Turning 11 years to 11 days - Workaround for slow Sol speed

I used to lean on Sol as my workhorse and let it manage everything. That has broken down with the slow speed. And at the worst time when I need throughput more than intelligence for a huge research epic that breaks down into many smaller but interdependent tasks. So I switched my approach. Disclaimer, this will not work for all situations and some may have figured this out as a normal way of working already. I am not claiming to have invented the wheel just sharing something useful. Normally, I would prefer Sol splitting work with Sol over Luna subagents for tough code issues. But for the current research I need bandwidth so Luna subagents with Sol orchestration intelligence works. But running Sol Ultra on the task even with guidance to scale out and use Luna did not really do it. It was slow and with the Sol orchestrator spinning up occasionally one or two subagents, sometimes Sol sometimes Luna. What did the trick was: Update the config.toml to include Luna as default and crank the threads up to 20 Using Sol Ultra with a goal that is focused on scale and not perfection. Guide it to quickly backlog work if it gets stuck, focus on the first 80% and park the tough 20%. The previous default setup with Sol Ultra predicted an ETA in 11 years , I kid you not. Now it has 10+ Luna subagents active most of the time. I have an ETA of 1-2 weeks. [agents] enabled = true default_subagent_model = "gpt-6-luna" default_subagent_reasoning_effort = "max" max_concurrent_threads_per_session = 20