2019 was the year the Lex Fridman Podcast found its shape: long, unhurried conversations with the people actually building the technology everyone else was just talking about. Compilers, self-driving cars, string theory, Watson, C++, Spotify's recommendation engine. It's a strange, wonderful mix, and going back through our full library of episode summaries made one thing obvious: some of these conversations hold up as some of the best technical interviews ever recorded, not just early podcast episodes.
We built this list straight from our own summaries of every 2019 episode, ranked by how much substance actually gets covered, not by view count or hype. Below are the fifteen that earn a real recommendation, whether you're into AI, physics, robotics, or just want to hear smart people think out loud for two hours.
Bjarne Stroustrup: C++ | Lex Fridman Podcast #48
The creator of C++ explains the zero-overhead principle that has kept his 40-year-old language running critical systems at Google, Facebook, and Amazon, and claims C++ matrix multiplication can actually beat Fortran when the right abstraction eliminates temporaries. He also names Bitcoin mining his least favorite use of C++, pointing out it burns as much energy as Switzerland. Anyone who writes software, or just wants to understand why some languages survive four decades, should listen.
Read the full episode notesRay Dalio: Principles, the Economic Machine, AI & the Arc of Life | Lex Fridman Podcast #54
Dalio walks through the 1982 call that wiped out his clients and left him borrowing $4,000 from his dad, then explains how that failure built the idea-meritocracy culture at Bridgewater. He lays out his contrarian AI rule (don't trust it when the future can differ from the past and you don't understand the cause and effect), and drops a genuinely surprising data point: the least happy years of life are 45 to 55, the happiest are 70 to 80. Essential listening for anyone interested in economics, failure, or how to actually build a truth-seeking organization.
Read the full episode notesMichio Kaku: Future of Humans, Aliens, Space Travel & Physics | Lex Fridman Podcast #45
Kaku extends the Kardashev scale to a type 4 civilization powered by dark energy after a kid's question at a planetarium stumped him, then goes further into digital immortality: digitizing the human brain and beaming it across space by laser into an avatar. He also explains, carefully, why he thinks the universe cannot be a simulation because no finite Turing machine could run it. This one is for anyone who wants big, unfiltered physics speculation without the hedging.
Read the full episode notesGeorge Hotz: Comma.ai, OpenPilot, and Autonomous Vehicles | Lex Fridman Podcast #31
The comma.ai founder reveals Elon Musk once offered him $12 million to match Mobileye's performance, losing $1 million for every month of delay, and lays out why he thinks lidar is a crutch and end-to-end learning is the only real path past human-level driving. He's candid about the company's finances too: roughly $200k a month burn against $100k in revenue. Recommended for anyone who wants the unvarnished, technical case against the industry's self-driving consensus.
Read the full episode notesDavid Ferrucci: IBM Watson, Jeopardy & Deep Conversations with AI | Lex Fridman Podcast #44
The engineer who built IBM Watson admits the system never came close to knowing all of Jeopardy, the correct answer was findable only about 85% of the time, and that he deliberately chose not to solve general language understanding, just to win the game by any means necessary. He closes with a wrenching personal story: doctors declared his father brain dead on a purely statistical argument, and Ferrucci refused to accept it. A must for anyone who wants to understand the real gap between prediction and reasoning in AI.
Read the full episode notesJeff Hawkins: Thousand Brains Theory of Intelligence | Lex Fridman Podcast #25
Hawkins lays out his Thousand Brains Theory: there's no single model of any object in your head, thousands of cortical columns each build a complete model and vote to reach consensus, all stored in reference frames like a CAD system. He also argues backpropagation flatly cannot happen in real brains, since biological learning is Hebbian synaptogenesis on dendritic segments. Listen if you want the neuroscience case against how today's deep learning actually works.
Read the full episode notesColin Angle: iRobot CEO | Lex Fridman Podcast #39
The iRobot CEO names five well-funded, brilliant robotics companies, Anki, Jibo, Mayfield Robotics, Sci-Fi Works, Rethink Robotics, that all went out of business, and admits Roomba only took off once he became, in his own words, a vacuum cleaner salesman. He also confesses that being upfront about iRobot's privacy commitments on the box would sell fewer units. A grounded, funny counterweight to the more speculative AI episodes on this list, good for anyone interested in what actually makes a robotics company survive.
