ParviDefenders · ParviRange · Chefe 2665 · 21 Aug 2026
I never meant two windows. I said two doors (how a brain arrives). The swarm is 65+ terminals at once: a parent wave in a window, children waves in more windows, in parallel, fast. For us (the house), our LLM-Hub is fine. For ParviDefenders and ParviRange, we must ask the customer first: their own hub with several AIs, or one AI as the brain those agents run with to build tools. We do not silently plug our house hub into their machine.
| Today in the engine | What you just set | |
|---|---|---|
| Live terminals at once | default 64 (65th waits unless we raise it) | 65+ must work |
| Children from one parent | default 8 | parent keeps spawning children until the job is done, inside the 65+ |
| Paid “pro” license | 50 terminals — that would block 65+ | too low for this promise |
| Brain on a customer install | still points at a hub URL; house is the default | must ask them first |
65+ is the working number. 64 in the engine is one short. I did not shrink your swarm to two windows.
Nothing invented. Your 2662 question was “who is coding.” This message is the product rule for whose brain and how many windows. Same Core. Two products (Defenders, Range) may not ride our house hub.
Only you — two yes/no for the next build:
If both are yes, that is the next change. No extra product. No Code-Hub as the pane brain.