How I Code with AI Using Cursor for Planning, Execution, and Reviews

TL;DR

AI can make experienced software engineers more productive by supporting project planning, execution, code reviews, and repetitive coding tasks. The creator explains how his Cursor workflow has changed almost two years after his 2023 Copilot video, including using planning for both small and large projects. The approach requires understanding the generated code and acknowledges investment-related biases, so read on for the workflow’s practical boundaries and techniques.

Transcript

Remember that video I did about how I was blown away with Copilot all the way back in like 2023? It's been almost 2 years since that video was published, and the way I use AI to write code has changed more than I ever would have imagined. I'm still on Cursor, but the ways I use it have changed so fundamentally that I wanted to try and break down wh... Read More

Key Insights

• AI tools like Copilot and Cursor have evolved, changing how developers write and review code.
• Integrating AI into coding workflows can increase productivity and make engineering tasks more enjoyable.
• AI coding tools are not a substitute for understanding code; they are best used by experienced developers to enhance productivity.
• Using AI for planning and execution allows developers to focus on higher-level tasks, while AI handles repetitive coding tasks.
• AI models can provide feedback and revisions, improving the quality of code through iterative enhancements.
• AI can help secure services by simplifying complex tasks like rate limiting and endpoint protection.
• Developers should use AI tools to complement their skills, not as a crutch for tasks they do not understand.
• AI tools are most effective when they are integrated into a developer's existing workflow and used for tasks that align with their strengths.

Questions & Answers

Q: How does an experienced software engineer use AI for coding?

The creator integrates AI into planning, execution, code reviews, editor work, Git, and work trees. He presents these tools as a way for an existing productive software engineer to build more efficiently and enjoy engineering work more.

Q: Which AI coding tool does the creator currently use?

He still uses Cursor, although his way of using it has changed fundamentally since his earlier Copilot video. He also notes that Cursor had recently caused problems and that he was working closely with its team to fix bugs.

Q: Is this AI coding workflow suitable for beginners?

No. The creator explicitly says this is not a vibe-coding tutorial or a method for writing code without understanding it. He advises people who cannot yet code to learn coding first and return afterward.

Q: How does planning help when coding with AI?

The creator says AI works especially well for greenfield and small projects. It can also work well for large projects, particularly when planning is used properly.

Q: How does the creator demonstrate his AI development workflow?

He starts with a new project to show how he uses AI to scaffold, plan, and build from the beginning. His example is a tool for comparing writing quality and styles across models, with one model potentially giving feedback to another model’s draft.

Q: How has AI changed the creator’s development process?

His workflow changed substantially in the almost two years following his 2023 Copilot video. He now incorporates AI across daily coding tasks, including planning, implementation, reviews, editor work, work trees, and Git.

Q: How can ArcJet protect an expensive AI-backed endpoint?

The demonstrated setup installs ArcJet’s package, adds an API key, and calls aj.protect with the request to receive a decision and its reason. The creator also shows a token bucket configured with a mode, user-identifying characteristics, refill rate, interval, and capacity for rate limiting.

Q: Did the featured AI coding tools pay for the video?

The creator says none of the tools discussed paid him to make the video, though ArcJet sponsored a segment. He also discloses that he invested money in several companies mentioned and asks viewers to account for those biases.

Summary & Key Takeaways

• AI tools have transformed the coding landscape by enabling developers to automate mundane tasks and focus on creative aspects. These tools, such as Copilot and Cursor, have evolved significantly, offering enhanced productivity and enjoyment in coding. However, they are not substitutes for foundational coding skills and are best used by experienced developers.

• Integrating AI into coding workflows involves using models for planning, execution, and code reviews. This integration can streamline the development process, allowing developers to focus on higher-level tasks. AI tools can also enhance security by simplifying complex tasks like rate limiting and endpoint protection.

• AI tools should be used to complement a developer's existing skills, providing support for repetitive and time-consuming tasks. They are most effective when used by those who understand coding fundamentals, as they enhance rather than replace these skills. Developers can leverage AI to improve productivity and code quality through iterative enhancements.