If the bubble bursts we club together to buy one of those big GPU servers (now available at rock bottom prices thanks to the bubble bursting) and run a shared instance of GLM-4.7 (the current best-at-coding Chinese open weight model) on it.