You might want to fix the Geim -> Gein autocorrect. I was horrified at the thought of what “experiments” Ed Gein might have been up to. en.wikipedia.org/wiki/Ed_Gein

Based on the snippets you edited into this video you’re only mentioning Claude and ChatGPT, not the specific models like GPT-4o, GPT-4.5, Claude Sonnet 4, Claude Opus 4.1, etc. The different versions have different capabilities, but I wouldn’t be surprised to find that they indeed all suck at music.

I wonder if the internal communication of this is more nuanced, putting it in the context of other metrics. One can only hope that it’s paired with “don’t regress on these other metrics related to quality.”

You can also look at it like this: gaining market share with unsustainably low prices is just another tool in service of corporate greed, but with extra steps.

Assuming you meant “health” in a more conceptual sense, as opposed to uptime, resource usage, etc.

In an attempt to quantify vibes, we are doing regular surveys within the Engineering department to identify problem areas within the code base, tooling, etc.