<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: sorenjan</title><link>https://news.ycombinator.com/user?id=sorenjan</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Mon, 17 Aug 2026 07:20:29 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=sorenjan" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by sorenjan in "Choosing an AI model: one prompt, 11 models, different results"]]></title><description><![CDATA[
<p>Just a few days ago I looked up a small coffee shop on Google maps which claimed it was closed, but their own website had different opening hours showing it was open. It's a seasonal business and Google had old data. Businesses should own their own communication and not be dependent on big tech companies.</p>
]]></description><pubDate>Fri, 14 Aug 2026 11:03:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=49297137</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49297137</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49297137</guid></item><item><title><![CDATA[New comment by sorenjan in "GDID Windows – Cut the tracker that follows you even under VPN"]]></title><description><![CDATA[
<p>This seems like something that should be easy to confirm, but I haven't seen anyone do it. Do they keep a database of every website visit all Edge users make?</p>
]]></description><pubDate>Sat, 25 Jul 2026 17:48:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=49049832</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49049832</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49049832</guid></item><item><title><![CDATA[New comment by sorenjan in "GDID Windows – Cut the tracker that follows you even under VPN"]]></title><description><![CDATA[
<p>That wouldn't explain how they connected the GDID to the visit to the retailers website<p>> Three hours later, the same GDID visited the retailer’s own website, through the same Tzulo proxy address used to set up the ngrok account.</p>
]]></description><pubDate>Sat, 25 Jul 2026 17:44:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=49049801</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49049801</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49049801</guid></item><item><title><![CDATA[New comment by sorenjan in "GDID Windows – Cut the tracker that follows you even under VPN"]]></title><description><![CDATA[
<p>> Microsoft provided the FBI with the history of IP addresses tied to that specific GDID.<p>This article, and most articles about this, doesn't explain where FBI got that GDID from. Ok, Microsoft has a list of IP addresses that has been used by a computer with a certain GDID, but FBI needs to get the GDID in the first place, and then try to bind that to a person.<p>I found another article that explains the process a bit better:<p>> Stokes got caught because he used the same Windows device for everything, and the GDID stitched all of it back together after the fact.<p>> Scattered Spider members phoned the jewelry retailer’s IT help desk from Google Voice numbers, posed as locked out employees, and talked support staff into resetting three accounts, two with administrator privileges. From there they installed a tunneling tool called ngrok to get past the retailer’s network defenses, moved roughly 77 gigabytes of data to Amazon cloud storage using ngrok [...]<p>> Investigators later subpoenaed ngrok and found the account used in the attack had been created on May 12, 2025, at 19:21 UTC from a VPN proxy IP address run by Tzulo, a hosting provider. The IP was a dead end. VPN proxies do that. But the GDID is built different.<p>> Microsoft’s records showed that at that exact same minute, a Windows device carrying GDID g:6755467234350028 had visited the ngrok signup page. Three hours later, the same GDID visited the retailer’s own website, through the same Tzulo proxy address used to set up the ngrok account. It gave the FBI a device, that don’t rotate the way VPN exit nodes do.<p><a href="https://www.windowslatest.com/2026/07/10/you-cant-fully-disable-microsofts-gdid-windows-11-tracker-but-these-settings-limit-what-it-captures/" rel="nofollow">https://www.windowslatest.com/2026/07/10/you-cant-fully-disa...</a><p>Although this doesn't explain where Microsoft got that traffic data from. How do Microsoft know which sites a computer visit?</p>
]]></description><pubDate>Sat, 25 Jul 2026 17:21:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=49049528</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49049528</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49049528</guid></item><item><title><![CDATA[New comment by sorenjan in "Test-time training 3D reconstruction"]]></title><description><![CDATA[
<p>I think the project page is a better URL, it contains more info and demos for what this is and it has a link to the code anyway:<p><a href="https://rover-xingyu.github.io/TTT3R/" rel="nofollow">https://rover-xingyu.github.io/TTT3R/</a><p>This looks really interesting, especially when combined with gaussian splats.</p>
]]></description><pubDate>Thu, 23 Jul 2026 13:57:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=49021694</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49021694</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49021694</guid></item><item><title><![CDATA[New comment by sorenjan in "Terence Tao's ChatGPT conversation about the Jacobian Conjecture counterexample"]]></title><description><![CDATA[
<p>Mathematical breakthrough, genius, trending on Artstation.</p>
]]></description><pubDate>Wed, 22 Jul 2026 18:54:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=49011673</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=49011673</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49011673</guid></item><item><title><![CDATA[New comment by sorenjan in "Transcribe.cpp"]]></title><description><![CDATA[
<p>That's up to you of course, but is it that much more work to compile the cli binary at the same time as you compile the libraries? How am I supposed to actually use the Cuda binaries available in the releases section, through a separately downloaded Python wheel?</p>
]]></description><pubDate>Mon, 20 Jul 2026 15:33:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=48980309</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48980309</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48980309</guid></item><item><title><![CDATA[New comment by sorenjan in "Annoying and alarming things about OpenCode"]]></title><description><![CDATA[
<p>For some reason OpenCode doesn't find my local models through LM Studio. Instead it shows three other models that I don't have. 
