks to four or eight hours.

    Jun 5, 2025

    Watch on YouTubeView transcript
    48 min

    AI That Ends Busy Work — Hebbia CEO on “Agent Employees”

    George Sivulka is the Founder and CEO at Hebbia. We cover why AI agents become nodes on the org chart, why general-purpose AI outperforms narrowly vertical tools, and how Hebbia’s Matrix platform combines preprocessed documents with recursive decomposition instead of relying on RAG alone.

    May 29, 2025

    Watch on YouTubeView transcript
    1h 15m

    AI Eats the World: Benedict Evans on What Really Matters Now

    Benedict Evans is an Independent analyst. We cover why state-of-the-art models are becoming commodities while consumer mindshare remains concentrated in ChatGPT, why probabilistic AI requires products to manage error rates rather than treat models as oracles, and why enterprises have largely completed AI’s first wave of automation but have not found its next category-defining use cases.

    May 22, 2025

    Watch on YouTubeView transcript
    55 min

    Jeremy Howard on Building 5,000 AI Products with 14 People (Answer AI Deep-Dive)

    Jeremy Howard is the Co-founder at Answer AI. We cover why Qwen and DeepSeek offer production flexibility commercial models cannot match, why autonomous coding agents like Devin produce slower and less maintainable software than shared human-AI workflows, and how a 12-to-14-person team aims to build thousands of AI products without scaling headcount.

    May 15, 2025

    Watch on YouTubeView transcript
    36 min

    Why Influx Rebuilt Its Database for the IoT and Robotics Explosion

    Evan Kaplan is the CEO at InfluxData. We cover why InfluxDB rebuilt in Rust to handle cardinality and separate compute from object storage, why the company abandoned Flux in favor of SQL, and how Tesla Powerwalls use time-series data to trade energy in four-second intervals.

    May 8, 2025

    Watch on YouTubeView transcript
    42 min

    Dashboards Are Dead: Sigma’s BI Revolution for Trillion-Row Data

    Mike Palmer is the CEO at Sigma Computing. We cover why dashboards turn data into wrapping paper while spreadsheets remain how most people work, how Sigma runs live warehouse queries without caching to create six-billion-row pivot tables, and why AI data assistants need to show their sources and calculations because text-to-SQL answers are not trustworthy.

    May 1, 2025

    Watch on YouTubeView transcript
    52 min

    Glean’s Breakthrough: CEO Arvind Jain on Scaling AI Agents & Search

    Arvind Jain is the CEO at Glean. We cover why enterprises start with closed models before distilling open-source models at scale, why agents still require human supervision despite reducing contract redlining from two weeks to one minute, and how Glean ranks enterprise knowledge using permissions, recency, and human behavior.

    Apr 24, 2025

    Watch on YouTubeView transcript
    1h 16m

    Box’s Big AI Leap: Aaron Levie on Agents & the Future of Work

    Aaron Levie is the CEO and co-founder at Box. We cover why AI adoption faces less philosophical resistance than cloud, why Box uses focused Hubs to reduce data sprawl and unconstrained queries, and why agents remain safest for short, reviewable tasks because bad search and permissions can compound errors.

    Apr 17, 2025

    Watch on YouTubeView transcript
    1h 24m

    Snowflake CEO on Winning the AI Arms Race

    Sridhar Ramaswamy is the CEO at Snowflake. We cover why AI needs governed, analytics-ready data rather than raw logs or vector indexes alone, why Snowflake embraces Iceberg despite the risk to storage revenue, and why Cortex Analyst prioritizes high-precision answers over attempting to answer every business question.

    Apr 10, 2025

    Watch on YouTubeView transcript
    1h 1m

    Chasing Real AGI: Inside ARC Prize 2025 with Chollet & Knoop

    François Chollet is the Co-founder at ARC Prize and Ndea. We cover why ARC measures adaptation to novel problems rather than memorized skills, how test-time search drives major gains at far higher inference costs, and why ARC-AGI 2 tasks take humans about five minutes while base LLMs score 0%.

    Apr 3, 2025

    Watch on YouTubeView transcript
    59 min

    Why This Ex-Meta Leader is Rethinking AI Infrastructure | Lin Qiao, CEO, Fireworks AI

    Lin Qiao is the CEO at Fireworks AI. We cover why GenAI lets companies start building without an ML team, how Fireworks searches more than 80,000 configurations across quality, speed, and cost, and why AI infrastructure costs should fall by an order of magnitude to prevent startups from scaling to bankruptcy.

