How to Learn Programming and CS in the AI Hype Era – Interview with Prof Mark Mahoney [Podcast #215]

Why It Matters

Understanding the limits of LLMs ensures developers retain core problem‑solving skills while safely integrating AI assistance, a balance critical for education, hiring, and corporate risk management.

Key Takeaways

• LLMs excel at low‑stakes visualizations, not complex production code.
• Experienced developers must still plan and review LLM‑generated code.
• Relying solely on LLMs hampers debugging skills and resilience.
• Traditional learning builds competence; LLMs provide an infinitely patient tutor.
• Corporate policies may restrict AI tools due to liability concerns.

Summary

The Free Code Camp podcast features an interview with computer‑science professor Mark Mahoney, who built the Playback Press platform and has taught thousands of developers. He discusses how the surge of large language model (LLM) code generators fits into modern programming education and why the fundamentals of computer science remain essential. Mahoney emphasizes that LLMs shine for low‑stakes tasks—quick visualizations, simple simulations, or classroom demos—but they falter when software complexity, safety, or maintainability are at stake. He advises experienced developers to treat LLM output as a draft: request a plan, iterate on it, and manually verify placement of data structures to avoid hidden technical debt. He also notes practical concerns such as token costs, subscription fees, and corporate policies that ban AI tools over liability fears. A memorable quote from Mahoney is that an LLM can act as an "infinitely patient tutor," yet it cannot replace the nuanced guidance a human instructor provides. He recounts using Claude Code to generate a pull‑request flow animation, and a recent mishap where the model stored data in the global document object, forcing him to intervene and correct the design. The takeaway for learners and educators is clear: blend AI assistance with traditional, hands‑on coding practice. Students who master debugging and architectural reasoning without over‑reliance on AI will be more resilient and attractive to employers. Institutions can leverage LLMs for low‑risk teaching aids while preserving human mentorship, and businesses must weigh cost, liability, and talent development when adopting these tools.

Original Description

Today Quincy Larson interviews Mark Mahoney. He worked as a dev before becoming a computer science professor. He's taught computer science for 23 years at Carthage College, a 180-year-old US university. He's also taught thousands of developers through his free programming courses built on top of his own open source course platform, Playback Press.

We talk about:

- Why learning programming the hard way is still the right way

- How to not deskill yourself when programming with LLM tools

- And why now is a great time to study computer science

Support for this podcast comes from the 10,113 kind folks who donate to our charity each month. Join them and support our mission at https://donate.freecodecamp.org

Get a freeCodeCamp tshirt for $20 with free shipping anywhere in the US: https://shop.freecodecamp.org

Links from our discussion:

- Playback Press, Mark's free interactive courses: https://playbackpress.com/books

- Mark's personal website: https://markm208.github.io/

- One of the many vibe-coded projects Mark mentions: https://markm208.github.io/vibeCodingInClassTools/git-workflow-simulator.html

- Mark's tutorials on freeCodeCamp: https://www.freecodecamp.org/news/author/markm208/

Community news section:

1. freeCodeCamp just published a new course on AI-assisted software development. You'll learn common terminal workflows and tips for “pair programming” alongside LLM tools. You'll also get exposure to tools like GitHub Copilot, Claude Code, Gemini CLI, and OpenClaw. At the end of the day, the entire goal of using these tools is to build more features without compromising the maintainability of your codebase. (90 minute YouTube course): https://www.freecodecamp.org/news/ai-tools-for-developers/

2. freeCodeCamp also published a beginner level course on AI literacy for everybody that you can also share with your family. First you'll learn about the two traits that definte artificial intelligence: autonomy and adaptivity. Then you'll build your own image classifier right on your own phone or laptop. This course also delves into considerations like algorithmic bias the environmental costs of training and running LLM systems. (1 hour YouTube course): https://www.freecodecamp.org/news/ai-literacy-for-everybody/

3. Learn how to build your own QR code generator using JavaScript. This tutorial will walk you through generating QR codes entirely in a browser without the need for a backend. You'll learn how to validate input, clear previous output, and use a JavaScript library to render the code instantly on the client side. Then you'll see how to extend the project with downloads, custom styling, WiFi support, and more. (7 minute read): https://www.freecodecamp.org/news/how-to-build-a-qr-code-generator-using-javascript/

4. I'm thrilled to announce that the Bad Website Club is back for another Responsive Web Design bootcamp based on freeCodeCamp's curriculum. It starts April 24 and runs for 10 weeks. You can join their Discord community and tune in for live streams. It's lead by volunteer devs who are passionate about helping folks learn CSS and JavaScript fundamentals. (5 minute read): https://www.freecodecamp.org/news/bad-website-club-bootcamp-based-on-freecodecamp-rwd-cert/

5. Today's song of the week is 2008's Strange Overtones. The Talking Heads singer David Byrne blends his voice with Brian Enos, who handles organs and synths. The entire affair plays over an infectious palm-muted guitar line, and driving bass. This is a perfect mid-week jam. Put it on during during your commute. https://www.youtube.com/watch?v=GvZhpCYWFzs