What AI tools do you use?
Short answer
Day to day I mainly use Cursor as my primary coding agent environment, alongside general-purpose LLMs for research and explanation. I integrate them with my existing toolchain (git, tests, linters, CI) rather than treating them as a separate system.
Longer answer
The core tools I rely on are:
-
Cursor IDE (coding agents)
- Context-aware code completion and refactoring.
- Multi-file edits with awareness of the repository.
- Automated test generation and updates.
- Architecture-aware refactors when I provide clear guidance and rules.
-
LLM-based assistants for design and documentation
- Exploring alternative designs and architectural patterns.
- Translating complex code or legacy logic into clearer documentation.
- Drafting RFCs, ADRs, and design docs that I then refine manually.
-
AI-assisted code search and navigation
- “Where is X done?” or “How does Y flow through the system?” questions over large codebases.
- Quickly finding all the places an abstraction is used before doing a refactor.
I don’t treat any of these tools as infallible. I combine them with:
- Static analysis and linters to enforce style and catch obvious issues.
- Unit/integration tests to validate behavior.
- Code reviews where humans make the final call.
This combination lets me get the speed benefits of AI while keeping a strong quality and safety net around the codebase.
Interview angle
- “Which AI tools do you use?” - name them specifically and say what each is for. Vagueness reads as not actually using them. Pair each tool with the task it fits: an agentic coding tool for multi-file changes, inline completion for flow, a chat model for design discussion and unfamiliar APIs.
- “How do you decide which to reach for?” - by task shape. Well-specified mechanical change: an agent. Exploration and naming things: chat. Line-level flow: completion. Reviewing a diff someone else wrote: chat with the diff pasted in.
- “What do you configure?” - project-level conventions (a
CLAUDE.md,.cursorrulesor equivalent) so the tool stops re-litigating style every session, and a scoped working directory for anything that edits files or runs commands. - “Any tool you tried and stopped using?” - having one is a credibility signal. It shows you evaluate rather than adopt, and the reason matters more than the tool.