Interview, Fireside Chat, Conference Presentation
The Future of Software Engineering: Cognition’s Russell Kaplan
- Capabilities in formally verifiable domains like coding and math are projected to expand rapidly, with AI agent complexity doubling every seven months, a rate exceeding Moore's Law.
- A paradigm shift from Q&A to autonomous action is expected to drive utility, where every software engineer manages a fleet of AI agents, potentially elevating individuals to tech lead or manager roles while non-coding skills become more prominent.
- Success in the AI agent era will correlate with the ability to decompose complex projects into discrete units, eventually enabling agents to solve all technical classes, though full autonomy for all tasks is not yet realized.
- Software creation will broaden to a wider population by reducing reliance on complex syntax, while agentic capabilities may extend beyond code into policy and process execution, with reasoning becoming more critical alongside documentation.
- Cognition aims to reach senior engineer proficiency for its product "Devin" within the next 12 months and plans to acquire Windsurf to enhance high-detail, human-in-the-loop workflows.
- Strategic focus will remain on integrating with large enterprises to address proprietary codebase challenges before expanding to other sectors.
- The industry faces anticipated energy constraints in the United States, with serious bottlenecks regarding power production expected to emerge approximately two years from now.