March 17, 2026
Code rocket or token bonfire?
Get Shit Done: A Meta-Prompting, Context Engineering and Spec-Driven Dev System
Builds code fast or burns cash faster? Devs are divided
TLDR: A new tool called GSD promises faster, more reliable AI-coded projects by managing context for assistants like Claude and Gemini. The community’s split: some brag about massive output, others call it a token-guzzling hype machine with a risky “skip permissions” vibe—big potential if you’re precise, big bills if you’re not.
“Get Sh*t Done” (GSD) dropped with big energy: a plug-in system that promises to make AI coding tools like Claude Code and Gemini actually reliable by fighting “context rot” — when the bot forgets what you said earlier. The creator says it skips the corporate cosplay and gives you a few simple commands, with claims of engineers at big-name companies using it and a repo you can poke at here.
But the comments? Pure chaos. One user flexed that GSD helped churn out “250K lines of code in less than a month,” crowning it the productivity cheat code. Immediately, skeptics crashed the party: another tester said it didn’t beat just prompting Claude directly, but obliterated their usage limits — “5-hour cap in 30 minutes, weekly max by Tuesday.” The word “token burner” kept popping up, with comparisons to rivals like Superpowers. Meanwhile, the name alone sparked its own micro-riot: “terrible name, DOA,” sneered one commenter, as others joked it sounds like a gym routine for your laptop.
More nuance came from the pragmatists: the tool shines if you can describe what you actually want. That’s the catch. Also raising eyebrows: the suggestion to run with a “dangerously skip permissions” flag — rocket fuel, sure, but buckle up. Verdict from the crowd: genius for laser-focused builders, risky wallet-drain for everyone else.
Key Points
- •GSD is a meta-prompting, context engineering, and spec-driven development system for Claude Code, OpenCode, Gemini CLI, and Codex.
- •It aims to mitigate 'context rot' and make AI code generation more reliable by handling context and verification behind the scenes.
- •The tool emphasizes simplicity for solo developers, contrasting with more complex spec-driven frameworks like BMAD and Speckit.
- •Cross-platform support is provided (Mac, Windows, Linux), with installation via npx and options for global or local setup and multiple runtimes.
- •Non-interactive and development installation methods are detailed, including GitHub cloning and a recommendation to run Claude Code with a permissions-skipping flag.