SkillOpt: Training Loop for Agent Skills

My morning ponder was RSI... Improvement or Immolation, wondering where ai working with its own output will head

Do people not look at their ai output? What are most people doing these days? Sorry, but I cannot read past the visual "aesthetics" of this site

I'm nearing the "RSI" point for my own context/engineering improvement setup, but I've remained very much in the loop, doesn't seem like a good idea to leave them to their own devices yet

Maybe I’m missing something, but it seems to me, the performance numbers on the main results charts are meaningless without any units or other ranges or indicating they are percentages. It’s also surprisingly non quantitative for basic quantitative type writing and work. Is that a 9 point gain on a range of 0 to 1,000? 9 percent gain?

github: https://github.com/microsoft/SkillOpt