I feel like you're ignoring the amount of training that an expert musician does to learn a specific piece of music and maintain their proficiency in it. For the most part, they don't go to professional gigs and play something novel.

Traditional musicians have a whole live or real-time performance aspect as do athletes, dancers, etc. I think the amount of time they spend preparing for this can be similar to the time we spend working on one programming task. Bigger problems take more preparation. The difference is we don't have to then do a live performance after we've figured out how to program it. We just accumulate a recorded artifact and ship it, rather than doing a live recital after we've figured out all the difficult bits.

So it's difficult to draw parallels. Programmers have more in common with writers, painters, and sculptors who all work on a tangible artifact that is delivered after the fact and which acts as an accumulator of time-shifted work product. Some crafts, like glass blowing, are more like live music in that you develop a skill but then have to make a real-time performance each time you produce the artifact.

I made something similar as well (that includes both 3.11, 95, 2000, XP, CDE and Mac OS 9, and also all the default color schemes of those): https://nielssp.github.io/classic-stylesheets/?theme=win9x&s...

My focus was not so much on pixel perfect, but instead on creating something that would also work and look aesthetically pleasing on modern systems, like with higher DPI monitors and such. So one of the the things I did was to recreate all the icons and symbols in SVG.

I tried posting it as a Show HN when I added XP and Mac OS 9, but it didn't get much attention. Maybe the title of the project isn't as catchy.

I was thinking of making a database of only studies that we can “trust”

What does HN think? Had a tough time determining the barometer for what is trustable and what isn’t

How are folks parsing PDFs with GPT4? Every 3rd party tool that supposedly uses GPT has failed for me on even the most regular of PDF bank statements. Currently I am parsing the text contents manually using pdfbox to reconstruct historical bank statements, but this is error-prone due to page breaks and run-on lines.