Latest Interviews
Showing 1–15 of 49 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 Capital17 min
How Companies Are Building Their Own Intelligence | Sonya Huang, Sequoia Capital
Approximately 80 portfolio company founders and AI leaders convened to strategize the adoption of Sovereign AI, a framework defined by vertical integration where organizations own model weights rather than relying on external APIs. The event combined high-level market analysis with technical workshops led by industry experts to outline a four-step roadmap for building custom intelligence capabilities. Participants explored critical architectural decisions regarding cost efficiency, latency reduction, and the necessity of dedicated research labs to leverage open-weight models for proprietary domain performance.
- Sequoia Capital1 min
Every CIO will have to answer for every token | Factory's Matan Grinberg
The factory router introduces a dynamic model selection strategy that optimizes enterprise AI efficiency by routing distinct tasks to specialized, cost-effective models rather than relying on a single frontier solution. This framework enables CIOs to address the critical need for granular token allocation, assigning specific capabilities such as lightweight generation for "vibe coding" or custom fine-tuned models for legacy COBOL maintenance. By matching model performance to precise organizational workflows, the approach allows for complex multi-model validation pipelines while avoiding the inefficiency of blanket usage caps.
- Sequoia Capital1 min
90% of AI tokens will be asynchronous | Matan Grinberg, Factory
The event outlines a critical shift in AI consumption from fragile, prompt-dependent synchronous models to robust asynchronous workflows where autonomous agents initiate tasks without human triggers. Industry forecasts predict that within 12 to 24 months, 90% of tokens will be generated by these autonomous systems, marking the transition from current "co-pilot" tools to fully "agent-native" operations. This evolution promises to decouple revenue growth from active human intervention by enabling systems to independently identify customer signals and execute first-pass solutions.
- Sequoia Capital2 min
Why "Tokens Aren't Fungible" - Anthropic's Angela Jiang
The organization is transitioning its strategic focus from knowledge retrieval to an execution layer powered by Cloud Managed Agents, which enables AI systems to string together tasks and edit files across multiple systems. This infrastructure serves as the foundation for a forthcoming coordination layer that will introduce strategies as a meta harness to orchestrate complex workflows through specialized token roles. The roadmap outlines a sequential evolution moving from the current execution capabilities toward this higher-level abstraction where high-level intent guides composed, autonomous systems.
- 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 Capital1h 3m
Notion’s Ivan Zhao: The Refounder
Ivan Zhao, Jack Dorsey, Brian Armstrong, B Halligan
Notion CEO Ivan Jawan has steered the company through two strategic refoundings, most recently pivoting to an AI-native model that replaces traditional hierarchies with a fluid "jazz band" structure relying on self-managed teams and AI as the central information processor. This transformation involves restructuring hiring to prioritize individual agency over experience, merging product roles to leverage AI agents, and shifting compensation toward a high-meritocracy "wartime" model. By treating product development as non-deterministic experimentation rather than rigid planning, Notion aims to scale effectively while maintaining a culture that functions as a shared belief system among its fifty to sixty acquired founder-employees.
- Sequoia Capital11 min
AI That Designs Its Own Chips: Ricursive's Anna Goldie and Azalia Mirhoseini
Anna Goldie, Azalia Mirhoseini
Founded by former leaders from Google Brain, Anthropic, and DeepMind, Recursive Intelligence has deployed its AlphaChip technology to optimize billions of transistors for Google's TPU and Axion lines while adopting by MediaTek. The company currently accelerates chip design by running static timing analysis 1,000 times faster than commercial tools, enabling AI agents to generate curved, organic layouts that reduce wire lengths and cut annual design cycles from months to days. Looking ahead, Recursive plans to democratize custom silicon through a platform model that delivers fabrication-ready layouts without in-house teams, ultimately aiming to vertically integrate design and fabrication into a self-reinforcing loop for frontier AI systems.
- Sequoia Capital9 min
Inside the Rise of Autonomous AI Hackers: XBOW's Oege de Moor
The presentation argues that the cybersecurity arms race has shifted to autonomous AI attacks, exemplified by an AI agent named XBO that recently achieved global dominance on HackerOne by discovering critical Microsoft Bing vulnerabilities through black-box testing. Because current defensive tools often fail to verify exploitability in live environments, the speaker urges organizations to immediately integrate autonomous AI into their workflows to counter negative exploit velocity before open-weight models close the capability gap within six to nine months. Ultimately, the event posits that future security success depends entirely on adopting AI-driven offensive and defensive systems rather than relying on traditional human-only methods.
- Sequoia Capital14 min
Why the Brain Computes 1,000,000x More Efficiently Than A GPU: Unconventional AI's Naveen Rao
Unconventional AI CEO Navin Rao is deploying a prototype that replaces traditional von Neumann architectures with nonlinear dynamical systems to overcome the impending energy saturation of current AI infrastructure. By leveraging Kuramoto synchronization models, the startup achieved functional generative capabilities in six months while demonstrating energy efficiency comparable to biological neural networks. This physics-based approach aims to bypass the thermodynamic limits of digital lithography, offering a viable pathway to artificial general intelligence within strict global power constraints.
- Sequoia Capital12 min
Starcloud's Philip Johnston: Why the Cheapest Compute Will Be in Space
StarCloud CEO Philip Johnston validated the technical feasibility of space-based high-performance computing through the successful "StarCloud 1" mission, which demonstrated thermal management and radiation tolerance while executing AI inference tasks. The company has filed an FCC application for an 88,000-satellite constellation capable of delivering 20 gigawatts of compute power with sub-50-millisecond latency, targeting a $100 billion capital expenditure that becomes economically viable once launch costs drop below $500 per kilogram. While current operations focus on inference workloads, the roadmap envisions future large-scale training structures that could catalyze a transition toward a Kardashev Type 2 civilization within decades.