<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: Casteil</title><link>https://news.ycombinator.com/user?id=Casteil</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 10 Jun 2026 17:55:25 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=Casteil" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by Casteil in "Running local models on an M4 with 24GB memory"]]></title><description><![CDATA[
<p>You can expect around 55-60t/s with Qwen3.5:35b-a3b or gemma4:26b-a4b Q4</p>
]]></description><pubDate>Mon, 11 May 2026 17:10:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=48097702</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=48097702</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48097702</guid></item><item><title><![CDATA[New comment by Casteil in "Running local models on an M4 with 24GB memory"]]></title><description><![CDATA[
<p>Qwen3.5/3.6 are really prone to looping and 'overthinking'. Gemma4 doesn't seem to have the same problems.</p>
]]></description><pubDate>Mon, 11 May 2026 16:55:35 +0000</pubDate><link>https://news.ycombinator.com/item?id=48097527</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=48097527</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48097527</guid></item><item><title><![CDATA[New comment by Casteil in "Running local models on an M4 with 24GB memory"]]></title><description><![CDATA[
<p>Why not 35b-a3b? ...or gemma4:26b-a4b? Both will be more capable than 9b and run at roughly similar (perhaps faster) speeds</p>
]]></description><pubDate>Mon, 11 May 2026 16:54:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=48097510</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=48097510</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48097510</guid></item><item><title><![CDATA[New comment by Casteil in "Ollama is now powered by MLX on Apple Silicon in preview"]]></title><description><![CDATA[
<p>It's gotten significantly better with the advent of local/offline MoE models (e.g. qwen3.5:35b-a3b, qwen3:30b-a3b, gpt-oss:20b-3.6b), which offer a good balance of prompt response speed and output quality.<p>'Dense' models of yesteryear (e.g. llama:70b, gemma2/3:27b) tend to be significantly slower by comparison, therefore, your hardware spends a lot more time 'maxed out' for a given prompt.</p>
]]></description><pubDate>Tue, 31 Mar 2026 18:55:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=47591833</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=47591833</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47591833</guid></item><item><title><![CDATA[New comment by Casteil in "Hetzner Prices increase 30-40%"]]></title><description><![CDATA[
<p>OVH is nearly doubling their some of their VPS pricing soon.</p>
]]></description><pubDate>Thu, 26 Feb 2026 03:17:33 +0000</pubDate><link>https://news.ycombinator.com/item?id=47161374</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=47161374</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47161374</guid></item><item><title><![CDATA[New comment by Casteil in "Microsoft Copilot AI Comes to LG TVs, and Can't Be Deleted"]]></title><description><![CDATA[
<p>100%. Roku's privacy policy is the most wildly invasive thing I've ever seen - basically everything that used to be just conspiracy theory.</p>
]]></description><pubDate>Tue, 16 Dec 2025 03:10:14 +0000</pubDate><link>https://news.ycombinator.com/item?id=46284303</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=46284303</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46284303</guid></item><item><title><![CDATA[New comment by Casteil in "X5.1 solar flare, G4 geomagnetic storm watch"]]></title><description><![CDATA[
<p>Nice. Looks like it was peaking around 21:00 - 22:00 local time, got pretty intense for a while.</p>
]]></description><pubDate>Wed, 12 Nov 2025 01:26:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=45895342</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45895342</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45895342</guid></item><item><title><![CDATA[New comment by Casteil in "GM will ditch Apple CarPlay and Android Auto on all its cars, not just EVs"]]></title><description><![CDATA[
<p>Yep.<p>Manufacturers ditching Apple Carplay/Android Auto support will, if not immediately, inevitably pursue rent-seeking behavior in the form of paid subscriptions for services people could otherwise just have for free (and likely better) via phone.</p>
]]></description><pubDate>Thu, 23 Oct 2025 01:28:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=45677191</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45677191</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45677191</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>...they're not. This is a release of a 14" with the base M5, alongside the other existing M4 Pro/Max models.<p>The Pro/Max rollout tends to lag behind by about 6 months.</p>
]]></description><pubDate>Wed, 15 Oct 2025 20:12:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597812</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597812</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597812</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>Good chance they'll introduce it with the upcoming M5 Pro/Max; the non-pro/max devices always tended to be a little lower spec all around.</p>
]]></description><pubDate>Wed, 15 Oct 2025 20:09:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597765</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597765</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597765</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>Pro/Max rollout tends to lag behind the 'base' by about 6 months<p>It used to be a little less 'weird' when the base M-chips were only available in the Air and 13" MBP.</p>