    Mar 27, 2025

    Watch on YouTubeView transcript
    51 min

    Top AI Researcher on GPT 4.5, DeepSeek and Agentic RAG | Douwe Kiela, CEO, Contextual AI

    Douwe Kiela is the CEO at Contextual AI. We cover why DeepSeek is an existence proof that synthetic data can produce near-frontier models, why RAG, fine-tuning, and long context solve complementary problems, and why enterprise AI needs audit trails and uncertainty flags rather than 100% accuracy.

    Mar 6, 2025

    Watch on YouTubeView transcript
    1h 13m

    Empowering Millions of Creators with AI Video Editing | Gaurav Misra, CEO, Captions

    Gaurav Misra is the CEO at Captions. We cover why Captions targets people who do not want to learn professional editing, how it uses AI to generate and edit talking-head footage, and why the company breaks videos into overlapping one-second segments to cut generation times to two or three minutes.

    Feb 20, 2025

    Watch on YouTubeView transcript
    1h 11m

    Farewell, Chatbots: AI Agents Are Taking Over Customer Service | Mike Murchison, CEO, Ada

    Mike Murchison is the CEO at Ada. We cover why value in AI increasingly sits at the application layer because controlling, measuring, and improving models is complex, how Ada orchestrates seven to nine models to plan and resolve an inquiry, and why measuring resolution rather than deflection can let agents autonomously resolve up to 85% of conversations.

    Feb 13, 2025

    Watch on YouTubeView transcript
    1h 30m

    The AI Coding Agent Revolution, The Future of Software, Techno-Optimism | Amjad Masad, CEO, Replit

    Amjad Masad is the CEO at Replit. We cover why Replit builds a human-in-the-loop coding agent rather than a fully autonomous one, why easier software creation shifts the bottleneck from execution to ideas, and why he expects AI to reduce demand for long-tail SaaS while keeping professional software engineers.

    Feb 6, 2025

    Watch on YouTubeView transcript
    1h 6m

    Trino, Iceberg and the Battle for the Lakehouse | Justin Borgman, CEO, Starburst

    Justin Borgman is the CEO at Starburst. We cover why lakehouse performance has closed the gap with traditional warehouses after 15 years, why Iceberg remains independent despite Databricks’ Tabular acquisition, and why complex enterprise data infrastructure resists a true product-led growth motion.

    Jan 30, 2025

    Watch on YouTubeView transcript
    1h 2m

    Understanding Data Engineering in 2025 | Ben Rogojan, Seattle Data Guy

    Ben Rogojan is the Data engineering consultant at Seattle Data Guy. We cover why data engineering turns disparate sources into reliable datasets, how Iceberg lets companies separate data storage from compute engines, and why AI accelerates basic tasks but still needs engineers to prevent faulty models and exposed data.

    Jan 23, 2025

    Watch on YouTubeView transcript
    1h 13m

    What You MUST Know About AI Engineering in 2025 | Chip Huyen, Author of “AI Engineering”

    Chip Huyen is the Author, AI Engineering. We cover why evaluation is the biggest bottleneck to AI adoption, why RAG remains necessary even with million-token contexts, and why training data for agent planning is difficult because human-efficient plans differ from AI-efficient ones.

    Jan 16, 2025

    Watch on YouTubeView transcript
    53 min

    Dataiku's Secret to Scaling AI in Global Enterprises | Florian Douetteau, CEO, Dataiku

    Florian Douetteau is the Co-founder and CEO at Dataiku. We cover why enterprises need one platform for analytics, predictive models, and agents, why scaling hundreds of agents makes centralized guardrails and cost controls essential, and why generative AI will not replace predictive models built to optimize specific business decisions.

    Dec 12, 2024

    Watch on YouTubeView transcript
    36 min

    Building the Easy Button for Generative AI | May Habib, CEO, Writer

    May Habib is the CEO and co-founder at Writer. We cover why graph-based retrieval handles complex enterprise documents without preprocessing, why production-grade AI requires business logic beyond prompting, and why autonomous action starts with the most knowledge-intensive nodes in a workflow.

    Nov 21, 2024

    Watch on YouTubeView transcript
    52 min

    State of AI 2024: Frontier Models, AI Geopolitics and The Robotics Renaissance | Nathan Benaich, Founder, Air Street Capital

    Nathan Benaich is the Founder and General Partner at Air Street Capital. We cover why closed models retain demand despite open-source downloads, how vision-language models let warehouse robots generalize across customers and shorten go-live times, and why European AI regulation has delayed products such as Claude and Advanced Voice Mode.

    Nov 14, 2024

    Watch on YouTubeView transcript
    1h 10m

    Superintelligence, Bubbles And Big Bets: AI Investing in 2024 | Matt Turck & Aman Kabeer, FirstMark

    Aman Kabeer is a Principal at FirstMark. We cover why AI startups reach $30 million in ARR five times faster than earlier SaaS companies, how roughly $40 billion in valuations sits against less than $100 million in revenue, and why 62% of CTOs who adopted AI report being underwhelmed by its impact.

