100 Years of Progress in 100 Days: Sequoia AI Ascent 2026 Keynote

The current AI wave is not an incremental step but a fundamental paradigm shift, distinct from previous technological advancements like the internet, cloud, and mobile. While past waves were additive and focused on the distribution of information—revolutions in communication—AI represents a revolution in computation, fundamentally changing how information is processed.

This wave is different in three key ways:

1. It’s the biggest yet, expanding beyond the traditional software market into a vast, multi-trillion dollar services sector. The market for legal services in the US alone rivals the entire global software TAM, illustrating the immense scale of the opportunity.
2. It’s the fastest wave, with the landscape of market leaders being populated at an unprecedented rate.
3. It’s a different shape, as a revolution in computation means the foundational technology layer is in constant flux, creating a dynamic and volatile environment.

The Arrival of Functional AGI

We’ve witnessed three major inflection points in recent years: the “ChatGPT moment” showcasing pre-training, the emergence of a second scaling law around inference-time compute, and most recently, the power of long-horizon agents. This third step represents a discontinuous leap forward. From a commercial and functional standpoint, this is AGI. If you can dispatch an agent to perform a job, and it can recover from failure and persist until the task is complete, it has achieved a practical form of general intelligence.

The “faster horses” of the last few years—tools that offered 10-40% productivity gains—are being replaced by “cars,” applications that deliver 10-40x improvements and fundamentally change the nature of work and organizations.

Strategy in a Computational Revolution: Get M.A.D.

For startups building on top of foundation models, the race is different. A winning strategy can be framed by the acronym M.A.D.: Moats, Affordance, and Diffusion.

• Moats: In a world where the technology changes daily, durable moats are built by being obsessively customer-back, not tech-out. Customer needs are far more stable than the underlying AI capabilities.
• Affordance: The most powerful tools are useless if inaccessible. The opportunity lies in creating brain-dead simple paths for specific customers to solve their problems, bridging the gap between a powerful terminal and a user’s desired outcome.
• Diffusion: The gap between the creation of new AI capabilities and their adoption by the average enterprise is constantly widening. This “diffusion gap” represents the core opportunity for application-layer companies.

The Age of Agents is Now

While early experiments with agents like AutoGPT in 2022 were “completely useless,” the underlying models have now matured. Today, agents are not just a future concept; they are here and they work. The key components of agency have all advanced significantly:

1. Reasoning & Planning: Models, the “brain” of the agent, can now sustain progress on complex, long-horizon tasks for hours instead of minutes.
2. Tools: The “arms and legs” of agents leverage the last two decades of software development, from terminals and IDEs to SaaS applications, which will explode in value as agent usage increases.
3. Harness: Persistence and the ability to iterate towards a goal are now being refined through reinforcement learning, creating a powerful feedback loop for improvement.

This is leading to a new “scale of agentness,” progressing from simple inline assistants to agentic teams, asynchronous background agents, and eventually “dark factories” where processes run without any human review. The core takeaway is that services is the new software. Agents can be “hired” to perform complex professional work in medicine, law, and science, scaling at a rate humans cannot match because they are paid in tokens, not salaries, and are infinitely scalable with compute.

The Coming Cognitive Revolution

This AI movement is best understood as a cognitive revolution, which will parallel the Industrial Revolution’s impact on physical work, but will be much bigger and faster. For millennia, human and animal muscle performed nearly all physical work; today, over 99% is done by machines. We are at the beginning of a similar transition where, in the near future, over 99% of cognition on Earth will be performed by machines.

Four stories help frame this future:

1. Aluminum as Intelligence: Aluminum was once the most precious metal on Earth, capping the Washington Monument. The invention of electrolysis made it so abundant it became disposable. AI is the electrolysis for intelligence, making PhD-level skills instantly invocable and universally accessible.
2. Alien Design: Human design is optimized for human cognition. When we hand over design to AI—as NASA did with an antenna—the results are often dramatically more effective but not intuitive to us. We must prepare for a world of “alien design” in chips, cars, and architecture.
3. Emerging Sciences: For over a century, engineers tinkered with combustion engines before the science of thermodynamics emerged to formalize the principles. We are currently in the “tinkering” phase of AI. A new, fundamental science will arise in the coming decades to help us master AI and even consciousness.
4. The Art of Unreason: When photography automated realism in art, painters didn’t stop painting. They responded with new forms like Impressionism and Cubism, focusing on capturing what the soul sees, not just the eye. As AI automates cognitive work, it forces humanity to focus on what gives that work meaning.

Ultimately, AI will do the work, but only human connection can provide a reason to care. In a world of abundant intelligence, the relationships we form and the meaning we create will be the things of enduring value.

• Tags:
• Foundation Models
• AGI
• Agents
• Cognitive Revolution
• Sequoia Capital