Keynote, Conference Presentation
The Golden Age of Software Engineering | Scott Wu, Cognition | RAISE Summit 2026
- All organizations, including firms like Goldman Sachs, Mercedes-Benz, and Lowe's, are projected to be classified as "software companies" by 2026 as software capabilities become ubiquitous across sectors.
- AI coding adoption is accelerating rapidly, with Cognition's software shipping volume already increasing 10-fold between November and the end of April, a trend expected to continue.
- The average autonomous task duration for AI is predicted to grow beyond the current 16-hour maximum, enabling agents to function continuously for multiple days without human intervention.
- The industry is shifting from utilizing AI as a 30-second chatbot assistant to employing full software engineers capable of owning tasks end-to-end, including codebase analysis, planning, writing, testing, and review.
- Productivity gains are anticipated to be substantial, with teams effectively achieving a 10x capacity multiplier, meaning 1,000 human engineers could operate with the capacity of 10,000.
- Specific productivity metrics, such as a 13% increase observed in December 2025, are expected to be replicated or exceeded by organizations that effectively integrate AI agents.
- Cognition's Devon agents are forecasted to contribute 90% or more to code commits and initiate the majority of development sessions autonomously.
- The "token-maxing era," characterized by measuring input metrics like token usage or lines of code, is projected to have lasted approximately six months from January to May and is expected to end in favor of measuring real business outcomes and end-to-end delivery.
- Major AI model labs are expected to reach a saturation point where their flagship models are all "good enough" to perform specific tasks, shifting the competitive focus to how agents "close the loop" on every step.
- Operational models are expected to transition to parallel and event-driven structures, allowing agents to autonomously handle bulk CVE remediation, major version upgrades, and massive codebase refactors via automated triggers.
- Human triage for customer reports is expected to be bypassed, with agents proceeding directly to solving issues and generating pull requests for diverse scenarios such as monitoring API routes and responding to Datadog errors.
- Human software engineers are projected to become ten times more productive while simultaneously facing ten times more work due to the expanded scope of software needs in under-served areas like Department of Motor Vehicles, medical records, and banking platforms.
- The "golden age of software engineering" is described as currently underway, contingent on the correct implementation of AI coding tools to enable teams to deliver 10 times the previous volume of work.