<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: nighthawk454</title><link>https://news.ycombinator.com/user?id=nighthawk454</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Fri, 20 Nov 2026 23:26:12 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=nighthawk454" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by nighthawk454 in "E-means: Outperforming k-means in one-tenth the epochs"]]></title><description><![CDATA[
<p>Yeah, it's not a well-defined problem, so there's no one true solution. Then only question left is "well what do you want?". Always have to pick something.<p><a href="https://en.wikipedia.org/wiki/Cluster_analysis#Definition" rel="nofollow">https://en.wikipedia.org/wiki/Cluster_analysis#Definition</a></p>
]]></description><pubDate>Tue, 10 Nov 2026 04:03:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=50435916</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=50435916</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50435916</guid></item><item><title><![CDATA[New comment by nighthawk454 in "E-means: Outperforming k-means in one-tenth the epochs"]]></title><description><![CDATA[
<p>The 'Elbow method' [1], yeah. Sweep a bunch of different k values and pick the one where your target metric starts having diminishing returns. There are a lot of heuristics you could use.<p>There's still not one particularly great way to pick k for this type of clustering. There kind of can't be as clustering is sort of inherently subjective.<p>Other methods like HDBSCAN can find the number of clusters themselves. There's still a 'one key parameter' you have to pick yourself, but it can be more intuitive and less sensitive for some datasets, when you don't know how many clusters there are / how many you want.<p>Sometimes, picking k exactly is useful for practical reasons too. Like fixing the size of your codebook. Or partitioning a dataset into k equal-ish pieces, say to spread across a pool of servers of a known/fixed size.<p>[1] <a href="https://en.wikipedia.org/wiki/Elbow_method_(clustering)" rel="nofollow">https://en.wikipedia.org/wiki/Elbow_method_(clustering)</a></p>
]]></description><pubDate>Tue, 10 Nov 2026 03:38:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=50435768</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=50435768</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50435768</guid></item><item><title><![CDATA[New comment by nighthawk454 in "What Is Gridifcation"]]></title><description><![CDATA[
<p>This guy did it to Van Halen, and it renders it utterly lifeless.<p><a href="https://youtu.be/p9rVKnEvAQ8?t=240" rel="nofollow">https://youtu.be/p9rVKnEvAQ8?t=240</a></p>
]]></description><pubDate>Fri, 23 Oct 2026 02:16:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=50203742</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=50203742</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50203742</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Neural networks are clean"]]></title><description><![CDATA[
<p>At the risk of oversimplifying, it seems the field looked at neuron activations and couldn't find the pattern, so a series of convoluted hypothesis were proposed to explain it away. When the actual answer was the pattern was very simple they just missed it. Little bit embarrassing really.<p>The good news I guess is the actual patterns make a lot of sense with what you'd expect the neurons to do, and don't require nearly as much 'black magic who knows what wonders scale can create'.</p>
]]></description><pubDate>Sat, 17 Oct 2026 19:35:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=50134338</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=50134338</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50134338</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Kev: Tiny Jev-like family of decision models built on top of Qwen3.5"]]></title><description><![CDATA[
<p>which amounts to nearest-neighbor</p>
]]></description><pubDate>Tue, 22 Sep 2026 03:53:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=49796621</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=49796621</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49796621</guid></item><item><title><![CDATA[New comment by nighthawk454 in "How concerned should we be about Astra's recurrent architecture?"]]></title><description><![CDATA[
<p>Anyone remember Universal Transformers paper (Dehghani et al) from back in 2018? Recurrent transformers have a history as long as transformers themselves.<p>Somewhat unclear how particularly novel this is vs a way to save compute.</p>
]]></description><pubDate>Thu, 03 Sep 2026 20:06:54 +0000</pubDate><link>https://news.ycombinator.com/item?id=49556090</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=49556090</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49556090</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Show HN: BitBang – Reach machines behind NAT from a browser, no account"]]></title><description><![CDATA[
