Latest Interviews
Showing 1–15 of 23 transcripts.
Clear all filters- Sequoia Capital22 min
Continual Learning: How AI Agents Get Better With Every Use | Arjun Karanam, Trajectory
Arjun Karanam, Ronak, Gabe, Harrison, Nico, Harvey
Trajectory, co-founded by Arjun and Ronak, addresses the lack of accumulated experience in AI by building a platform that enables models to continuously learn from the 100 trillion daily tokens generated by real-world agent interactions. The company utilizes a dual-learning architecture combining differential privacy with reinforcement learning on user-corrected traces, allowing organizations to transition from static models to systems that compound capability through automated post-training and flexible harness optimization. By abstracting complex training parameters into a 15-minute workflow, Trajectory empowers enterprises to retain ownership of their specialized models while refining agent performance directly against production traffic.
- Sequoia Capital24 min
When to Build Your Own Agent Harness | Harrison Chase, LangChain
The framework defines autonomous agents as systems built from three owned components: the model, context, and a prioritized harness that orchestrates data flow through an iterative LLM loop. Organizations can customize this harness via middleware for domain-specific optimizations or maintain off-the-shelf versions for in-distribution tasks, ensuring compatibility through dynamic model profiles. Continuous improvement is driven by a flywheel where trace data from evaluations using the Harbor benchmark feeds into an automated engine that identifies failures and suggests prompt, code, or context fixes.
- Sequoia Capital26 min
RL Environments Explained: How AI Agents Learn Real-World Work | Brendan Foody, Mercor
Mercore has expanded its revenue run rate to $2 billion by transitioning the AI data market from basic crowdsourcing to high-skilled "agentic data" services that enable frontier labs to build complex reinforcement learning environments. The company leverages expert networks of lawyers, engineers, and doctors to create realistic simulated worlds with precise human-verified rubrics, demonstrating a fivefold increase in model performance on specific legal tasks during recent training. As the primary data vendor for major application layer companies, Mercore addresses the industry's need for ultra-long horizon tasks and social dynamics evaluations that synthetic models cannot yet self-generate.
- Sequoia Capital28 min
Post-Training Is How You Keep Your Taste | Fireworks CEO Lin Qiao
Fireworks CEO Linh Nguyen advocates for a strategic industry shift from relying on rented APIs to owning intelligence through deep model customization, enabling companies to preserve unique business judgment while reducing inference costs by five to ten times. This approach utilizes a structured lifecycle of data curation, fine-tuning, and serving loops to transition from generic prompting to specialized models, as demonstrated by success stories like Cursor and niche vertical leaders in healthcare and security. Ultimately, post-training is positioned as the critical mechanism for startups to scale after product-market fit by converting proprietary user data into unclonable domain expertise before high API expenses threaten unit economics.
- Sequoia Capital29 min
How Harvey Built a Research Lab on a Budget | Gabe Pereyra
Harvey, Gabe Pereyra, Brendan, Julio, Ross, Brock
Harvey differentiates itself from well-funded frontier labs by leveraging an application-layer strategy that combines synthetic data generation guided by domain experts with post-training on open-source models. The company builds specialized legal benchmarks and utilizes infrastructure partnerships to train agents on complex tasks like contract negotiation without exposing sensitive client information. By deploying these capabilities across multiple vendors and product surfaces, Harvey aims to solve organizational productivity challenges while mitigating the performance gaps inherent in current long-context environments.
- Sequoia Capital45 min
Memory and Continual Learning: Engram's Dan Biderman and Jessy Lin
Dan Biderman, Jessy Lin, Sonya Huang, Shaun Maguire
Ngram addresses the scalability and cognitive limitations of current retrieval-augmented generation by training custom, continually learning models directly within workspace environments. This approach utilizes adapter fine-tuning to internalize organizational knowledge into model weights, reducing inference token consumption by a factor of 100 while enabling true intuition rather than static fact retrieval. By shifting the focus from pre-training generic AI to perpetual, private adaptation, the platform aims to create personalized neural interfaces that evolve alongside a team's data.
- 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 Capital28 min
OpenAI's Greg Brockman: Why Human Attention Is the New Bottleneck
OpenAI executives estimate they are 80% toward achieving AGI, a milestone driven by aggressive compute acquisition strategies and architectural innovations that currently enable AI to autonomously engineer software kernels. The organization is pivoting toward an enterprise and consumer model where a single AGI entity handles complex goals, urging startups to leverage agentic coding tools for massive productivity gains while navigating a landscape where human attention has become the primary bottleneck. This shift is underpinned by anticipated scientific breakthroughs in physics and biology, alongside internal reforms designed to manage the risks of autonomous agents within a future where humans oversee teams of AI rather than writing code manually.
