Latest Interviews
Showing 46–60 of 375 transcripts.
Clear all filters- Sequoia Capital27 min
Waymo's Dmitri Dolgov: 20 Million Rides and the Road to Full Autonomy
Dmitri Dolgov, Konstantine Buhler
Dmitry Dolgov, Waymo's co-founder, outlines a 21-year evolution from DARPA challenges to a sixth-generation autonomous fleet that has achieved 200 million fully autonomous miles and delivered over 20 million rides with a safety profile 13 times superior to human drivers. The company's architecture utilizes a multimodal "Foundation Model" integrating physics reasoning and spatial understanding to power a structured, end-to-end system across 11 cities, now accelerating toward global expansion in London and Tokyo through the newly launched Waymo Origin vehicle. This mission-driven approach bypasses incremental improvements to rapidly scale commercial deployment, treating full autonomy as an existential imperative to eliminate the 13 million annual road fatalities while transitioning the core technology into a primary family transportation tool.
- Sequoia Capital20 min
Robotics' End Game: Nvidia's Jim Fan
NVIDIA Robotics is transitioning from language-centric VLA models to World Action Models like Dream Zero, which prioritize learning physical laws through video prediction to enable zero-shot generalization. This architectural shift is supported by a new data collection strategy replacing low-efficiency teleoperation with scalable egocentric video and neural simulators, establishing a log-linear scaling law between training hours and dexterity. Looking toward 2040, Jim Phan projects the industry will achieve a "Physical Turing Test," an orchestration "Physical API" for automated fleets, and "Physical Auto-Research" where robots autonomously improve their own designs.
- a16z1h 6m
Balaji Srinivasan on The Future of AI | The a16z Show
Balaji Srinivasan, Erik Torenberg
The speaker redefines human-AI collaboration by positioning humans as strategic leaders who direct specialized AI "actors," creating a trusted ecosystem where value shifts toward personalized, private instances and verification services. Parallel to this technological evolution, the discussion details a bifurcated future for digital currency in which Bitcoin serves as transparent institutional collateral while Zcash-powered tools like Zedal provide quantum-resistant privacy for individual transactions. Ultimately, the presentation argues that political constraints and resource scarcity will prevent centralized AI dominance, fostering a polytheistic landscape of decentralized models and democratized global entrepreneurship.
- The Economist8 min
Will the Iran war spread across the Middle East?
Greg, Anshul, Adam, Pete Hexeth
Experts warn that regional chaos in Iran could trigger a prolonged conflict, spiking energy prices and threatening the security models of Gulf states like Saudi Arabia and Qatar. While historically neutral, these nations are shifting toward military alignment with the US and abandoning diplomatic restraint as Iranian strikes backfire by uniting regional populations against Tehran. This escalation occurs amidst a US strategy driven by legacy considerations rather than broad public support, aiming to transform Iran's political landscape while balancing fears of a quagmire against the need to signal power to global rivals like China.
- Y Combinator8 min
The New Way To Build A Startup
Companies such as GigaML, Legion Health, and PhaseShift are redefining operational scale by automating all internal workflows—from code integration to clinical operations—rather than merely augmenting isolated functions. This "20X company" model enables lean teams of just a few employees to service Fortune 500 clients and achieve fourfold revenue growth while maintaining flat headcounts that would traditionally require hundreds of staff. By treating internal processes as the primary product, these organizations achieve AGI-level productivity, allowing them to compete directly with massive incumbents without the corresponding payroll burdens.
- a16z13 min
AI in 2026: 3 Predictions For What’s To Come (a16z Big Ideas)
Oliver Hsu, Bryan Kim, David Haber
Speakers Oliver Hsu, Brian Kim, and David Haber analyze distinct trajectories for AI adoption, ranging from autonomous scientific discovery in life sciences to consumer platforms shifting focus from productivity to digital connectivity. Hsu identifies the maturation of robot learning and simulation as prerequisites for closing the loop on self-driving labs, while Kim argues that future consumer value hinges on AI agents facilitating human relationships through deep emotional engagement. Concurrently, Haber demonstrates that the most robust business models integrate AI to directly reinforce revenue generation and proprietary data loops, as seen in contingency law firms and loan servicing platforms that scale success rather than merely cutting costs.
- a16z14 min
How AI Agents Will Transform in 2026 (a16z Big Ideas)
Marc Andrusko, Stephanie Zhang, Olivia Moore, Steph Ng
The event details a fundamental shift from reactive prompt-based interfaces to proactive AI agents that autonomously diagnose issues and execute actions, expanding the addressable market from software spending to the broader labor economy. Speakers discuss how design priorities are evolving toward machine legibility to enable high-autonomy applications in sales, security, and content discovery, while voice AI agents achieve scalable enterprise deployment in healthcare, finance, and customer service by outperforming humans in compliance and multilingual accuracy. This transition is redefining the human-in-the-loop dynamic, where critical decisions retain oversight while routine tasks are increasingly automated to drive efficiency across nearly every industry vertical.
