No Priors• May 2, 2024• 29:23Interview

No Priors Ep. 62 | With Cognition CEO and Co-Founder Scott Wu

From No Priors

Scott Wu•CEO & Co-Founder, Cognition

Executive Summary

This episode features an interview with Scott Wu, CEO of Cognition, the AI lab behind Devon, an AI software engineer.
Wu discusses Devon's capabilities, including its ability to handle end-to-end software tasks autonomously and its leading performance on the SWE-bench benchmark.
He explains that Devon is designed as a collaborative tool, with a UI that allows for human oversight and feedback.
Wu predicts that AI agents will augment, not replace, engineers, shifting their role towards high-level problem-solving and architecture, ultimately increasing productivity and democratizing software development.

12 quotes

Concerns Raised

Public misconception that AI will eliminate all software engineering jobs.
Skepticism about the current capabilities of AI agents.

Opportunities Identified

Massively increasing the productivity of existing software engineers.
Democratizing software development for a wider audience.
Solving complex DevOps and data analysis tasks that are tedious for humans.

Key Themes

The Emergence of AI Software Engineers

Cognition launched Devon, an AI agent capable of handling complete software engineering tasks from planning to debugging. This represents a step-change from simple code completion tools to autonomous agents that can reason and use tools like a human.

This signals a major shift in software development, where AI agents become active teammates rather than passive tools.

Human-in-the-Loop AI Collaboration

Devon's UI is intentionally designed to be interactive, allowing a human to observe its plan, code, and browser actions. This paradigm of 'looking over the shoulder' enables course correction and makes the agent more effective than a black-box system.

Effective AI agent design requires transparent and interactive user experiences to build trust and handle complex, multi-step tasks.

The Evolving Role of the Software Engineer

Scott Wu predicts that AI will not eliminate engineering jobs but will instead elevate the role. Engineers will spend less time on implementation and more on creative problem-solving, system architecture, and product definition, akin to a hybrid technical architect and product manager.

Professionals should focus on developing skills in system design and problem formulation, as implementation details will be increasingly automated.

The Future of Human-Computer Interaction

The development of agents like Devon points to a future where natural language (like English) becomes the primary way to instruct computers. This will make complex software creation more accessible and change the fundamental definition of 'coding'.

The barrier to creating software will decrease, unlocking latent demand and enabling more people to build custom tools and applications.

Get started free

Topics

Cognition Devon AI Agents Software Engineering Future of Work AI-assisted Development SWE-bench Reasoning Autonomous Agents DevOps Automation Human-Computer Interaction Competitive Programming

Processed Feb 25, 2026 yt-dlp + mlx-whisper + Gemini