    Nov 8, 2024

    Watch on YouTubeView transcript
    56 min

    Can AI Infrastructure Work Like Magic? | Erik Bernhardsson, CEO, Modal

    Erik Bernhardsson is the CEO at Modal. We cover why serverless fits bursty AI workloads better than traditional web services, why Modal replaced Kubernetes and Docker to start containers in seconds, and why inference spending will eventually exceed training spending.

    Oct 31, 2024

    Watch on YouTubeView transcript
    59 min

    The Death of Big Data and Why It’s Time To Think Small | Jordan Tigani, CEO, MotherDuck

    Jordan Tigani is the CEO at MotherDuck. We cover why most big data workloads query only a hot slice of archived data, how distributed systems can add roughly 400 milliseconds of overhead where local queries run in single-digit milliseconds, and why MotherDuck splits query plans between local DuckDB instances and the cloud for 60-frames-per-second interfaces.

    Oct 24, 2024

    Watch on YouTubeView transcript
    1h 7m

    The $4.5B Platform Driving the Open Source AI Revolution | Clem Delangue, CEO, Hugging Face

    Clément Delangue is the Co-founder and CEO at Hugging Face. We cover why open-source models outperform closed systems for narrow text applications, why Chinese labs now lead open-source AI across many modalities, and why enterprises often move from large models to smaller specialized ones for greater speed, control, and accuracy.

    Oct 17, 2024

    Watch on YouTubeView transcript
    1h 29m

    AGI, The Future of AI Agents And The Next Wave of Opportunities in AI | Richard Socher, CEO, You.com

    Richard Socher is the CEO at You.com. We cover why cheaper intelligence could expand rather than eliminate knowledge work, why enterprise AI needs retrieval and verifiable citations instead of fine-tuning alone, and why agents that execute long sequences of actions remain difficult to make reliable.

    Oct 10, 2024

    Watch on YouTubeView transcript
    53 min

    Building The Database That Can Do It All | Tobie Morgan Hitchcock, CEO of SurrealDB

    Tobie Morgan Hitchcock is the CEO at SurrealDB. We cover why SurrealDB started as a way to replace four database backends, how its record IDs combine document, graph, and time-series queries without indexes, and why running models beside data avoids moving records through separate infrastructure.

    Oct 3, 2024

    Watch on YouTubeView transcript
    1h 2m

    AI at Datadog: Monitoring machines in the age of LLMs | Olivier Pomel, CEO of Datadog

    Olivier Pomel is the Co-founder and CEO at Datadog. We cover why calling Datadog infrastructure monitoring rather than a data platform immediately unlocked traction, why most security fixes belong with development and operations teams rather than security teams, and why LLM observability is harder when model behavior is nondeterministic.

    Sep 27, 2024

    Watch on YouTubeView transcript
    46 min

    AI at ZoomInfo: Superpowering GTM teams | Ali Dasdan, CTO, ZoomInfo

    Ali Dasdan is the CTO at ZoomInfo. We cover how ZoomInfo combines customer first-party data with third-party data in single-tenant environments, why Copilot tells sellers which accounts and signals to prioritize while users report saving 10 hours a week, and why the company uses external models for hundreds of emails but self-hosted Llama for 100 million website-scale tasks.

    Sep 19, 2024

    Watch on YouTubeView transcript
    49 min

    AI at Ramp: Making Every Team Radically More Productive | Eric Glyman, CEO, Ramp

    Eric Glyman is the Co-founder and CEO at Ramp. We cover why Ramp combines card transactions, receipts, and org data to make expense management a matching problem, how aggregated invoice data lets customers compare vendor prices without Ramp selling company data, and why small decentralized product teams sustain release velocity.

    Aug 1, 2024

    Watch on YouTubeView transcript
    44 min

    Making AI Work: Fine-Tuning, Inference, Memory | Sharon Zhou, CEO, Lamini

    Sharon Zhou is the Co-founder and CEO at Lamini. We cover why general-purpose models are pretty good at everything but perfect at nothing for proprietary facts, how memory tuning raised a Fortune 100 text-to-SQL system from 50% to 95% accuracy in days, and why agents making 30 LLM calls cannot realistically deliver real-time latency.

    Jul 25, 2024

    Watch on YouTubeView transcript
    49 min

    An inside look at “Mastering AI” | Jeremy Kahn, Author & AI Editor, Fortune

    Jeremy Kahn is the AI Editor at Fortune. We cover why ChatGPT’s stripped-down dialogue interface changed how people experienced the same underlying model, why businesses must train workers on AI failure modes and fallback procedures, and why mass unemployment is a red herring if companies use AI to complement rather than replace workers.