- Sequoia Capital32 min
This is AGI: Sequoia AI Ascent 2026 Keynote
Pat Grady, Sonya Huang, Konstantine Buhler
This analysis outlines a $10 trillion market opportunity driven by a paradigm shift from information distribution to autonomous computation, where AI agents are rapidly replacing cognitive labor with agentic systems capable of long-horizon execution. Founders are advised to leverage the "MAD" model—focusing on customer moats, immediate affordance, and bridging the adoption diffusion gap—to capitalize on a timeline that compresses years of work into days. The event further projects that by 2026, the convergence of these technologies will trigger a cognitive industrial revolution, fundamentally redefining human value from task execution to relationship building and strategic oversight.
- 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.
- Sequoia Capital38 min
Why the Next AI Revolution Will Happen Off-Screen: Samsara CEO Sanjit Biswas
Sanjit Biswas, Sonya Huang, Pat Grady
Samsara leverages its 90-billion-mile fleet dataset to drive a "third shift" in autonomous logistics and physical AI, prioritizing edge-computed safety and behavioral coaching over full automation replacement. With $3 billion reinvested into R&D, the company transforms legacy operations across trucking, construction, and public sector transit by integrating sensor telemetry with cloud-based video language models. This strategy aims to unlock 24/7 productivity while mitigating risk through distributed architectures that balance real-time inference with scalable digital workflow modernization.
- Sequoia Capital38 min
OpenAI Codex Team: From Coding Autocomplete to Asynchronous Autonomous Agents
Hanson Wang, Alexander Embiricos, Sonya Huang, Lauren Reeder
OpenAI has rebranded its Codex system into an agentic coding suite specifically RL-tuned to autonomously execute complex enterprise development tasks like debugging, testing, and deployment within isolated cloud environments. Internal adoption data indicates that professional engineers now leverage the tool to generate multiple parallel code iterations daily, effectively shifting their primary responsibility from writing code to validating agent outputs. This strategic pivot aims to redefine 2025 as the "year of agents" by lowering barriers to bespoke software creation while anticipating a market where human developers manage high-level workflows through future interfaces that blend in-IDE pairing with long-running background automation.
- Sequoia Capital32 min
OpenAI’s Sam Altman on Building the ‘Core AI Subscription’ for Your Life
Sam Altman, Jens Nordvigsen, DAN GALPIN, SAM SACCONE
OpenAI has evolved from a 14-person research lab into a commercial powerhouse driven by the GPT-3 API and the viral ChatGPT platform, which now serves 500 million weekly active users. Under Sam Altman's leadership, the organization prioritizes high-impact small teams and a vision of becoming a personalized AI operating system while navigating a market where startups outpace legacy enterprises in agility. The company's roadmap emphasizes the transition from text-based assistants to autonomous agents by 2025, with future infrastructure expected to facilitate seamless agent-to-agent communication and eventual robotics applications.
- Sequoia Capital31 min
Google's Jeff Dean on the Coming Transformations in AI
This presentation outlines the trajectory of deep learning from its 2012 scaling breakthroughs toward a future of multi-modal agents and physical robotics capable of performing thousands of tasks within two years. It details a converging industry landscape where a handful of foundational models drive a secondary ecosystem of efficient, specialized architectures supported by Google's upcoming Ironwood TPU generation and Pathways system. The discussion further projects how these advancements will revolutionize scientific discovery, democratize virtual engineering through junior-level AI assistants, and reshape economic structures via dynamic compute allocation.
- Sequoia Capital24 min
Anthropic CPO Mike Krieger: Building AI Products From the Bottom Up
Mike Krieger, Trevor Johnsen, Sam Nelsons, Emily Fortuna, Dave Elliott Smith, Mike McDonald Jr.
At a recent strategic discussion, Anthropic executive Mike McDonald Jr. outlined a paradigm shift where the distinction between AI-generated and human-created content becomes irrelevant, emphasizing instead that provenance via blockchain and compelling human storytelling will define future value. The organization is operationalizing this vision through a "bottoms-up" product philosophy and the development of the Model Context Protocol (MCP), which evolved from engineer-led integrations to establish a standardized, open framework for autonomous agent interactions and economic transactions. Internally, this approach has resulted in massive AI adoption with over 70% of code reviews generated by machines, though leaders note that non-technical organizational bottlenecks now pose the primary constraint on high-velocity development.