Mastodawn

How do you get better-than-frontier performance with models? You get a panel of them.

If you’re a #pi user, give this extension a spin: https://github.com/leblancfg/pi-fusion

Compound model panel for pi: parallel model calls, then one synthesis response. - leblancfg/pi-fusion

GitHub

#OpenRouter has model fusion, which this recreates in pi.

This tracks with internal evals: the “million monkies” approach to some use cases is better than a frontier model with ‘xhigh’ reasoning.

While it has slick and useful #tui panes, it’s purpose is production workloads: all output goes to the session file, and is resumable.

You control everything: prompts, models, reasoning, presets, tools it uses.