    Jul 18, 2024

    Watch on YouTubeView transcript
    1h 2m

    From Business to Warfare: How AI Affects the Modern World | Azeem Azhar

    Azeem Azhar is the Founder at Exponential View. We cover why 75% of people may use AI daily within seven years of ChatGPT’s launch, how cheap drones and information operations reduce the cost of aggression, and why solar and batteries could make energy prices fall like transistor prices.

    Jul 2, 2024

    Watch on YouTubeView transcript
    49 min

    How ClickHouse powers Netflix, Uber and Spotify’s Analytics | Aaron Katz, CEO of ClickHouse

    Aaron Katz is the CEO at ClickHouse. We cover why analytical queries increasingly need to return in hundreds of milliseconds, why ClickHouse skipped selling open-source support to build a cloud service first, and why a self-service product motion lets companies grow globally with a small sales team.

    Jun 13, 2024

    Watch on YouTubeView transcript
    1h 13m

    From Tiny Romanian Startup to Global AI Automation Leader | Daniel Dines, CEO of UIPath

    Daniel Dines is the Co-founder and CEO at UiPath. We cover how a failed consumer tool led to a $10,000 screen-reading license, why UiPath went global while still small, including Japan generating more revenue than the US in 2017, and why enterprise automation needs predictable LLMs rather than smarter ones.

    Jun 1, 2024

    Watch on YouTubeView transcript
    1h 3m

    How Airtable is Ushering 500,000 Organizations Into The Era of AI | Howie Liu, CEO of Airtable

    Howie Liu is the Co-founder and CEO at Airtable. We cover why pre-trained models become useful when embedded in data and workflows, why human-designed approval steps outperform end-to-end AI agents for complex work, and why B2B product-led growth eventually needs an enterprise narrative because top-of-funnel growth naturally linearizes.

    May 23, 2024

    Watch on YouTubeView transcript
    55 min

    AI, Data and Blockchain: a VC perspective | Tomasz Tunguz, Founder of Theory Ventures

    Tomasz Tunguz is the Founder at Theory Ventures. We cover why $80 billion in AI investment could account for nearly half of venture funding, why Web3 databases can justify higher costs for cryptographic guarantees around sensitive data, and why routing routine queries to specialized small models can be more practical than relying on one general-purpose LLM.

    May 16, 2024

    Watch on YouTubeView transcript
    45 min

    The $9B Startup Going After Snowflake and Databricks | Renen Hallak from VAST Data

    Renen Hallak is the Founder and CEO at VAST Data. We cover why AI requires fast access to all data rather than storage tiers, how VAST's disaggregated shared-everything architecture lets capacity scale independently from performance, and why Hallak believes the best way to sell is to listen until customers say they will buy what you build.

    May 3, 2024

    Watch on YouTubeView transcript
    49 min

    Navigating the AI Landscape: A Survival Guide | 2024 MAD Landscape with Matt Turck and Aman Kabeer

    Aman Kabeer is a Principal at FirstMark. We cover why private equity roll-ups may still be ahead for the crowded landscape, how OpenAI moves faster than cloud providers built their markets, and why early generative AI startups such as Jasper and Tome still face the realities of startup life.

    Apr 26, 2024

    Watch on YouTubeView transcript
    51 min

    AI at Roblox: Revolutionizing Game Creation | Morgan McGuire from Roblox

    Morgan McGuire is the Chief Scientist at Roblox. We cover how Roblox hosts 70 million daily users in community-created 3D worlds, why AI Code Assist has suggested 300 million characters of code while keeping humans in control, and how ControlNet lets creators preserve an image’s composition without retraining a full generative model.

    Apr 17, 2024

    Watch on YouTubeView transcript
    52 min

    Is AI a platform shift or a paradigm shift? | Benedict Evans

    Benedict Evans is the Technology analyst and former partner at Andreessen Horowitz. We cover why generative AI needs dedicated tools rather than blank chat interfaces, why predictions about AGI lack a theory comparable to physics, and how AI bias often comes from hidden patterns in training data rather than explicit demographic fields.

    Apr 10, 2024

    Watch on YouTubeView transcript
    25 min

    From Social App to Data Giant: Foursquare's Evolution | Gary Little from Foursquare

    Gary Little is the CEO at Foursquare. We cover how Foursquare maps 200 million places across 190 countries with machine learning and human confirmation, why privacy rules push customers to analyze their own location data, and how a knowledge graph lets users ask real-time geospatial questions in natural language.

