Doing a demi-Gruber; sitting on a beach, not earning 20%.
Search
5 results for “looptron”
-
⏭️ Skip an interview question or manual test, approve a plan with gaps, cancel, or finish before changes are merged? LoopTroop saves your reason. The note stays with the task, so you can see what happened without digging through a long chat or guessing which step got skipped.
💬 When several AI models ask you questions, each now gets its own tab. Choose how long LoopTroop should wait, from 1 to 60 minutes. The countdown stops as soon as you start replying, so the timer won't run out while you type. You can compare and answer each model separately.
🔌 OpenCode v2 now works with LoopTroop, and v1 still works too. 💾 If a run is interrupted, canceled, or restarted, LoopTroop is better at picking up where it stopped. It can spot when GitHub merges a code change, even if LoopTroop is closed, and mark the task done for you.
📜 Logs and exports keep more details through reconnects, retries, and cancellations, so you can see where a long run stopped. ✨ Keyboard use is easier, and drafts and answers stay with the right task when you move between tasks. You can see what happened before coming back.
🪟 On Windows, you can now install LoopTroop with Chocolatey or WinGet; Scoop is fixed too. On Node 24.18.0 or newer, installs can use the npm that comes with Node, so you don't need to pick a separate npm version. One less version detail to sort out on a new machine.
🛑 These fixes are part of the 0.6 release: stop or retry one AI model without stopping the others, and skip every question that's waiting for you with one action. Useful when one model hangs or you want to keep the rest of the task moving while you sort out one response.
🔍 When something breaks, v0.6.1 gives you clearer errors and a better way to continue. It fixes OpenCode startup and model-detection problems, improves how LoopTroop handles OpenCode v2, and reads AI replies that confused the app. There are fixes for Windows and Linux, too.
⚡ Terminal commands can be up to 20× faster in v0.6.1. In the release's recorded checks, status and help took about 0.3 seconds. Install packages are about 30 MB smaller, so there’s less to download. Those are measured examples; not every command will be 20× faster.
-
🧪 In 0.6.0, passing tests went from 3,495 to 8,099 across 500 test files, checking 93.19% of the code lines. A speed check that read the same 64-step plan again and again went from 24.3 ms to 0.68 ms. That's about 36× faster for that check, not a whole coding task.
LoopTroop helps with coding tasks too big for one prompt: several AI models help plan, OpenCode works through small steps in its own folder, and you review the changes before sharing them on GitHub. v0.6.1 is the current version to install. Full notes: https://github.com/looptroop-ai/LoopTroop/releases/tag/v0.6.1