Head of AIOps @IBM, Andreas Horn | 228K followers on Linkedin 🤯 𝗖𝗹𝗮𝘂𝗱𝗲 𝗖𝗼𝗱𝗲 𝗼𝗻 𝘀𝘁𝗲𝗿𝗼𝗶𝗱𝘀 - 𝘂𝘀𝗶𝗻𝗴 𝗦𝗽𝗲𝗰-𝗗𝗿𝗶𝘃𝗲𝗻 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 (𝗦𝗗𝗗) 𝗶𝗻𝘀𝘁𝗲𝗮𝗱 𝗼𝗳 “𝘃𝗶𝗯𝗲 𝗰𝗼𝗱𝗶𝗻𝗴.” That's what GSD (Get Shit Done) feels like after using it for the last two weeks. Didn't expect much from another Claude Code repo, after testing hundreds. GSD proved me wrong. I’m genuinely impressed. 𝗪𝗵𝗮𝘁 𝗶𝘀 𝗚𝗦𝗗? Repo that adds a context engineering layer that turns AI-coding into actual shipping. Claude Code is brilliant - until context rot kicks in. 50k tokens in, it's "being more concise." 100k in, it's hallucinating functions. 150k in, you're debugging its debugging. GSD's brutal fix: every task gets a fresh 200k context window. No accumulated garbage and no degradation. 𝗧𝗵𝗲 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄 𝗶𝘁 𝗳𝗼𝗹𝗹𝗼𝘄𝘀: ✦ Discuss → AI asks until it truly gets your vision ✦ Research → Parallel agents investigate the domain ✦ Plan → Atomic tasks with built-in verification ✦ Execute → Fresh context per task, one surgical commit each ✦ Verify → You confirm it works, AI auto-generates fix plans if not It feels like magic, but it's just scaffolding designed for how LLMs actually behave - not how we wish they did. Install with one command: npx get-shit-done-cc Repo: Anyone here experimenting with this? What’s your take so far?