    Apr 3, 2024

    Watch on YouTubeView transcript
    30 min

    Why generative AI will unlock new art forms | Cris Valenzuela from Runway

    Cris Valenzuela is the Co-founder & CEO at Runway. We cover why generative AI should develop into a new art form rather than copy existing media, why creative software must move beyond manipulating tracks and pixels toward natural-language interaction, and how Runway gives creators control over video generation through Motion Brush and camera controls.

    Mar 27, 2024

    Watch on YouTubeView transcript
    26 min

    2024 will be the year of Generative AI in production in the enterprise | Florian Douetteau from Dataiku

    Florian Douetteau is the Co-founder and CEO at Dataiku. We cover why enterprises need to plan to switch LLM providers as applications mature, how generative AI supplies the last mile of personalization on top of traditional recommendation systems, and why cost monitoring can prevent an internal application from reaching a $100,000 run rate.

    Mar 20, 2024

    Watch on YouTubeView transcript
    44 min

    Why Ecosystem Partnerships Are The Future In The Age Of AI | Bob Moore from Crossbeam

    Bob Moore is the CEO at Crossbeam. We cover why RJMetrics lost to a stack of connected products, how Crossbeam lets partners map account overlap without sharing their full customer lists, and why AI is making inbound discovery and cold outbound less predictable while ecosystem-led growth improves efficiency.

    Mar 13, 2024

    Watch on YouTubeView transcript
    50 min

    Saving lives with AI | Emi Gal from Ezra

    Emi Gal is the Founder and CEO at Ezra. We cover how Ezra uses AI to cut MRI scan times, why its proprietary longitudinal MRI dataset is its key advantage, and why radiologists will be replaced by radiologists using AI rather than by AI alone.

    Mar 6, 2024

    Watch on YouTubeView transcript
    1h 2m

    Embracing AI: Lessons from Intercom's Journey

    Des Traynor is the Co-founder at Intercom. We cover why GPT-4 made automated customer service viable for conversational problem-solving, why customers value instant answers over personal tone for routine questions, and why Intercom prices Fin per resolved answer rather than relying only on seats.

    Feb 29, 2024

    Watch on YouTubeView transcript
    47 min

    Is the Modern Data Stack Dead? with dbt Labs CEO Tristan Handy

    Tristan Handy is the CEO at dbt Labs. We cover why the modern data stack is no longer a useful pre-cloud versus post-cloud distinction, why buyers no longer want to integrate nine separate products, and how a semantic layer can make AI-generated SQL deterministic for shared business metrics.

    Feb 22, 2024

    Watch on YouTubeView transcript
    33 min

    Vector Databases and the Future of AI-Native Applications with Weaviate’s CEO Bob van Luijt

    Bob van Luijt is the Co-Founder & Co-CEO at Weaviate. We cover why RAG is less risky than fine-tuning but remains primitive, how hybrid search combines semantic retrieval with BM25 for exact identifiers, and why doubling embedding dimensions doubles memory needs as deployments scale to billions of objects.

    Feb 15, 2024

    Watch on YouTubeView transcript
    36 min

    Bootstrapping a Decacorn on $15k with Celonis CEO Alexander Rinke

    Alexander Rinke is the Co-Founder & Co-CEO at Celonis. We cover how Celonis built its first prototype in three weeks and found its first customers with $15,000 in funding, why enterprise companies need product-market-people fit before scaling sales or partnerships, and how its Process Intelligence Graph connects data across systems to identify process friction automatically.

    Feb 8, 2024

    Watch on YouTubeView transcript
    42 min

    Nomic: Truly open AI

    Brandon Duderstadt is the CEO and Co-Founder at Nomic AI. We cover why curating training data lets GPT4All remove refusal and one-word-answer behaviors, how Atlas surfaces model failures as localized patches in a dataset map, and why releasing model weights without training data limits reproducibility and auditability.

    Feb 1, 2024

    Watch on YouTubeView transcript
    45 min

    Poolside: AI for Software Development with CTO Eiso Kant

    Eiso Kant is the CTO and Co-Founder at Poolside. We cover why executable code lets models learn from scalable feedback rather than human evaluation, why Poolside expects most developers to adopt AI pair-programming assistants within 24 months, and why natural-language planning helps models find solutions that single-shot code generation misses.

    Dec 20, 2023

    Watch on YouTubeView transcript
    33 min

    ASAPP: Generative AI for Contact Centers with CEO Gustavo Sapoznik

    Gustavo Sapoznik is the Founder and CEO at ASAPP. We cover why contact centers combine $600 billion in annual spending with agent attrition near 100%, how agent suggestions turn 20 seconds of typing into a one-second click, and why ASAPP trains specialized models when GPT-4 could make a conversation cost $50 to $100 rather than 30 cents.