]]></description><pubDate>Wed, 15 Oct 2025 20:05:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597716</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597716</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597716</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>Correct. Sometimes much later.<p>Still no M4 Ultra Studio available.</p>
]]></description><pubDate>Wed, 15 Oct 2025 20:02:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597679</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597679</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597679</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>I wish they'd bring Space Gray back. Not a huge fan of Silver, and the 'Space Black' apparently tends to show smudges more.</p>
]]></description><pubDate>Wed, 15 Oct 2025 20:00:52 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597669</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597669</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597669</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>The non-pro/max chipped MBPs have always been a little 'lower spec' in several regards. There used to be a little more separation though, with the non-pro chips available only in the Air & 13" MBP, but back then people complained about Apple having 'too many models'...<p>I suspect the M5 Pro/Max chipped MBPs will bring some of these improvements you're looking for.</p>
]]></description><pubDate>Wed, 15 Oct 2025 19:58:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597646</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597646</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597646</guid></item><item><title><![CDATA[New comment by Casteil in "M5 MacBook Pro"]]></title><description><![CDATA[
<p>Part of me misses my OG base 14" M4 Pro. The battery on that thing was absolutely phenomenal - literal 12-14+ hours of real-world use. Not so much on the 14" M1 Max (64GB) that I upgraded to after about 2 yrs.<p>'Real-world idle' efficiency on the newer chips is the main reason I've got the (slight) itch to upgrade, but 64GB+ MBPs certainly don't come cheap.</p>
]]></description><pubDate>Wed, 15 Oct 2025 19:49:05 +0000</pubDate><link>https://news.ycombinator.com/item?id=45597515</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45597515</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45597515</guid></item><item><title><![CDATA[New comment by Casteil in "I'm spoiled by Apple Silicon but still love Framework"]]></title><description><![CDATA[
<p>Charging is a lossy process. Charging a battery, then using it to charge another battery isn't as efficient.<p>Plus, depending on how/where it's used, having to wait for it to recharge while connected to a power bank might be a non-starter. You also don't necessarily want to recharge while transporting in a bag either because of heat concerns.</p>
]]></description><pubDate>Tue, 23 Sep 2025 19:43:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=45351824</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=45351824</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45351824</guid></item><item><title><![CDATA[New comment by Casteil in "Notion releases offline mode"]]></title><description><![CDATA[
<p>Came here to say as much. No proprietary 'lock-in' with Obsidian, just plain .md, and honestly (if I recall correctly - it's been years now) I felt like I struggled using Notion's 'fancier' features. It seemed like copying/pasting out of it was often problematic.<p>Like others are saying, though, Notion is much more geared for collaboration. I couldn't care much less about that, so here I am with Obsidian, still, years later.</p>
]]></description><pubDate>Wed, 20 Aug 2025 00:03:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=44957391</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=44957391</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44957391</guid></item><item><title><![CDATA[New comment by Casteil in "Upgrading an M4 Pro Mac mini's storage for half the price"]]></title><description><![CDATA[
<p>>Esp when the lowest setting isn't even meant to be purchased<p>All current models ship with 16GB so this isn't really true anymore.. 16GB will be good for most folks for years and years (most folks aren't doing heavy virtualization or other highly ram-intensive use cases).<p>>and the specs are so low they will underperform, or be obsolete in a few years.<p>...what?? People are still using OG M1 macs because they're still very capable machines. These things - especially the M4s - have crazy amounts of performance headroom.</p>
]]></description><pubDate>Sat, 12 Jul 2025 19:16:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=44544338</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=44544338</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44544338</guid></item><item><title><![CDATA[New comment by Casteil in "What Google Translate can tell us about vibecoding"]]></title><description><![CDATA[
<p>> It then writes a prompt based on your answers, sends the text to models from OpenAI, Anthropic, and Google, creates a combined draft from the three translations, and then sends that draft back to the three models for several rounds of revision, checking, and polishing.<p>RIP global power consumption</p>
]]></description><pubDate>Wed, 18 Jun 2025 19:40:41 +0000</pubDate><link>https://news.ycombinator.com/item?id=44312735</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=44312735</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44312735</guid></item><item><title><![CDATA[New comment by Casteil in "Have I Been Pwned 2.0"]]></title><description><![CDATA[
<p>Might be related to the 'breach disclosed/discovered' date? I noticed some of mine appeared out of chronological order, too.</p>
]]></description><pubDate>Mon, 19 May 2025 23:48:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=44036235</link><dc:creator>Casteil</dc:creator><comments>https://news.ycombinator.com/item?id=44036235</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44036235</guid></item></channel></rss>