Chaltteok Labs
Say it roughly.
Get it done right.
We are building an autonomous workroom: you describe what you want in plain, imperfect words, and an AI manager works out the intent, plans the job, hands pieces to worker models, checks the results, and brings back finished work.
The name comes from a Korean saying — even a request as lumpy as gaetteok should come back as neat as chaltteok.
How it works
- A manager seat, run by Claude. The manager reads the request, decides what matters, and only asks when the intent is truly unclear or a hard limit is touched.
- Project leads and workers. Larger jobs are split into projects. Leads assign tasks to worker models, review what comes back, and send it for rework until it passes.
- Handovers that keep the thread. Long-running work outlives any single session. The context lives on the local machine and follows the task, so whoever sits in the seat next can pick up without carrying everything.
- The person stays out of the loop. The goal is that the owner is not called in — results appear, with a short report of what was done and why.
Status
In daily use and active development since September 2026. Private alpha, running on a single local machine. Not yet open to outside users.