Life got busy. I don't have the hours to run long AI sessions anymore, so I built something to handle the repetitive parts for me. Looping, prompt queues, personas, crash recovery, planning. Works across ChatGPT, Claude, Gemini, Perplexity, Grok, Copilot, DeepSeek and a few others.
It's called Ghost in the Loop. Free, no account, installs like any userscript.
New prototype at the repo: https://raw.githubusercontent.com/MShneur/ghost-in-the-loop/main/dev/ghost-in-the-loop.user.js
GitHub: https://github.com/MShneur/ghost-in-the-loop
What I actually want is simple: show me if it fails in your browsers, dev tool errors, html errors, or your personal read on it.
I built this around my own workflows, which means I've probably baked in my own blind spots without realizing it. If you work differently, use different platforms, chain tasks in weird ways, or have a prompting style I haven't thought of, I want to see where it fits and where it falls apart.
Less "please find my bugs" and more "what slot is missing from this thing."
I'll take anything. Friction points, feature gaps, workflow ideas. Weirder the better..
[link] [comments]