Sequoia's Pat Grady and Sonya Huang declare AGI has arrived — and issue a call to arms for founders

Summary

Sequoia partners Pat Grady and Sonya Huang have published a piece declaring that AGI has arrived, framing the announcement less as a philosophical milestone and more as a tactical signal to founders. Their core argument is that the technology is no longer the bottleneck — human agency and application are.

The Three Inflection Points

Grady maps the current moment against two earlier turning points. The first was ChatGPT, which delivered baseline world knowledge and instinctive judgment through pretraining. The second was o1 in late 2024, which introduced reasoning — the ability to work through complex problems with deeper conclusions. The third, and the one Sequoia identifies as the AGI threshold, is the emergence of long-horizon agents over the past several months, driven by tools like Claude Code and Opus 4.5. These systems can fail, recover, stay on task, and persist through to an outcome without continuous human supervision.

The practical implication, per Grady, is a shift from AI as a productivity tool to AI as a parallel workforce. A user can give an agent instructions, run multiple instances simultaneously, and return to completed work — not a progress report.

From Talkers to Doers

Huang frames the generational shift in founder pitches as moving from "talkers to doers." Early-wave AI companies, even strong ones, largely produced a more capable version of SaaS — faster interfaces over familiar workflows. The new generation is selling completed work. She cites Traversal, which automates the incident response process that typically takes engineering teams several hours to a full day — root cause analysis, trace review, and remediation — as an example of a product with no SaaS analog. It functions as a coworker, not a dashboard.

Harvey is cited as a company that bridges both paradigms: structured workflows for users who want control, and background agents that complete legal tasks — data room review, deposition analysis — and surface results only when finished.

Agent-Led Growth Replaces Product-Led Growth

Huang identifies a structural shift in enterprise go-to-market. The progression runs from sales-led to product-led to what she calls agent-led growth. Agents like Claude Code are already making infrastructure choices autonomously — recommending Supabase for databases, Vercel for hosting — and this pattern will extend across verticals. The implication for software vendors is significant: the agent becomes the procurement layer, and the best product for the agent's task may win over the best-marketed product. Moats shift from user interface stickiness to contextual depth and feedback loops built up over time.

Grady draws a direct parallel to the on-premises-to-cloud transition, arguing that incumbent SaaS companies face the same structural difficulty making this leap that on-prem vendors faced a decade ago. The design paradigm and the business model are both fundamentally different.

The AGI Resistance Argument

On the question of whether AGI will move measurable GDP — a concern raised by economist Tyler Cowen, who points to healthcare, nonprofits, and other sectors as AI-resistant — Huang concedes the digital-physical gap is real. She argues, however, that robotics is advancing fast enough to narrow it, pointing to Recursive Intelligence, a Sequoia-backed company founded by alumni of Google's AlphaChip team, which uses reinforcement learning to generate chip layouts that bear no resemblance to conventional Manhattan-grid designs.

Huang explicitly distances the Sequoia position from ASI or recursive self-improvement scenarios. The stake in the ground is not that AI is about to go vertical — it is that the capabilities available today are sufficient to build transformative companies, and waiting for a more dramatic signal is a mistake.

The Risk of Passivity

Both Grady and Huang express concern about a cohort of founders and young people who have concluded that AGI will solve major problems autonomously and that building now is therefore pointless. Huang argues that even if all research progress froze at current capability levels, the addressable value remains enormous and underleveraged. The Sequoia piece is, by their own description, a call to action directed at that audience.