My permanent music snapshot
Every number below is counted from my real library export: 5,262 unique songs across 1,119 artists, deduplicated and broken down honestly. I invented nothing here, and there's no proprietary audio-features data because my export doesn't contain it. Just what's really in my library, plus a hand-picked set of favourites with the actual track, a lyrics excerpt, and a link to the rest.
The pipeline behind this is unglamorous on purpose. I exported my library, wrote a script to dedupe it and pull the genre and decade breakdowns, then built this page to render the real output. I didn't guess at anything, and I didn't fill a single gap with something that merely looks plausible. It's the same self-hosted habit as everything else here, and there's more on how that side of things runs on the Infrastructure page.
🔁 On rotation right now
What I'm listening to at the moment. That's a different question from what I own the most of, and this one has a shelf life, so it's dated.
📊 Library overview
🎯 Genre Spotlight
The same real per-artist counts as the ranked list below, split out by genre. Rock/Hardcore/Metal and Rap are hand-curated rather than a strict count sort, and say so under the list; the rest are a straight top 10 by count.
🏆 Every artist, ranked
My top 489 artists, ranked by how many of their songs are in the library. Not vibes, not memory, just the real count. A few of these placings surprised me and I left them where the count put them. Search for someone specific, filter by genre, or click any row to see some of what I saved by them. Going from 250 to 489 covers 88% of the library instead of about three quarters, and it costs something worth saying out loud: I have a genre on file for 247 of these, so about half the new arrivals sit in Other until I classify them. The count is still real for every one of them. The colour is the part that's incomplete.
📉 The long tail
What does my taste look like when I plot every last bit of it? The ranked list above only shows the top of the library. This is all 1,119 artists at once, so you can see the real shape: a small handful I have gone completely feral over, and then a very long, very flat tail of artists I saved one song from. Hover anywhere on the curve to find out who's at that position.
🌌 Genre galaxy
My whole library as a map instead of a list. Every star is one of my top 150 artists, sized by how much of my library they take up and colored by genre, but where they sit is the real point. I researched documented links between them (a song together, a tour together, something one said publicly about the other) and let those pull the map into shape, so artists who are genuinely connected end up near each other. Nothing here's placed by feel. If two stars sit close together, I have a reason on file. Hover a star to light up its connections, or click one for the receipts.
🎨 Genre breakdown
📅 Music by decade
💭 What this says about my taste
🧭 The emotional compass
Two axes: calm to intense across, melancholy to bright up and down. Every genre and a few dozen of my most-saved artists, placed by feel. This one is opinion, not measurement, and I want that said before the chart, not after it. The circle sizes are the real song counts. The positions are mine. Hover or tab through any point for the numbers behind it.
🔤 Words I keep saving
Straight text analysis over every unique track title in the library: which words recur across five thousand songs, once the filler and the format noise are stripped out. There's no interpretation in the counting, though I have plenty of opinions about the result.
The one thing I get asked for and can't build? A timeline. There's no discovery chart on this page, no listening-evolution curve, and no library-growth graph, because my library export is nothing but "Artist - Title" rows, with no date added, no timestamp, no play count, and no release year. There's genuinely nothing to plot a time axis against. I could have used the order of lines in the file as a stand-in for time and nobody would have checked, but I have no way to prove that file is even in chronological order, and a chart that might be fiction is worse than no chart. I listen on YouTube Music now, which does keep real timestamps, so the honest version of this section is something I can build from my own data later instead of something I fake today.
📜 Nine years of receipts
Everything above is derived from a library export. This is the opposite kind of evidence: two real, dated posts I wrote on Reddit, years apart, pulled from my own account export and quoted here instead of reconstructed from memory. Nobody wrote these for this page. They were already true before it existed.
⭐ Favourites, up close
Specific songs, not artists. Ordered by how many of that artist's songs are in my library, which is the only number here that means anything. There are no position numbers on the cards, because which song I picked from each artist is a judgment, not a rank. Every video is a verified official upload, checked against YouTube's own metadata and not a search snippet. Where I couldn't verify one, the card says so and links out instead of embedding a guess. Start any video and it becomes the first song in a continuous Top Picks queue: the floating mini-player keeps going through every other verified favourite while you browse lilharper.dev, then loops around. Minimize it to a bar, bring it back, or close it whenever you want. Lyric excerpts load on request, and the Genius link works regardless.