What I'm using
to make the work.
Inspired by Wes Bos's /uses tradition — the hardware, software, and small daily tools that go into building things. Updated when something changes, which is rarer than I'd like to admit.
Hardware
- Laptop
- MacBook Pro M2, 16GBThe screen is half the desk.
- Display
- Dell U2723QE 27"4K, USB-C, one cable in.
- Keyboard
- Keychron K3 ProBrown switches, ortholinear curious.
- Mouse
- Logitech MX Master 3SOr a track-pad when commuting.
- Headphones
- Sony WH-1000XM4For coffee shops.
- Camera
- Olympus OM-10 filmBought from a man in Patan.
Editor & shell
- Editor
- Neovim + LazyVim+ Cursor for AI-heavy days.
- Coding agent
- Claude CodeAnthropic's terminal agent.
- Terminal
- Ghostty+ tmux for sessions.
- Shell
- zsh + starshipMinimal prompt, fzf for jumps.
- Font
- JetBrains MonoLigatures off.
- Theme
- Tokyo NightSame in nvim, terminal, OS.
AI & ML stack
- Models
- Claude, GPT-4o, local LlamaReach for what fits the task.
- Framework
- PyTorchSometimes JAX when curious.
- Orchestration
- LangChain & LlamaIndexSparingly. Roll your own past prototype.
- Vector DB
- Chroma / pgvectorPostgres wins by default.
- Tracing
- Langfuse / OpenTelemetryFor when prompts go strange.
- Eval
- Custom (see prompt-bench)Pytest + a small YAML.
Web stack
- Backend
- FastAPI / DjangoFlask for tiny things.
- Database
- PostgreSQLPlus SQLite for prototypes.
- Front-end
- React / HTMXVanilla when I can get away with it.
- Styles
- Tailwind / hand-rolledDepends on the team.
- Deploy
- Cloudflare + HetznerOr Fly for the smaller things.
- CI
- GitHub ActionsBoring on purpose.
Daily small stuff
- Notes
- ObsidianVault sync via iCloud.
- Tasks
- Things 3+ a paper notebook.
- Browser
- ArcSafari for reading.
- Reader
- Reeder + RSSNo algorithmic feeds.
- Password
- 1PasswordFamily plan, family approved.
- Coffee
- AeroPress, dark roastTwice in the morning, never after 2pm.
What I learn from
- Book
- Designing ML SystemsChip Huyen.
- Course
- nanoGPT walkthroughKarpathy is a teacher.
- Newsletter
- Latent Space, Tldr.aiThe two I actually read.
- Papers
- via arxiv-sanityQuality over quantity.
- Podcast
- Latent Space, DwarkeshFor runs.
- People
- Hamel, Jeremy, ChipTheir writing shaped my approach.
Disagree?
Tell me what you'd swap.
I rebuild this list every year or so, mostly when I realize something on it has been collecting dust. If you've got a tool I should be using, say so.
Send me a tip →