Getting This Site Online

I first wrote this article while getting the site online and experimenting with several IDE assistants. The tools have changed considerably since then. I now use Cursor as my primary coding editor, supported by ChatGPT and other tools when I want a second perspective on a design or implementation.

The most useful IDE assistants understand more than the file currently open. Codebase awareness, project-specific instructions, fast completion, and natural-language editing all reduce the amount of context I have to reconstruct manually while working.

Cursor has become the best fit for that workflow. I use its autocomplete for small local changes, codebase-aware conversations for exploration, and agent workflows for larger tasks that need searching, editing, and validation. I still review the resulting diff and run the relevant tests rather than treating generated code as automatically correct.

I also use ChatGPT for brainstorming, comparing approaches, and checking assumptions outside the editor. The combination matters more to me than expecting one assistant to handle every part of research, implementation, and review.

These products continue to evolve quickly, so I avoid anchoring recommendations to a temporary model list, price, or usage-statistics service. The durable question is whether the tool fits the repository, preserves enough operator control, and makes its changes easy to inspect.

Add your thoughts or questions about the topic 2024-09-02 AI Coding Assistants