Keynote, Conference Presentation
The Golden Age of Software Engineering | Scott Wu, Cognition | RAISE Summit 2026
- Company Scope and Growth: Cognition, founded 2.5 years ago, operates globally (US, Europe, Latin America, Asia, Middle East) with major clients including Goldman Sachs, Mercedes-Benz, and Lowe's.
- Core Thesis: By 2026, all major corporations will be defined as software companies, necessitating massive engineering capacity to manage complex internal and external software systems.
- Productivity Explosion: In a six-month period (November to April), Cognition increased its software shipping volume by 10x while simultaneously growing its engineering headcount by only 40% (from ~45 to 60 engineers).
- Current Agent Capabilities: The median duration of autonomous AI tasks has shifted from ~30 seconds (chatbot era) to 16+ hours, enabling agents to own end-to-end projects rather than just assist with isolated queries.
- Model Saturation: Recent improvements across all major AI model providers have created a "saturation" point where multiple models are sufficiently capable to handle complex software engineering tasks without human intervention.
- Autonomous Loop Closure: Meaningful gains are achieved when agents autonomously execute the full SDLC lifecycle: understanding codebases, planning, writing code, testing, and reviewing, rather than just generating code snippets.
- Commitment Metrics: At Cognition, 90% of committed code is now generated by their AI engineer, "Devin."
- Proactive Execution: The majority of Devin sessions are now initiated asynchronously via automated triggers (e.g., API performance drops or Datadog error alerts) rather than interactive human requests.
- Parallel Workflows: Organizations can deploy agents to run parallel, event-driven workflows such as bulk CVE remediation, large-scale version upgrades, and immediate bug triage without human orchestration.
- Capacity Scaling: A 13% monthly increase in AI-assisted output in December 2025 is comparable to adding 130 engineers to a 1,000-person team, effectively creating a 10:1 human-to-AI engineering ratio.
- Metric Shift: The industry is transitioning away from "token-maxing" and lines-of-code metrics (which incentivize volume over quality) toward outcome-based business impact and project delivery efficiency.
- Human Impact: Human engineers are projected to become 10x more productive, with the primary demand shifting from "hiring more people" to "building 10x more software" (e.g., modernizing DMVs, medical record systems, and banking tools).
- Future Outlook: The "Golden Age of Software Engineering" is described as imminent, characterized by an abundance of software capacity where the primary constraint is identifying valuable problems to solve rather than available engineering bandwidth.