DOP 313: Harnessing AI for Smarter Development

Episode 313

Show Notes

#313: In this episode, Darin shares his recent experiences using AI tools Cursor and Claude Code to improve and refactor Jenkins plugins. After receiving a recommendation to try out Cursor for code improvements, he tests it alongside Claude Code, comparing their functionalities and effectiveness. He describes his process and observations, noting that both tools helped identify performance improvements in the code. While Cursor provided quick initial feedback, Claude Code offered a slightly better quality of suggestions but required nudging to get accurate results. Darin also mentions the practicality of integrating these tools with his existing setups and the importance of having issues documented for better management. Moreover, he discusses the benefits of AI-assisted PR descriptions and emphasizes the need for caution when using such tools for proprietary code without corporate approval. Overall, he concludes that transitioning to these advanced AI tools can significantly improve productivity in open-source projects.

Frequently Asked Questions

How do Cursor and Claude Code compare for refactoring work?

Darin Pope ran the same questions through both on DevOps Paradox episode 313, against Jenkins plugins he helps maintain. Both found the significant problem immediately and both needed nudging toward the right fix. Claude Code surfaced a few extra items, which he attributes to the model behind it. Viktor Farcic points out Cursor’s default setting picks the model for you and leans heavily toward the cheapest option.

Should you point an AI coding tool at your employer's code?

Darin Pope is emphatic on DevOps Paradox episode 313 that you get approval first, because the code leaves your machine. He does none of this against proprietary work. His suggestion for anyone wanting to learn without that constraint is open source, where a public repository has almost certainly been ingested already, so there is nothing left to leak and no compliance conversation to have.

Why write a detailed pull request description if AI generated the code?

Viktor Farcic gives a second reason on DevOps Paradox episode 313 beyond helping the human reviewer. Start a fresh session weeks later with no memory of the work, point the agent at an issue in that pull request, and the description tells it exactly what was going on. He suggests generating those descriptions regardless of whether a person or an agent opened the request.

Do AI coding agents agree with you too easily?

Viktor Farcic suggests a test on DevOps Paradox episode 313: tell the agent something obviously wrong, such as rewriting the whole plugin in a different language being a five minute job, and watch it agree that you are correct. Darin Pope had already noticed the pattern, describing how it thanked him for corrections it should have caught itself on a null check and a missing final.

Should you record AI findings as issues when working alone?

Viktor Farcic answers with a question on DevOps Paradox episode 313: without AI, having found six performance problems yourself, what would you do? Darin Pope concedes he would write them down somewhere, which makes the answer the same either way. The cost of not doing it landed immediately, when exiting the session lost the context and re-asking produced a different list.

What should you let a coding agent do without asking?

Viktor Farcic’s rule on DevOps Paradox episode 313 is to approve anything exploratory for the whole project, such as listing files or reading cluster state, and to review every change. Darin Pope found the interruption useful for a different reason: pausing on each edit gave him time to notice one he wanted to revisit, where a batch of five arriving at once would have hidden it.

What is the DevOps Paradox podcast?

DevOps Paradox is a weekly podcast co-hosted by Darin Pope and Viktor Farcic, covering DevOps, platform engineering, and modern software delivery. Episode 313 follows Darin Pope’s first day with Cursor and Claude Code on real open source code, covering what each found, what needed correcting, and which habits carry over from working without them. Every episode page carries the audio, the video, and a full transcript.

Episode Transcript

Read the Full Transcript

Topics

Share and Download

Hosts

Darin Pope

Darin Pope

Viktor Farcic

Viktor Farcic

Viktor Farcic is a member of the Google Developer Experts and Docker Captains groups, and published author.

His big passions are DevOps, Containers, Kubernetes, Microservices, Continuous Integration, Delivery and Deployment (CI/CD) and Test-Driven Development (TDD).

He often speaks at community gatherings and conferences.

He has published DevOps Paradox and Test-Driven Java Development.

His random thoughts and tutorials can be found in his blog The DevOps Toolkit.