Read the full episode notesChris Lattner: Compilers, LLVM, Swift, TPU, and ML Accelerators | Lex Fridman Podcast #21
The creator of LLVM and Swift reveals he started both projects quietly, nights and weekends, without telling anyone at Apple, and that doing a new language felt 'heretical' since the team believed the iPhone's success was tied to Objective-C. He also recounts his five-month stint as VP of Tesla Autopilot software, where he says he witnessed the highest employee turnover he'd ever seen. Worth it for anyone curious how the tools that build modern software actually got made.
Read the full episode notesYann LeCun: Deep Learning, ConvNets, and Self-Supervised Learning | Lex Fridman Podcast #36
LeCun explains why massive over-parameterized neural nets working at all broke every pre-deep-learning textbook rule, and argues, contrary to popular assumption, that human intelligence isn't actually general, it's highly specialized. He also reveals neural nets fell out of favor around 1995 partly because AT&T's lawyers blocked an open-source release. A sharp pick for anyone who wants the philosophy behind deep learning from one of its founding fathers, not just the hype.
Read the full episode notesGustav Soderstrom: Spotify | Lex Fridman Podcast #29
Spotify's R&D chief admits the company's playlist-based recommendation engine was built on 'dumb luck,' billions of user-made playlists turned out to be a goldmine of semantically labeled tracks nobody planned for. He also reveals Spotify's original edge over piracy was a peer-to-peer streaming stack that started playback in about 250 milliseconds. Good listening for anyone interested in music, recommender systems, or how a legal streaming product actually beat free piracy.
Read the full episode notesFrançois Chollet: Keras, Deep Learning, and the Progress of AI | Lex Fridman Podcast #38
The Keras creator makes the case against an AI intelligence explosion, arguing recursively self-improving systems hit exponential friction rather than runaway growth, and that even science itself is a recursively self-improving system whose output stays stubbornly linear. He also warns, bluntly, that combining AI with our digital lives already enables mass manipulation of behavior at population scale. Recommended for anyone who wants a rigorous skeptic's view inside the deep learning field itself.
Read the full episode notesDrago Anguelov (Waymo) - MIT Self-Driving Cars
Waymo's principal scientist details a fleet simulating roughly 25,000 virtual cars driving ten million miles a day, over seven billion simulated miles total, built to expose rare edge cases like a cyclist carrying a stop sign or a falling pole. He also describes applying Google's AutoML to lidar segmentation and lane detection, finding architectures that beat human engineers on both quality and latency. A solid pick for anyone who wants the actual engineering behind self-driving, not the marketing version.
Read the full episode notesLeonard Susskind: Quantum Mechanics, String Theory and Black Holes | Lex Fridman Podcast #41
One of the founders of string theory explains that the real power of quantum computers will be simulating quantum systems, not factoring numbers, which he calls almost a fluke of popular attention. He also reveals a second job most listeners wouldn't expect: consulting for Google X as senior academic advisor to a group of machine learning physicists. Recommended for anyone who wants physics explained through intuition rather than equations.
Read the full episode notesSean Carroll: Quantum Mechanics and the Many-Worlds Interpretation | Lex Fridman Podcast #47
Carroll makes his case for the many-worlds interpretation of quantum mechanics, insisting it does not violate conservation of energy since the universe splits into thinner branches rather than duplicating itself. He also offers a startlingly concrete estimate for the dimensionality of our observable Hilbert space: 10 to the power of 10 to the power of 122. A great listen for anyone who wants quantum mechanics explained in plain language by someone who also hosts his own interview podcast.
Read the full episode notesRosalind Picard: Affective Computing, Emotion, Privacy, and Health | Lex Fridman Podcast #24
The founder of affective computing admits she's deliberately putting the brakes on her own research given how emotion-recognition technology is being used for surveillance and control in places like China. She also traces how a strange skin-conductance spike in an autistic child's wrist led to detecting hidden seizures, work that eventually produced the FDA-cleared Embrace wristband. Worth it for anyone interested in the ethics of AI that reads human emotion, not just the technology itself.
Read the full episode notesThat's fifteen from a single year, and it barely scratches what's in our full library. Browse the rest of our episode summaries to find the conversation that fits whatever you're curious about next.