I guess I should add the models myself in opencode.json, but I generally dislike hardcoding things when computers are really good at listing things. The model list is right there in the API, please just use it.</p>
]]></description><pubDate>Mon, 20 Jul 2026 15:04:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=48979892</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48979892</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48979892</guid></item><item><title><![CDATA[New comment by sorenjan in "Transcribe.cpp"]]></title><description><![CDATA[
<p>Why not include transcribe-cli in the release archives to make it easier to use for people that can't compile it themselves? I downloaded the Cuda version but it's only the dll files, I don't really want to have to deal with Cuda SDK, I doubt most people want to.</p>
]]></description><pubDate>Sun, 19 Jul 2026 13:33:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=48968101</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48968101</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48968101</guid></item><item><title><![CDATA[New comment by sorenjan in "We charge $10k a week to delete AI-generated code"]]></title><description><![CDATA[
<p>- What you do at Initech is you take the specifications from the customer and bring them down to the prompt engineers?<p>- Yes, yes that's right.<p>- Well then I just have to ask why can't the customers take them directly to the vibe coding software people?<p>- Well, I'll tell you why... because... engineers are not good at dealing with customers...<p>- So you physically take the specs from the customer?<p>- Well... No. My secretary does that... or they're faxed.<p>- So then you must physically bring them to the software people?<p>- Well... No. ah sometimes.<p>- What would you say you do here?<p>- Look I already told you, I deal with the @#$% customers so the prompt engineers don't have to. I have people skills! I am good at dealing with people, can't you understand that? WHAT THE HELL IS WRONG WITH YOU PEOPLE?!</p>
]]></description><pubDate>Wed, 08 Jul 2026 02:15:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=48826690</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48826690</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48826690</guid></item><item><title><![CDATA[New comment by sorenjan in "CarPlay Is Additive"]]></title><description><![CDATA[
<p>This is the problem with so many modern companies. Instead of focusing on building great products that their customers will enjoy using, they're trying to build relationships with users.<p>Maybe I'm overthinking things, but to me "user" implies a different more subordinate role where you take what you get, there's no "the customer is always right" when it comes to "users".</p>
]]></description><pubDate>Fri, 03 Jul 2026 12:16:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=48774120</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48774120</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48774120</guid></item><item><title><![CDATA[New comment by sorenjan in "Grok's Traffic Is Mostly Driven by Adult Content"]]></title><description><![CDATA[
<p>> It turns out that NSFW activities account for "well over half" of Grok's traffic, according to a new report in The Information that cites two former employees of the SpaceX-owned company.<p><a href="https://www.engadget.com/2202099/the-most-popular-grok-feature-is-apparently-exactly-what-you-think/" rel="nofollow">https://www.engadget.com/2202099/the-most-popular-grok-featu...</a></p>
]]></description><pubDate>Tue, 30 Jun 2026 20:54:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=48739068</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48739068</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48739068</guid></item><item><title><![CDATA[New comment by sorenjan in "Grok's Traffic Is Mostly Driven by Adult Content"]]></title><description><![CDATA[
<p>> Well over half of Grok's overall traffic is driven by pornographic images and videos, adult role-play chats or other such activity, according to the report<p>This is worth $2.25 trillion?</p>
]]></description><pubDate>Tue, 30 Jun 2026 20:22:40 +0000</pubDate><link>https://news.ycombinator.com/item?id=48738664</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48738664</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48738664</guid></item><item><title><![CDATA[Grok's Traffic Is Mostly Driven by Adult Content]]></title><description><![CDATA[
