Is AI Coding Making Developers Obsolete?

TL;DR

AI coding is not making developers obsolete, but overreliance can weaken their skills and disconnect them from the codebase. The speaker sees major productivity gains, even citing a CEO producing 37,000 lines of code a day, and finds AI especially effective for tedious technical-debt work. Yet repeatedly rerunning agents when something fails creates cognitive debt, unpredictable results, and added costs, making the tradeoffs worth examining closely.

Transcript

AI code is here and it's here to stay. Who hasn't seen the productivity gains? We're talking CEOs of companies doing 37,000 lines of code every single day. I mean, what else is there to our lives as developers? If we can put out that much code, why would we do anything else? This totally isn't going to come back to bite our asses, right? It's not l... Read More

Key Insights

• AI coding tools have changed how developers approach coding, often reducing the time spent writing code directly.
• There is a growing concern about skill atrophy as developers rely more on AI to handle coding tasks.
• AI tools can lead to cognitive debt, where developers lose the ability to understand and manage the codebase effectively.
• The non-deterministic nature of AI increases complexity and potential vendor lock-in for coding teams.
• Effective use of AI in coding requires a balance between automation and maintaining fundamental coding skills.
• AI tools can accelerate productivity but may not always prioritize code quality and understanding.
• Developers should use AI to enhance their understanding and problem-solving capabilities, not replace them.
• Vendor lock-in and fluctuating costs are concerns when relying heavily on specific AI coding tools.

Questions & Answers

Q: Is AI coding making developers obsolete?

The speaker does not argue that developers are becoming obsolete. Instead, the concern is that developers may lose skills and understanding when they spend less time writing or inspecting code and repeatedly ask AI to try again when something fails.

Q: How does AI coding affect developer productivity?

AI coding can produce dramatic output gains, with the transcript citing a company CEO generating 37,000 lines of code every day. It can also handle labor-intensive maintenance tasks that previously required substantial engineering time.

Q: What is cognitive debt in AI coding?

Cognitive debt is the growing disconnection between developers and the codebase as AI performs more of the work. The speaker describes barely looking at code anymore and sometimes rerunning an agent repeatedly instead of understanding why something failed.

Q: Can relying on AI coding tools cause skill atrophy?

Yes, the speaker personally feels some skills atrophying while spending less time writing code. When an output fails, repeatedly pulling the metaphorical slot-machine lever can replace deliberate debugging and deeper engagement with the code.

Q: What development skills can increase when using AI agents?

The speaker says AI-assisted work has led to more experience with Git, SSH, and remote machines. It has also encouraged experimentation with new technologies, showing that AI can shift skill development rather than simply eliminate it.

Q: What coding tasks is AI especially good at?

AI is described as particularly effective at tedious technical-debt work. Examples include applying a new lint rule across an entire codebase or updating a package change affecting 5,000 locations.

Q: How did developers handle large migrations before AI coding tools?

The speaker recalls a Twitch GraphQL migration that broke roughly 8,000 TypeScript files. The broken files were placed in a Google Sheet and assigned in chunks to engineers for manual updates on a shared branch.

Q: What are the risks of repeatedly rerunning AI coding agents?

The results are non-deterministic, so developers do not know what each new run will produce. Repeated attempts also cost money and may leave developers with less understanding of why the code failed or how it was fixed.

Summary & Key Takeaways

• AI coding tools provide significant productivity improvements but can lead to skill atrophy among developers. As developers rely more on AI, they risk losing critical thinking and coding skills, making them less effective in the long run. Balancing AI use with active coding engagement is essential to maintain and enhance developer skills.

• AI coding tools can create cognitive debt, where developers become disconnected from the codebase and lose the ability to manage it effectively. The non-deterministic nature of AI increases complexity and potential vendor lock-in, posing challenges for development teams.

• Developers should leverage AI to enhance their understanding and problem-solving abilities, not replace them. Effective use of AI requires balancing automation with maintaining fundamental coding skills, ensuring AI tools contribute to better quality work without sacrificing essential skills.