AI tools in a developer's arsenal
The internet is full of "best AI tools for programmers" lists. Most written by people who don't use them in real work. We do - and we have opinions.
What we actually use
Claude Code (Anthropic)
Our main coding partner. Not perfect, but it does three things well:
- Understands project context - you can feed it your entire codebase and it will remember what's where
- Writes sensible code - not "it works but I don't know why", but code you can understand
- Helps with debugging - paste an error, get an explanation and a fix
VS Code + Copilot
Autocomplete that speeds up writing repetitive code. Won't design your architecture for you, but saves typing boilerplate.
Marked + custom scripts
Simple tools for building static sites. AI helps write build scripts, but the architecture is ours.
What we've learned
AI is good at repetitive tasks. Generating components, writing tests, converting formats. This is where it saves hours.
AI is weak at architectural decisions. It'll suggest a solution but won't take responsibility for the consequences. That's still your job.
AI requires supervision. AI-generated code needs to be reviewed just as carefully as code from a junior developer. Sometimes it works, but does it wrong.
Practical advice
Don't start with "AI will build my entire project." Start with one task that frustrates you - writing tests, documentation, refactoring legacy code. Use AI for that one task. Evaluate the result. Then expand.
AI tools are a multiplier of your skills, not a replacement. A good developer with AI is faster. A bad developer with AI makes the same mistakes, just faster.
