Two hours with Cursor changed my mind about AI coding
8 days ago
- #AI Coding Tools
- #Software Engineering
- #Productivity
- The author initially resisted AI coding tools due to skepticism about their effectiveness and a love for traditional coding.
- Early experiences with GitHub Copilot reinforced skepticism, as it disrupted workflow and reduced productivity.
- Trying Cursor in agent mode for two hours changed the author's perspective, as it handled complex codebases and generated relevant tests.
- Cursor's ability to understand feedback and correct mistakes impressed the author, leading to daily use of both Cursor and Claude Code.
- Writing precise prompts for AI tools is likened to writing code, requiring patterns, good practices, and testing.
- The author maintains a document of good practices for AI tools, sharing them with colleagues due to the lack of up-to-date resources.
- AI tools are seen as valuable for delegating systematic, time-consuming tasks, such as migrating UI libraries, while still requiring expert oversight.
- The author now enjoys interacting with AI tools like Cursor and Claude Code, seeing them as enhancing rather than replacing the craft of coding.
- Uncertainty remains about the future of software engineering, but the author is excited and committed to evolving with the craft.