<p>Article URL: <a href="https://www.forbes.com/sites/maryroeloffs/2026/06/24/groks-traffic-is-mostly-driven-by-adult-content-report-says/">https://www.forbes.com/sites/maryroeloffs/2026/06/24/groks-traffic-is-mostly-driven-by-adult-content-report-says/</a></p>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48738652">https://news.ycombinator.com/item?id=48738652</a></p>
<p>Points: 25</p>
<p># Comments: 8</p>
]]></description><pubDate>Tue, 30 Jun 2026 20:21:28 +0000</pubDate><link>https://www.forbes.com/sites/maryroeloffs/2026/06/24/groks-traffic-is-mostly-driven-by-adult-content-report-says/</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48738652</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48738652</guid></item><item><title><![CDATA[New comment by sorenjan in "Anthropic says Alibaba illicitly extracted Claude AI model capabilities"]]></title><description><![CDATA[
<p>I think companies should do this too, in a smaller scale. Proxy all LLM traffic to and from your employees, and use it to fine tune a smaller local model.</p>
]]></description><pubDate>Thu, 25 Jun 2026 11:40:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=48671980</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48671980</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48671980</guid></item><item><title><![CDATA[New comment by sorenjan in "Anthropic says Alibaba illicitly extracted Claude AI model capabilities"]]></title><description><![CDATA[
<p>Doesn't "real" distillation use the logits instead of the final tokens? I would classify this more like using a model to generate synthetic training data.</p>
]]></description><pubDate>Thu, 25 Jun 2026 11:34:22 +0000</pubDate><link>https://news.ycombinator.com/item?id=48671935</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48671935</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48671935</guid></item><item><title><![CDATA[New comment by sorenjan in "Show HN: TikZ Editor – WYSIWYG editor for figures in LaTeX"]]></title><description><![CDATA[
<p>Looks really nice. You might consider adding some presets to make it easier to get started, like some common neural net architectures and other use cases for TikZ.</p>
]]></description><pubDate>Tue, 23 Jun 2026 15:35:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=48646718</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48646718</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48646718</guid></item><item><title><![CDATA[New comment by sorenjan in "VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO"]]></title><description><![CDATA[
<p>How would you best utilize a model like this for coding? I take it it's not meant for vibe coding a full app, and the reasoning probably makes it unsuitable for autocomplete. Would you use it to implement specific functions? I looked at one of the coding benchmarks used, Live Code Bench, and it seems to be problem descriptions with sample input and output, and then a solution with a single function or class.<p>Seems like a really good model to use in an IDE when you still want control over the code structure then.</p>
]]></description><pubDate>Tue, 23 Jun 2026 13:09:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=48644431</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48644431</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48644431</guid></item><item><title><![CDATA[New comment by sorenjan in "Running MicroVMs in Proxmox VE, the Easy Way"]]></title><description><![CDATA[
<p>Am I understanding the ballooning part right that this doesn't allocate all of the VM memory from the host until it's needed, and releases memory automatically when it's not needed anymore? So you can overprovision memory with multiple guests as long as the guests aren't using the memory at the same time?</p>
]]></description><pubDate>Sun, 21 Jun 2026 15:20:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=48619712</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48619712</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48619712</guid></item><item><title><![CDATA[New comment by sorenjan in "AI Compute Extensions (ACE) Specification"]]></title><description><![CDATA[
<p>AVX512 isn't available on most new CPUs, I'm guessing ACE will only be available on server CPUs for at least a couple of years at launch?</p>
]]></description><pubDate>Thu, 18 Jun 2026 03:44:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=48580547</link><dc:creator>sorenjan</dc:creator><comments>https://news.ycombinator.com/item?id=48580547</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48580547</guid></item></channel></rss>