// the take
try is a single-file Ruby CLI that gives every throwaway experiment a date-prefixed home in one directory (~/src/tries) and a fuzzy-search TUI to jump back into them, plus shortcuts for cloning repos or creating git worktrees into the same scheme. It's for developers who constantly spin up test-this and test-that folders and lose track of where the good one was.
Zero-dependency single Ruby file that ships as a gem, a raw curl-able script, or a compiled native binary, so there's no real install friction and it's trivial to read end to end. The fuzzy matcher is recency-weighted rather than pure string distance, so a 2-day-old redis-pool experiment outranks a 2-week-old one on a partial match, which is the actual UX problem this tool exists to solve. Git handling is more thought-through than you'd expect from a 'just one file' project: it parses HTTPS/SSH/GitLab/custom-host URLs, strips .git suffixes correctly, and can clone a PR URL straight into a detached-HEAD checkout.
The native-binary build path depends on an unmerged Spinel PR (matz/spinel#3906) for IO#tty?/winsize support, so the faster binary isn't actually available to anyone who isn't willing to build a patched Ruby implementation from source — most users are stuck on MRI startup time. There's no pruning, archiving, or size management: directories accumulate under TRY_PATH forever, and the only cleanup is a manual Ctrl-D per entry, so 'thousands of experiments' (which the FAQ brags about handling) is really thousands of directories you'll eventually have to deal with by hand. No Windows support is mentioned anywhere — install paths are Homebrew, Nix, RubyGems with bash/zsh/fish shell hooks, so Windows/PowerShell users are out unless they're in WSL. By the project's own admission in the FAQ, it's explicitly not meant for anything you intend to keep, so it solves exactly one narrow problem and nothing adjacent to it.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →