[wip] Eval and Vibes

[WIP]

notes

• vibes are everything
• Teams use evals as a goal and we get very spotty performance
• Hard to get universal perofiemance right , or specific per mask masquerading as universal perf mask
• Anthropic and cursor as a good example , because power users love it there is a vibe
• Lots of normal people use ChatGPT in ways that no one anticipates
• Vibes are getting more important
• Comparative vs absolute
• LLMs as Mirrors effect the better you are
• What do we care about
• Human standards keep rising , models scores lower but only in comparison. To previous models
• We measure humans based on what they are given and what tbet can do