<p>Ah good point, no udp. I was looking at webtorrent but it does share the same limitation, unless you’re on the nodejs (non-browser) version.<p>Looking forward to trying bitbang! Seems super neat</p>
]]></description><pubDate>Sat, 01 Aug 2026 00:57:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=49130133</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=49130133</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49130133</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Show HN: BitBang – Reach machines behind NAT from a browser, no account"]]></title><description><![CDATA[
<p>> it still needs to bootstrap the connection via central server. I think that always might be a limitation of P2P<p>I think Iroh has a way of doing it using either DNS or BitTorrent DHT. Not perfect but fairly decentralized</p>
]]></description><pubDate>Fri, 31 Jul 2026 19:20:17 +0000</pubDate><link>https://news.ycombinator.com/item?id=49127452</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=49127452</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49127452</guid></item><item><title><![CDATA[New comment by nighthawk454 in "More Tailscale tricks for your jailbroken Kindle"]]></title><description><![CDATA[
<p>You can add a library view! Checkout this Bookshelf plugin:<p><a href="https://github.com/AndyHazz/bookshelf.koplugin" rel="nofollow">https://github.com/AndyHazz/bookshelf.koplugin</a></p>
]]></description><pubDate>Wed, 29 Jul 2026 18:52:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=49101486</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=49101486</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49101486</guid></item><item><title><![CDATA[New comment by nighthawk454 in "As downtown Seattle offices empty, city facing years of 'zombie' towers"]]></title><description><![CDATA[
<p>This tracks with what I’ve heard around as well. What changed in the financial instruments?<p>My understanding is a lot of the loans have gone PIK or otherwise essentially aren’t serviceable at current prices. Do you think that’s resolvable somehow or just lagging implosion?</p>
]]></description><pubDate>Sat, 04 Jul 2026 20:39:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=48788864</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48788864</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48788864</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Dispersion loss counteracts embedding condensation in small language models"]]></title><description><![CDATA[
<p>Some potentially related stuff on the topic:<p>Anisotropy in word embeddings dates back to at least 2017 with word2vec - where there were zero layers.<p>The cone-shaped anisotropy in transformers is known since at least Gao et al. 2019. That lineage explained it fairly intuitively as an artifact of word frequency and softmax geometry (so a training dynamic).<p>A variety of papers followed up by adding post-hoc ‘whitening’ steps (from classical statistics/NLP), then adding regularizers to the loss to penalize the anisotropy, eventually penalizing the covariance matrix (a la VICReg), and then the SIGReg method as a computationally much cheaper way to approximate the full covariance.<p>As another commenter pointed out it’s also similar to the InfoNCE/contrastive learning objectives. Where terms to increase uniformity (spread out evenly) on the hyper sphere were added. Like the SimCSE (Gao 2021) paper or the excellent alignment/uniformity breakdown from Wang & Isola 2020.<p>This proposed dispersion loss seems to be similar in that it pushes things apart by penalizing cosine similarity. Although this one works on the tokens within one sequence. Usually contrastive methods mean pool the sequences and then contrast against the other pooled sequences in the batch.</p>
]]></description><pubDate>Sat, 04 Jul 2026 05:21:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=48782836</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48782836</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48782836</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Alan Greenspan has died"]]></title><description><![CDATA[
<p>> Greenbacks … a form of fiat money<p><a href="https://en.wikipedia.org/wiki/Greenback_(1860s_money)" rel="nofollow">https://en.wikipedia.org/wiki/Greenback_(1860s_money)</a></p>
]]></description><pubDate>Mon, 22 Jun 2026 19:33:03 +0000</pubDate><link>https://news.ycombinator.com/item?id=48634919</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48634919</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48634919</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>Ah, I hadn’t thought about the generational aspect that’s interesting. The aesthetic totally makes sense to me when the music is intended for it / designed with it in mind, which I guess quite a lot of music is.<p>I particularly dislike when old intentionally-dynamic music is remastered to be “modernized” into a brick, which is sort of the opposite direction.<p>> Cinema mixes<p>I didn’t know about these, that’s neat! Makes sense that the levels can’t really be the same in my living room as a theater. Is it really a whole separate mix or just some compression in mastering?<p>I really hope that’s not another masterings collection rabbit hole I’m about to fall down haha. I’ll look out for some Dolby certified venues in my area too</p>
