Podcast Summary: "How Devin Replaces Your Junior Engineers with Infinite AI Interns That Never Sleep"

Podcast: How I AI
Host: Claire Vo
Guest: Scott Wu (CEO of Cognition Labs)
Date: September 8, 2025

Episode Overview

This episode dives deep into how Cognition’s AI agent, Devin, is reinventing the way software engineering teams operate. Host Claire Vo sits with Scott Wu (CEO and founder, Cognition Labs) for a hands-on exploration of Devin—dubbed the ultimate “infinite intern”—and its transformative workflows. They discuss how Devin mimics a junior engineer, enabling asynchronous, multi-threaded productivity, and touch on best practices for task scoping, integrating AI into team culture, and even leveraging voice AI for collaborative meetings.

Key Discussion Points

Introducing Devin: The Infinite Intern

• Devin as Async Junior Engineer

  • Devin acts like a junior engineer: given a well-scoped task, it works independently and asynchronously.
  • “Devin's going to start working and looking through the code... it's just as if you gave your intern a project and your intern is going and working on it.” — Scott Wu [00:00]

• Multi-threading Tasks

  • Teams can spin up multiple Devin sessions in parallel, substantially increasing throughput without needing to “babysit” each task.
  • “You can multi-thread a lot with tools like this and set 2, 3, 4, 5, 10 of these going at once on different projects and not feel like you have to sit there and babysit things.” — Claire Vo [00:32, 13:01]

Scoping Work for AI

• Tasks vs. Problems

  • Devin excels at executing clear, specific tasks built from well-defined specs—its strength is in “tasks, not problems.”
  • “Devin is not going to go and solve some, you know, really hard architectural problem... But where Devin really shines... is kind of like tasks, not problems.” — Scott Wu [04:04]

• Using AI-Generated ‘Deep Wiki’

  • Cognition's Deep Wiki creates AI-generated documentation for any codebase. It's a foundation for both understanding context and refining tasks for Devin.
  • “You can come in and get a full AI generated documentation of the repo.” — Scott Wu [06:48]

• Prompt Refinement

  • Instead of sending five-word prompts, users are encouraged to gather context, build a richer prompt, and then hand off to Devin for asynchronous execution.
  • “Take this prompt and turn it into an effective prompt given the context... It feels like extra friction ... but I think pretty soon is one going to be the job to be done of the tool itself.” — Claire Vo [10:27]

The Async Engineering Workflow

• Synchronous Setup, Asynchronous Execution

  • The ideal workflow: research with Deep Wiki/search (sync), build the prompt, then let Devin work independently (async).
  • “If you had an intern... Often what you would actually do is sit down with them, talk it through for two minutes... then you kind of hand off there.” — Scott Wu [11:24]

• Freeing Up Human Engineers

  • While Devin runs, humans can focus on meetings or other tasks, then check back to review outcomes (e.g., pull requests, bug fixes, front-end changes) generated by Devin.

Operationalizing Devin in the Team

• Devin as First-Line Responder

  • Devin is tagged in Slack channels for new issues, crash monitoring, and even mundane engineering toil—becoming the “first person” on the task before escalating to a human.

• Collaborative, Public Workflows

  • Encourages using Devin in public channels rather than private DMs to increase AI adoption and cross-team learning.
  • “Hiding your AI use is kind of the worst thing you can do... So I say do it all in public.” — Claire Vo [24:56]
  • “Devin is a naturally multiplayer experience... you’ll often have a few different folks going back and forth... Devin is just one of the players in [the] thread.” — Scott Wu [19:19]

Top Use Cases for Devin

Scott Wu’s Top 5:

1. Front-End Fixes

   • “You tag Devin, you explain 'here's a screenshot, I want to make this button rounder'... and it'll go and do that.” [26:09]

2. Version Upgrades & Migrations

   • Automates tedious dependency updates and finds necessary changes across the repo.

3. Documentation Generation

   • Writes and maintains technical docs, including for DeepWiki itself.

4. Incident Response

   • Acts on PagerDuty alerts, investigates crashes, and generates incident reports—often before a human is notified.

5. Adding & Running Tests

   • Creates unit tests, runs them locally, and iterates until CI passes.

Claire Vo’s Additions:

• Rubber Duck Debugging
  • Devin is always available for troubleshooting and ideation, even during off-hours.

Making Voice AI Socially “Normal” in Meetings

• ChatGPT Voice as an “Extra Attendee”
  • Scott uses ChatGPT’s voice interface during meetings to surface info or answer questions instantly—making AI participation a shared, inclusive experience.
  • “I almost think of ChatGPT voice as a better Google. You can get an even faster answer ... fully synchronous, you can do it in the conversation.” — Scott Wu [32:36]
  • Claire: “If you flip it ... this is just another meeting participant that I'm putting into the room, it actually is more socially inclusive.” [35:15]

Future of AI Engineering Interfaces

• Form Factor Evolution

  • Scott envisions a future where the “agent” is the atomic interface, erasing the need for IDEs or even code:
    • "Tony Stark doesn't have a laptop. ... At some point, if you have your Jarvis plugged in... you’re just looking at your own product and saying, hey, let’s make this button rounder." [36:14]

• For Today: Start Where You Work

  • Slack/issue trackers for leads/managers; IDE extensions for IC engineers.

Tips for “Prompting” and Unblocking Devin

• When Devin Gets Stuck
  • Trace through Devin’s steps to see where it erred, provide missing info, and reinstruct—akin to pair debugging with a junior teammate.
  • “With an agent... you can go through and look through all the history... and then you understand, oh, Devin was missing the link to this page.” — Scott Wu [38:51]

Memorable Quotes & Moments

• “Devin's my favorite intern on my team and I have infinite of them.” — Claire Vo [00:10, 06:09]
• “Devin is a junior engineer. We're working on getting Devin to senior engineer—obviously, we'll get Devin the promotion!” — Scott Wu [04:04]
• “I DM Devin all the time. It's because I have no employees, no one to talk to. He's my only buddy.” — Claire Vo [22:30]
• “Hiding your AI use is kind of the worst thing you can do... do it all in public.” — Claire Vo [24:56]
• “The way that I like to say it is Tony Stark doesn't have a laptop. ...At some point, if you have your Jarvis plugged in... you're just looking at your own product.” — Scott Wu [36:14]

Notable Timestamps

• 00:00–00:32 — Introduction to Devin and async, multi-threaded workflows
• 06:09–08:40 — Devin’s strength in executing “tasks, not problems”
• 10:27–13:01 — Importance of context-rich prompts and async handoff
• 14:10–15:31 — Real-life async-use; checking in after meetings
• 18:17–22:30 — How Devin is institutionalized as the first responder
• 24:56–25:49 — The value of public, multiplayer AI collaboration
• 26:09–30:41 — Scott and Claire’s top Devin use cases
• 32:36–35:15 — Using ChatGPT voice in meetings, lowering social frictions
• 36:14–38:51 — The future interface of AI engineering and handling AI frustrations

Conclusion

Scott Wu demonstrates how Devin can act as an “infinite intern,” powering productivity by tackling well-scoped engineering tasks asynchronously. From automated bug fixes to documentation and incident response, Devin frees up human engineers for higher-order problem-solving. Integrating Devin publicly across teams enhances AI adoption and organizational learning; meanwhile, the embrace of voice AI hints at a future where human-computer collaboration is seamless, social, and increasingly natural.

Where to find them:

• Twitter: @Cognition, Devin AI
• Podcast and transcripts: How I AI