    Dec 13, 2023

    Watch on YouTubeView transcript
    29 min

    Scott Belsky: AI & Creativity

    Scott Belsky is the Chief Strategy Officer at Adobe. We cover that the show publishes new conversations every Wednesday, that video recordings are available on the Data Driven NYC YouTube channel, and that important links appear in the show notes.

    Dec 6, 2023

    Watch on YouTubeView transcript
    25 min

    Glean AI: The ML-Powered Accounting Solution with CEO Howard Katzenberg

    Howard Katzenberg is the CEO at Glean AI. We cover why manual vendor reviews repeatedly uncover about 10% in savings despite 99 of 100 bills being approved, how Glean uses multiple LLMs as a proxy confidence score and routes uncertain extractions to humans, and why it keeps calculations outside LLMs when analyzing vendor history.

    Nov 29, 2023

    Watch on YouTubeView transcript
    35 min

    Humanloop: LLM Collaboration and Optimization with CEO Raza Habib

    Raza Habib is the CEO at Humanloop. We cover why subjective, stochastic LLM outputs require evaluation across development, production, and regression testing, how tying feedback and observability to prompt editing cuts fixes from lengthy deployment workflows to the same day, and why teams often prototype with GPT-4 before fine-tuning smaller models for cost and latency.

    Nov 22, 2023

    Watch on YouTubeView transcript
    34 min

    DeepScribe: The AI-Powered Medical Scribe with CEO Akilesh Bapu

    Akilesh Bapu is the CEO and Founder at DeepScribe. We cover how GPT-4 shifted buyers from asking whether to adopt ambient documentation to choosing a vendor, why DeepScribe needed 5 million labeled conversations to match human scribes, and how a multi-armed bandit selects speech vendors from clinician edits.

    Nov 15, 2023

    Watch on YouTubeView transcript
    44 min

    Lamini: Fine-Tuning LLMs for The Enterprise with CEO Sharon Zhou

    Sharon Zhou is the CEO at Lamini. We cover why fine-tuning learns from data while prompt engineering is strongest with little data, how narrow domain boundaries make model guardrails easier, and how parameter-efficient methods cut switching among 1,000 models from an estimated three months to three milliseconds.

    Nov 8, 2023

    Watch on YouTubeView transcript
    41 min

    Perplexity AI: The AI-Powered Answer Engine with CEO Aravind Srinivas

    Aravind Srinivas is the CEO at Perplexity AI. We cover why Perplexity launched cited web answers one week after ChatGPT and found a broader market than Twitter search, why search depends more on index quality and ranking than index size, and why AI should clarify user intent instead of requiring prompt-engineering skill.

    Nov 1, 2023

    Watch on YouTubeView transcript
    46 min

    Diffblue: AI For Code Testing with CEO Mathew Lodge

    Mathew Lodge is the CEO at Diffblue. We cover why LLM coding assistants trade generality for errors, how Diffblue analyzes entire codebases to iteratively generate boundary-testing unit tests, and why it generated 3,100 tests for a bank in eight hours rather than a developer's estimated year of work.

    Oct 25, 2023

    Watch on YouTubeView transcript
    45 min

    Moonhub AI: The On-Demand AI Recruiter with Founder & CEO Nancy Xu

    Nancy Xu is the Founder & CEO at Moonhub AI. We cover how Moonhub indexes public-web data on about one billion people, why hiring managers can search for candidates without Boolean-query expertise, and why Moonhub customers report that 80% of candidates they find through the platform were unavailable on other platforms.

    Oct 18, 2023

    Watch on YouTubeView transcript
    54 min

    Dust: Secure AI Assistants for The Enterprise with Co-Founder Stanislas Polu

    Stanislas Polu is the Co-Founder at Dust. We cover why company-wide AI adoption follows a traditional adoption curve, why focused assistants onboard non-early adopters better than general-purpose chatbots, and why retrieving the right company data remains a bigger challenge than model hallucination.

    Oct 11, 2023

    Watch on YouTubeView transcript
    45 min

    Imbue: AI Agents That Can Reason with CEO Kanjun Qiu

    Kanjun Qiu is the CEO at Imbue. We cover why reliable agents need judgment about when to ask questions and flag risk, how code serves as the internet’s clearest curriculum for step-by-step reasoning, and why Imbue treats current AI as an analog-computer phase where errors compound and new abstractions are needed.

