newsfilter
← Interviews
Sequoia Capital32 min

OpenAI’s Sam Altman on Building the ‘Core AI Subscription’ for Your Life

Sam Altman, Jens Nordvigsen, DAN GALPIN, SAM SACCONE

Conference Presentation, Fireside Chat, Interview, Keynote

Historical Context & Product Origins

  • OpenAI began in 2016 in its first office with a team of approximately 14 people, operating as a research lab without a concrete product or business plan.
  • The first consumer-facing product was DALL-E, not ChatGPT; the initial commercial offering was the GPT-3 API released in June 2020.
  • The decision to pursue a business model was driven by the need to fund "billion-dollar" scale models, as pure research could not sustain the required investment.
  • ChatGPT launched on November 30, 2022, emerging from user demand in the GPT-3 playground despite the model originally being "terrible at chat."
  • As of the current timeframe, ChatGPT has 500 million weekly active users.

Strategy & Operational Philosophy

  • Growth Velocity: Sam Altman asserts that large companies fail when they scale headcount without scaling output ("shipping"), leading to internal friction and low impact.
  • Core Strategy: OpenAI aims to become the "core AI subscription" and a "personalized AI" operating system, rather than just an API provider.
  • Platform Vision: Future infrastructure will likely evolve into a federated protocol (analogous to HTTP) allowing agents to exchange tools, authentication, and payments seamlessly.
  • Talent Focus: OpenAI prioritizes small teams with massive responsibility to ensure researchers and engineers remain high-impact and busy, avoiding "molasses" management styles.
  • Strategic Approach: The company rejects working "backwards" from a complex master plan, preferring to navigate the next 1–2 steps with adaptability as conditions change.

Market Dynamics & Competition

  • Legacy Enterprises: Altman predicts large companies will face "another couple of years of fighting" against AI disruption before a "last-minute scramble" and inevitable capitulation.
  • Generational Divide: A stark contrast exists between older users who treat AI as a "Google replacement" and younger users (20s–30s) who use it as a "life advisor" or "operating system."
  • Innovation Gap: Startups continue to "beat the crap out of larger ones" due to organizational rigidity and slow security/IT decision cycles in established firms.

Product Roadmap & Technical Focus

  • Coding & Agents: Coding is identified as a central, not vertical, capability essential for moving from "assistants" to "agents" that can actuate the physical world via code.
  • Voice Interaction: Voice is a top priority but is currently in a "work in progress" phase; the goal is to enable simultaneous voice and graphical user interface (GUI) interaction.
  • Sensor Data: Models are increasingly integrating physical sensor data, with plans to bake this capability more explicitly into future architectures.
  • Customization vs. Generalization: The long-term "platonic ideal" is a single, untrained model with massive context windows containing all user data, rather than specialized fine-tuned models for every use case.
  • Upcoming Milestones:
    • 2025: Projected as a year dominated by AI agents performing work, particularly in coding.
    • 2026–2027: Anticipated shift from intellectual AI tasks to significant value creation in the physical world, including robotics.

Research & Organizational Resilience

  • Research Leverage: Altman identifies "big algorithmic breakthroughs" as the highest leverage factor, estimating that a few 10x or 100x improvements remain in the roadmap.
  • Academic Collaboration: 90% of OpenAI's focus is on improving general model capabilities to serve the academic sector broadly, though custom research partnerships exist.
  • Founder Resilience Advice: Altman notes that while acute crises are manageable with adrenaline and support, the "fallout period" (Day 60+) is psychologically harder and requires specific practice in rebuilding.
  • Future Model Speculation: Altman expresses unwavering confidence in the research roadmap, noting GPT-5 is a primary focus but remains part of a continuous improvement cycle beyond current capabilities like O3.