Vertical agents trained end-to-end on specific workflows using reinforcement learning and user data can outperform the best humans at those tasks, with evidence emerging in security (Expo agents outperforming penetration testers), DevOps (Traversal vs troubleshooters), and networking (Meter vs engineers).

factualpending

Speaker

Sonia

Evidence Quote

“Expo is now showing that they have an agent that can outperform human penetration testers. In DevOps, Traversal is showing that they can create an AI troubleshooter that is better than the best human troubleshooters. Um, same goes in networking with meter and networking engineers.”

Source

AI's Trillion-Dollar Opportunity: Sequoia AI Ascent 2025 Keynote— Sequoia Capital
Created: 8/11/2026, 7:13:14 AM

Your notes

Loading notes...