Every year I try a bunch of new tools, and most of them do not stick. They are either too complex for what I need, too opinionated about workflow, or just marginal improvements over what I already use. But a few tools from 2021-2022 have genuinely changed how I work. Here are the ones that stayed.

Warp Terminal

I switched from iTerm2 to Warp about six months ago and have not looked back. The key difference is that Warp treats terminal output as blocks. You can select, copy, and scroll through the output of a single command without touching the output of the previous one. It sounds small, but when you are running multiple commands and referencing their output, it is a significant improvement.

The built-in command search and AI-powered command suggestions are nice, but the block-based output model is what keeps me using it daily.

Fig (Now Amazon CodeWhisperer for CLI)

Fig adds IDE-style autocomplete to your terminal. It works with bash, zsh, and fish, and it provides context-aware suggestions for hundreds of CLI tools. When you type git checkout, it shows your branches. When you type docker run, it suggests image names. When you type kubectl get, it lists resource types.

The autocomplete specs are open-source and community-maintained, so coverage keeps improving. It has saved me from checking --help output countless times.

Raycast

I replaced Alfred with Raycast as my application launcher and general productivity tool. The built-in extensions are what sold me - clipboard history, window management, snippets, and a calculator all work without installing separate apps. The developer experience for building custom extensions is also solid, built on React and TypeScript.

The one feature I use most is the window management. I used to run a separate tiling window manager, but Raycast handles my basic needs (halves, thirds, centering) without another tool in the stack.

HTTPie Desktop

I have used HTTPie on the command line for years as a cleaner alternative to curl. The desktop app brings the same philosophy - clean interface, sane defaults, good formatting - to a GUI. It is what Postman should be: fast, focused, and not trying to be an entire API development platform.

I still use curl and the CLI version for quick one-off requests, but for anything involving multiple headers, auth tokens, or request bodies, the desktop app is faster.

Obsidian for Technical Notes

I moved my technical notes from Notion to Obsidian. The reason was simple: Notion is slow, and my notes are plain text. Obsidian stores everything as local Markdown files, so it opens instantly and works offline. The backlink feature helps me connect related concepts across notes, and the graph view occasionally surfaces connections I had forgotten about.

My setup is minimal: one vault, a flat folder structure with a few top-level categories, and the daily notes plugin for quick captures. No fancy templates or complex plugins.

Turborepo

For monorepo management, Turborepo has been excellent. I was using Lerna before, and the difference in build speed is dramatic. Turborepo's remote caching means that if a teammate has already built a package with the same inputs, my build pulls the cached output instead of rebuilding from scratch.

The configuration is minimal - a single turbo.json file that defines your pipeline - and it works well alongside existing package managers like pnpm.

Tailscale

Tailscale is a mesh VPN built on WireGuard. I use it to connect my development machines, home server, and cloud instances into a private network. No port forwarding, no firewall rules, no fiddling with IP addresses. Every device gets a stable hostname, and connections are encrypted end-to-end.

The practical result: I can SSH into my home machine from anywhere using ssh home-desktop, access my local services from my phone, and share development servers with teammates without deploying to a staging environment.

What Did Not Stick

For honesty's sake: I tried GitHub Codespaces (too slow for my connection), Neovim with LSP (too much configuration time for marginal gains over VS Code), and Deno (the ecosystem is still too thin for production work). All interesting, but not yet ready to replace what I already use.

The theme across the tools that stuck is that they reduce friction without adding complexity. They do one thing well and integrate with existing workflows instead of demanding you rebuild your workflow around them. That is the bar I set for any new tool now.