Raw Content
Cognition’s CEO on What Comes After Code
Overview
This podcast episode features Scott Wu, CEO of Cognition (maker of the autonomous coding agent Devin), discussing the future of software engineering with Dan Shipper.
Key Themes
Computer Science Fundamentals Still Matter
Wu argues that despite AI coding tools becoming prevalent, core CS knowledge remains essential. He compares AI tools to calculators in math education—they shift what skills matter rather than eliminating their necessity. As Wu explains, “the reason for that is because you’re still the one at the helm making the decisions.”
The Future: Programming by Description
Wu envisions a steady climb toward higher abstraction levels. Just as programming evolved from assembly to Python, the next step involves describing what you want built in plain English, with AI agents handling implementation. He likens this to Iron Man’s relationship with J.A.R.V.I.S.—”you’re not looking at your code, you’re just looking at your own product.”
Near-Term Reality: A Spectrum of Tools
Currently, engineers need both synchronous tools (tab completion, code chat) and asynchronous systems (agents like Devin running independently). Wu states this hybrid approach will persist “until agents are capable enough to handle everything.”
The AI-First Workflow
Rather than jumping straight to coding, engineers should first map out plans with AI assistance—discussing user groups, UI behavior, and onboarding flows conversationally. Once detailed, AI handles implementation, with human review following.
Devin vs. Claude Code
Devin operates in dedicated cloud environments like “the software engineer sitting next to you,” while Claude Code gives “the reins over to your AI buddy” for direct computer control.
AGI Has Arrived
Wu argues we’ve already met decade-old AGI benchmarks: models pass Turing tests, win elite math competitions, and operate through reasoning agents.