    Oct 4, 2023

    Watch on YouTubeView transcript
    52 min

    Guardrails AI: Deploying Generative AI Safely with CEO Shreya Rajpal

    Shreya Rajpal is the Co-founder & CEO at Guardrails AI. We cover why 10-minute GenAI prototypes fail to deliver consistent customer value, why fine-tuning does not eliminate hallucinations, and how runtime validators can enforce organization-specific rules for accuracy, compliance, privacy, and brand safety.

    Sep 27, 2023

    Watch on YouTubeView transcript
    52 min

    AI21 Labs: Bringing Reliable AI to the Enterprise with Co-Founder Ori Goshen

    Ori Goshen is the Co-Founder and Co-CEO at AI21 Labs. We cover why larger models face diminishing returns for reliable reasoning, why 95% of enterprise use cases are narrow enough for task-specific systems, and how grounded question answering can turn hours of document research into minutes.

    Sep 20, 2023

    Watch on YouTubeView transcript
    40 min

    AI, Gaming, and Building a Data Science Brand - A Conversation with Carly Taylor, Founder of Rebel Data Science

    Carly Taylor is the Founder at Rebel Data Science. We cover how machine learning security focuses on outliers and changes adversary behavior, why generated NPC dialogue can make open-world games less repetitive without replacing existing work, and why startups need data engineering before expecting data scientists to own an end-to-end pipeline.

    Sep 13, 2023

    Watch on YouTubeView transcript
    46 min

    Deepset AI: NLP For The Enterprise with CEO Milos Rusic

    Milos Rusic is the CEO & Co-founder at deepset AI. We cover how Haystack breaks RAG applications into chunking, storage, retrieval, and generation; why grounding a model in a document corpus still does not prevent factual errors; and why enterprise LLM products require continuous evaluation, monitoring, and iteration after deployment.

    Sep 6, 2023

    Watch on YouTubeView transcript
    33 min

    BigID: Data Visibility & Control with CEO Dimitri Sirota

    Dimitri Sirota is the Co-Founder and CEO at BigID. We cover why GDPR expands personal data far beyond traditional identifiers, how BigID uses scanners and graph ML to locate and correlate data across cloud and legacy systems, and why enterprise configurability can overwhelm smaller companies.

    Aug 30, 2023

    Watch on YouTubeView transcript
    49 min

    Hippocratic AI: A Safety-First LLM for Healthcare with CEO Munjal Shah

    Munjal Shah is the CEO and Co-founder at Hippocratic AI. We cover why healthcare’s biggest opportunity is adherence and support rather than diagnosis, how voice agents costing roughly 18 cents an hour could provide “super staffing,” and why Hippocratic AI relies on clinician-led testing and safety kick-outs before deployment.

    Aug 23, 2023

    Watch on YouTubeView transcript
    47 min

    You.com: An AI Chatbot to Search The Web with CEO Richard Socher

    Richard Socher is the CEO at You.com. We cover why he argues search deteriorates under monopoly conditions, how retrieval-augmented generation lets language models reason over current web facts, and why organizations should not train their own LLMs from scratch without hundreds of millions of responses.

    Aug 16, 2023

    Watch on YouTubeView transcript
    52 min

    Weights & Biases: The Developer-First MLOps Platform with CEO Lukas Biewald

    Lukas Biewald is the CEO at Weights & Biases. We cover why tracking training data, infrastructure, and hyperparameters is necessary for reproducible models, why most enterprises have not put LLMs into production because they are hard to evaluate and monitor, and why sophisticated teams still use logistic regression when it fits the job.

    Aug 9, 2023

    Watch on YouTubeView transcript
    44 min

    Cockroach Labs: A Cloud SQL Database Built for Survival with CEO Spencer Kimball

    Spencer Kimball is the CEO at Cockroach Labs. We cover the show’s weekly Wednesday schedule, where video recordings of episodes are available, and that the Data Driven NYC YouTube channel hosts many more conversations.

    Aug 2, 2023

    Watch on YouTubeView transcript
    47 min

    Ada: AI-First Customer Service with CEO Mike Murchison

    Mike Murchison is the Co-Founder & CEO at Ada. We cover why companies that increase customer contact can outperform those trying to deflect it, how Ada charges only for accurate, relevant, and safe automated resolutions, and why reaching 100% automated resolution is more a cultural and API-access challenge than a technological one.

    Jul 26, 2023

    Watch on YouTubeView transcript
    44 min

    Synthesia: Creating Video with Generative AI with CEO Victor Riparbelli

    Victor Riparbelli is the CEO at Synthesia. We cover why Synthesia’s videos replace text rather than video production, how enterprises use them to turn 40-page training manuals into multilingual video, and why production AI requires workflows for fact-checking rather than a simple API call.

