Superconductor adds Qwen3.8-Max multi-agent support

As LLMs gain the ability to execute autonomously over long horizons, traditional synchronous chat interfaces become bottlenecks for developer productivity. Stacking agents across isolated git worktrees prevents active LLM runs from locking up local developer environments. Session resumability and task completion notifications solve the friction of losing context or missing completed runs when alt-tabbing. Broadening model support to frontier options like Qwen3.8-Max highlights the shifting landscape toward native, model-agnostic agent orchestration platforms.