What I use.

Primary machine. The M4 chip handles PyTorch training, Docker containers, and 47 browser tabs simultaneously.

One wide monitor. No more alt-tabbing between terminals and notebooks.

Tactile switches. The clickety-clack is 40% of the productivity.

Active noise cancellation — the second-best productivity tool after good coffee.

VS Code with AI superpowers. This entire website was built in it.

With zsh and Oh My Zsh. Theme: minimal. Plugins: git, z, syntax highlighting.

Terminal multiplexer. Keeps sessions alive when I inevitably close the wrong window.

Primary language. NumPy, Pandas, scikit-learn, PyTorch, FastAPI — the whole ecosystem.

For anything web-facing. Types save future-me from past-me.

PostgreSQL day-to-day. BigQuery for large-scale. Still the most underrated skill in data science.

For statistical analysis and publication-quality visualisations. ggplot2 is genuinely beautiful.

Automating everything that shouldn't need automating but somehow does.

Deep learning framework of choice. Dynamic graphs, Pythonic, excellent ecosystem.

Still the gold standard for classical ML. Pipelines are underused.

SQL transformation layer. Version-controlled data models — finally.

For exploration and sharing analysis. VS Code notebooks for production.

Dashboards for non-technical stakeholders. Sometimes the right tool is the accessible one.

Framework of choice for web. Pages Router, server components, fantastic DX.

Utility-first styling. Initially sceptical. Now converted.

Deploy → done. Preview URLs on every push. The best CI/CD experience I've used.

Design and prototyping. Also for communicating with designers who don't speak code.

Project notes, research logs, and second brain. Imperfect but flexible.

Issue tracking for Mapiva. Fast, minimal, opinionated — very Nothing OS.

Password manager. Non-negotiable.

Window management on macOS. Free. Simple. Indispensable.