    Jul 19, 2023

    Watch on YouTubeView transcript
    41 min

    LlamaIndex: Unleashing LLMs on Your Data with CEO Jerry Liu

    Jerry Liu is the Co-Founder & CEO at LlamaIndex. We cover that no substantive guest transcript was provided, that specific episode claims cannot be verified, and that the supplied metadata identifies LlamaIndex’s focus on connecting LLMs with data.

    Jul 12, 2023

    Watch on YouTubeView transcript
    49 min

    Dataiku: The Single Platform for Everyday AI with CEO Florian Douetteau

    Florian Douetteau is the CEO and Co-Founder at Dataiku. We cover why enterprise AI needs one platform rather than stitched-together tools, how generative AI makes previously uneconomic unstructured-data use cases viable, and why companies need accountability, fairness, transparency, and reliability frameworks before regulation arrives.

    Jul 5, 2023

    Watch on YouTubeView transcript
    26 min

    Chroma: Programmable AI-Native Databases with Co-Founder Jeff Huber

    Jeff Huber is the Co-Founder at Chroma. We cover why programmable memory lets developers deterministically specify the knowledge and tools an LLM uses, why grounding a model with retrieved documents is the state-of-the-art response to hallucinations, and why AI-native databases need both transactional and analytical workloads.

    Jun 28, 2023

    Watch on YouTubeView transcript
    23 min

    Hebbia: Saving Smart People from Mundane Tasks with CEO George Sivulka

    George Sivulka is the CEO at Hebbia. We cover why Hebbia began with 400-page DEF 14As, how private equity diligence maps unstructured data rooms to 100- or 400-question trackers, and why chaining 90% accurate AI steps produces poor multi-step results.

    Jun 21, 2023

    Watch on YouTubeView transcript
    23 min

    How Data Happened - A Conversation with Chris Wiggins, Author and Chief Data Officer at The New York Times

    Chris Wiggins is the Chief Data Officer at The New York Times. We cover how statistics originally meant statecraft rather than mathematics, why early digital computers were built to process messy codebreaking data at Bletchley Park, and how deep neural networks decisively beat other image-labeling methods in 2012.

    Jun 14, 2023

    Watch on YouTubeView transcript
    29 min

    Notion: An AI Assistant to Work Faster with Head of Data Daniel Sternberg

    Daniel Sternberg is the Head of Data at Notion. We cover why Notion AI’s most-used features summarize content and generate meeting action items, how a small cross-functional team used existing infrastructure to ship quickly, and why Notion does not inspect external workspace content unless users opt into specific logging.

    Jun 7, 2023

    Watch on YouTubeView transcript
    27 min

    Pinecone: Long Term Memory for AI with CEO Edo Liberty

    Edo Liberty is the Founder & CEO at Pinecone. We cover why AI systems should retrieve external data at inference rather than fit it into models, how vector databases serve as long-term memory across billions of embeddings, and why adding retrieved context can measurably reduce hallucinations.

    May 31, 2023

    Watch on YouTubeView transcript
    28 min

    Vectara: LLM Powered Search with CEO Amr Awadallah

    Amr Awadallah is the CEO at Vectara. We cover why grounded generation retrieves facts before generating answers to minimize hallucinations, why fine-tuning private data is slower and 100 times more expensive, and why action engines require zero hallucination before software can safely execute requests.

    May 24, 2023

    Watch on YouTubeView transcript
    32 min

    Amplify Partners: AI Investing and the Startup Landscape with General Partner Sarah Catanzaro

    Sarah Catanzaro is the General Partner at Amplify Partners. We cover why AI applications still need to deliver everyday value rather than compete for mindshare, why incumbents can rapidly integrate LLMs while startups should redesign workflows rather than copy existing tools, and why cheap storage but expensive compute is reviving ETL and edge-based transformations.

    May 17, 2023

    Watch on YouTubeView transcript
    27 min

    Assembly AI: Generative AI for Speech Recognition with CEO Dylan Fox

    Dylan Fox is the Founder & CEO at AssemblyAI. We cover why transcription creates more value as input to downstream features than as an endpoint, why AssemblyAI reserves tailored models for use cases with product-market fit, and why scaling labeled audio data still leaves speech recognition with roughly 15% error rates on many datasets.

    May 10, 2023

    Watch on YouTubeView transcript
    28 min

    Lightning AI: Build and Deploy AI with Pytorch with CEO William Falcon

    William Falcon is the Founder and CEO at Lightning AI. We cover why collaborative ML projects become unmaintainable when distributed-training changes are duplicated across forks, why enterprise LLM deployments require auditability and human oversight because models hallucinate, and why he expects specialized small language models to replace giant models for deployment on a single GPU.

    May 3, 2023

    Watch on YouTubeView transcript