]]></description><pubDate>Fri, 12 Jun 2026 07:24:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=48500937</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48500937</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48500937</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>> 1 … they will often chose the more compressed/louder one<p>I’ve always been curious - but presumably that’s true even after volume matching?<p>> 3 Compressed sound works better in noisy environments and as background music<p>I’ve heard this is also why film and video game soundtracks are often very compressed, even when orchestral, because they have to fit in the background with dialog/sfx</p>
]]></description><pubDate>Fri, 12 Jun 2026 00:12:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=48498193</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48498193</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48498193</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>ReplayGain is nice - but note it doesn’t ’fix the compression’. Compression and dynamic range is about loud/quiet _within_ the track. ReplayGain just turns the volume up and down for the entire track, the point being so all your tracks play back at about the same level. It saves a preset on the volume knob for you essentially.<p>If you remember making a playlist where one song is suddenly much louder than the last, and you’re riding the volume knob on every other song, you’ll see why this is nice!</p>
]]></description><pubDate>Fri, 12 Jun 2026 00:05:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=48498128</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48498128</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48498128</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>It really does on some records, if you’re interested check out some comparisons on YouTube. Many times it’s subtle eq tweaks, granted, and that won’t much matter. But a lot of older rock and pop records for example go from being super dynamic and well produced to completely crushed with boosted bass and treble to ‘modernize’ the sound.<p>You can see some examples of how dynamic range (they don’t track ‘mastering’ overall) varies across releases on this site: <a href="https://dr.loudness-war.info/" rel="nofollow">https://dr.loudness-war.info/</a></p>
]]></description><pubDate>Thu, 11 Jun 2026 23:41:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=48497928</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48497928</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48497928</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>It’s completely true, when the vinyl has a different mastering. It can be a completely different version. It’s not because it’s vinyl</p>
]]></description><pubDate>Thu, 11 Jun 2026 19:02:24 +0000</pubDate><link>https://news.ycombinator.com/item?id=48494914</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48494914</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48494914</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Vinyl succumbs to Loudness War: more than just collateral damage (2025)"]]></title><description><![CDATA[
<p>Short answer, no not really. It won’t ever be as good as a proper uncompressed mastering</p>
]]></description><pubDate>Thu, 11 Jun 2026 19:00:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=48494886</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48494886</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48494886</guid></item><item><title><![CDATA[New comment by nighthawk454 in "πFS"]]></title><description><![CDATA[
<p>And also all the days you don’t, so, by itself not very meaningful. Especially since you can’t tell which one is right in advance. In some sense, so does a calendar</p>
]]></description><pubDate>Wed, 10 Jun 2026 20:46:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=48482460</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48482460</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48482460</guid></item><item><title><![CDATA[New comment by nighthawk454 in "Show HN: Rocksky – Music scrobbling and discovery on the AT Protocol"]]></title><description><![CDATA[
<p>I’m running MultiScrobbler to scrobble from multiple sources (Spotify/PlexAmp) to multiple sinks (Last.fm/ListenBrainz/Maloja). Looks like they already support writing to Rocksky!<p><a href="https://github.com/FoxxMD/multi-scrobbler" rel="nofollow">https://github.com/FoxxMD/multi-scrobbler</a></p>
]]></description><pubDate>Sun, 17 May 2026 02:06:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=48165467</link><dc:creator>nighthawk454</dc:creator><comments>https://news.ycombinator.com/item?id=48165467</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48165467</guid></item></channel></rss>