How Should Junior Engineers Use Claude Code? With Cat Wu

Why It Matters

Empowering junior engineers with QuadCode shortens onboarding, improves code reliability, and eases senior reviewers' workload, driving faster, higher‑quality product development.

Key Takeaways

• Use QuadCode for quick codebase orientation and on‑demand queries.
• Validate AI answers with seniors and update QuadMD documentation accordingly.
• Begin with low‑risk “paper‑cut” fixes to build tool fluency.
• Progress to larger changes once confidence in QuadCode’s output grows.
• Teach QuadCode testing routines via /verify commands to reduce review load.

Summary

The discussion centers on how junior engineers can effectively adopt Claude’s QuadCode tool to accelerate onboarding and contribute meaningfully to a codebase. The speaker emphasizes starting with basic navigation—asking the AI any “dumb” questions to map the terrain, then cross‑checking its responses with senior engineers and recording corrections in the QuadMD for future users. Key practices include validating AI‑generated answers, updating documentation, and tackling low‑risk "paper‑cut" bugs as an initial hands‑on exercise. These small fixes let newcomers gauge QuadCode’s strengths and limitations while minimizing exposure to critical failures. As confidence builds, engineers can graduate to more substantial changes, always ensuring thorough testing before submitting pull requests. A recurring metaphor treats QuadCode as an intern that needs mentorship: senior engineers review AI PRs, juniors add /verify commands, and the team continuously refines the agent’s knowledge base. Examples such as fixing minor UI glitches and embedding verification steps illustrate a feedback loop that improves both the developer’s skill set and the AI’s accuracy. The approach promises faster ramp‑up times, higher code quality, and reduced review overhead for senior staff. By institutionalizing AI‑assisted learning and documentation, organizations can scale engineering productivity while keeping the AI model aligned with real‑world code standards.

Original Description

There’s a lot of talk about AI coding tools like Claude Code replacing junior engineers, but they can also help less-experienced developers deepen their understanding of codebases and processes. Here’s Anthropic’s Cat Wu from AI Codecon on a few ways juniors should use these tools today.

First, ask Claude Code all your “dumb” questions. As Cat points out, it doesn’t care—“it's very happy to answer any time of the day”...unlike your coworkers. And if you think Claude Code is wrong (and it might be), take a screenshot and follow up with a senior engineer to make sure you understand all the nuances. Then, most importantly, use that feedback to update the CLAUDE.md file so it gets it right next time. You’ve just improved the codebase for everyone.

Once you’re more comfortable, you can start fixing smaller-stakes, lower-risk “paper cut” problems to really get a feel for what Claude Code is good at and what it’s not. Finally, you can really show your initiative and support your senior engineers by using Claude Code to thoroughly test your AI PRs before submitting them, and even teaching it how to test code as part of the creation process. As Cat suggests, “Think of Claude Code as your intern that you're trying to level up.”

Follow O'Reilly on:

LinkedIn: https://www.linkedin.com/company/oreilly/

Facebook: http://facebook.com/OReilly

Instagram: https://www.instagram.com/oreillymedia

BlueSky: https://bsky.app/profile/oreilly.bsky.social