- Jane Street48 min
Making GPUs Actually Fast: A Deep Dive into Training Performance
Corwin, Savant Diaz, Sylvain De Wecker
Jane Street engineers optimize deep learning infrastructure by eliminating CPU-GPU synchronization bottlenecks and fusing PyTorch operations via `torch.compile` and Triton to maximize throughput. When automated compilation fails on complex Python logic, they deploy custom CUDA kernels that leverage shared memory and warp-level reductions to achieve nearly 1,000x speedups in specialized tensor operations. This hierarchical approach, ranging from standard PyTorch to hand-optimized C++, ensures efficient utilization of the H100's 132 Streaming Multiprocessors and strict memory bandwidth constraints.
- a16z39 min
Marc Andreessen and Ben Horowitz on the State of AI
Marc Andreessen, Ben Horowitz, Erik Torenberg
Industry leaders assess that AI product evolution will likely follow historical cycles of radical interface shifts while facing immediate constraints from a severe talent shortage and a looming hardware transition from chip scarcity to power availability. Geopolitical analysis warns that although the U.S. currently leads in software innovation, a six-month gap combined with American deindustrialization risks allowing China to overtake the nation in the critical embodied robotics phase. Despite speculation about a market bubble, the consensus remains that demand currently exceeds supply, with long-term success ultimately determined by whether customers validate the technology and by how effectively legacy incumbents and new entrants execute superior product strategies.
- Y Combinator39 min
From Idea to $650M Exit: Lessons in Building AI Startups
Case Text, led by a founder who pivoted the company to artificial intelligence in summer 2022, was acquired by Thomson Reuters for $650 million following the successful launch of its Co-Counsel legal assistant. The speaker outlines a methodology for building reliable AI applications by combining domain expertise with granular workflow decomposition, rigorous evaluation metrics, and iterative prompt engineering to achieve near-human accuracy. Key strategic insights for scaling include adopting value-based pricing, prioritizing side-by-side human-AI pilots to build trust, and viewing proprietary data integration and evaluation frameworks as the primary sources of long-term defensibility.
- All-In Podcast11 min
Bryan Johnson’s Best Health Hack: Lower Your RHR Before Sleep = Sleep Better and Live Longer
Brian Johnson presents an ambitious framework positioning the indefinite extension of human life as a new global ideology, utilizing biological age testing to illustrate that organs age independently and require targeted management. He identifies resting heart rate as the primary health metric and details a specific protocol involving strict sleep hygiene, diet timing, and screen discipline to lower this rate by approximately 10%. This approach aims to reverse aging trends while simultaneously treating sleep as the fundamental intervention for anxiety and depression, ultimately preparing humanity for a future defined by superintelligence.
- RAISE Summit20 min
A Keynote by Thomas Sohmers, Co Founder & CTO of Positron
Positron AI, a startup founded in 2023, is deploying specialized hardware to address the inference bottleneck that now drives global GDP growth and limits traditional GPU efficiency for Transformer models. Their Atlas server and upcoming Titan system prioritize massive memory bandwidth over raw compute, delivering 70% higher throughput per dollar while operating at 150 watts to eliminate the need for liquid cooling. By enabling concurrent execution of multiple models up to 512 billion parameters on standard air-cooled infrastructure, the company aims to drastically reduce AI operating costs and reshape the economic landscape of software development.
- RAISE Summit18 min
Cresta: The Path to 90% Service Automation: Pitfalls, Secrets & Lessons from Enterprise Deployments
Ping Wu, CEO of Cuesta, addresses the unique challenges of automating contact centers by distinguishing private, legacy-bound support data from the public datasets that drove coding automation. His platform deploys a three-part architecture of conversation intelligence, agent assistants, and native AI agents to achieve high task-level automation metrics while maintaining human oversight for complex escalations. This strategy aims to shift customer experience from defensive cost-cutting to an era of abundant, proactive engagement where AI handles 90% of sub-tasks, lowering interaction costs and enabling more frequent customer contact.
- RAISE Summit16 min
June Paik (FuriosaAI): Tensor Contraction Processor: NextGen AI Inference Chip for Data Centers
Seoul-based Furiosa AI CEO Jun Paek unveiled the Renegade, a second-generation silicon chip engineered to solve energy bottlenecks in the shift toward frontier agentic model inference. Built on TSMC's 5-nanometer node, this Tensor Contraction Processor delivers 190% superior tokens-per-second-per-watt performance compared to NVIDIA H100 units by optimizing for 1.5 terabytes per second memory bandwidth and 180-watt power consumption. Currently in the sampling phase with global enterprise clients, the chip enables manufacturers and sovereign sectors to own their full AI stack while drastically reducing the capital and operational costs associated with scaling massive data centers.
- RAISE Summit20 min
Des Traynor, Co-Founder & CSO @ Intercom & Creator of Fin.ai: The Death of SaaS, The Dawn of Agents
Industry leaders identify AI agents as an existential force converging fragmented B2B SaaS workflows, presenting a critical "adapt or die" choice where obsolescence looms for firms unable to deploy robust, job-executing systems within two years. While strategic ambition ranges from efficiency-boosting copilots to full organizational replacements, success hinges on overcoming severe operational hurdles regarding unstructured data, compounding error rates in multi-step chains, and balancing agent agency with strict reliability controls. Although pioneers like Intercom report five-year growth through this paradigm, widespread adoption requires fundamentally rewriting pricing, sales, and validation architectures to bridge the gap between experimental prototypes and production-ready performance.