<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: urbsgpw</title><link>https://news.ycombinator.com/user?id=urbsgpw</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Tue, 25 Aug 2026 05:48:23 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=urbsgpw" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by urbsgpw in "Cursor removed cost information from the usage page and CSV export"]]></title><description><![CDATA[
<p>Huh, never thought about that 2nd point - I'm also transitioning to pi, but didn't think to change to neovim as well. I guess your logic holds for claude code users as well though.</p>
]]></description><pubDate>Sat, 01 Aug 2026 17:19:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=49136383</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=49136383</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49136383</guid></item><item><title><![CDATA[New comment by urbsgpw in "Is AI reasoning right for the wrong reasons?"]]></title><description><![CDATA[
<p>But isn't it the case that we can't reach this safeguard with the current architecture? I remember Karpathy making an interesting point 2 years (cca) back, that I would summarize somehow like this: the mechanics behind every LLM answer are the same, what you then call hallucination is more or less a consequence of whether or not the answer was factually correct/useful.<p>Which would mean, as is so often the case, that the "killer feature" of the LLMs is also its biggest weakness and the two can't be disentangled. Now, we are inventive creatures and we might come up with a remedy for these issues, but what you basically see so far is more guardrails, the use of harnesses and building a whole bunch of infrastructure around the LLMs to get useful work out of them.<p>Which, btw is not a critique, I do it as well and it's a fun engineering challenge.</p>
]]></description><pubDate>Sat, 01 Aug 2026 06:52:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=49131794</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=49131794</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49131794</guid></item><item><title><![CDATA[New comment by urbsgpw in "Nvidia, Microsoft, Meta warn against overregulating open-weight models"]]></title><description><![CDATA[
<p>Maybe less shocking, but still, so is google. I know they are one of the labs, but that being said they rely far less on gemini for revenue as opposed to the big 2.</p>
]]></description><pubDate>Fri, 24 Jul 2026 19:40:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=49040681</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=49040681</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49040681</guid></item><item><title><![CDATA[New comment by urbsgpw in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"]]></title><description><![CDATA[
<p>It seems like they're sticking to a static pricing plan that was made when the only relevant competition were the US labs (im not counting deepseek 2025 as serious competition -> glm and then kimi on the other hand, now that's a different story).</p>
]]></description><pubDate>Tue, 21 Jul 2026 21:04:10 +0000</pubDate><link>https://news.ycombinator.com/item?id=48998289</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48998289</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48998289</guid></item><item><title><![CDATA[New comment by urbsgpw in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"]]></title><description><![CDATA[
<p>This was exactly my thought process. Coupled with a less idealistic one: mid 2025 if you looked arena ELO scores google seemed to be dominating and I was sure the trend would continue. But they really dropped the ball on coding and tool use.<p>That being said, controlling android and apple mobile devices is kind of a big deal. And their video models are still top notch.</p>
]]></description><pubDate>Tue, 21 Jul 2026 21:01:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=48998259</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48998259</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48998259</guid></item><item><title><![CDATA[New comment by urbsgpw in "Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber"]]></title><description><![CDATA[
<p>Exactly my experience. I'm building an AI document-extraction platform, so I had to benchmark a bunch of models — on the cost:quality:latency:adherence picture, flash wins hands down for structured extraction. (Caveat: I've only tested the three US labs and Mistral.). So like u said, totally viable in prod for a relatively static tool. Didn't build the tool suite with gemini, but if you use service mode it currently mainly runs on flash.<p>Haven't done any serious coding work with the flash models though — but I'm seeing more and more HN comments from people who seem to have picked it up for that in the last couple of months.</p>
]]></description><pubDate>Tue, 21 Jul 2026 20:56:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=48998180</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48998180</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48998180</guid></item><item><title><![CDATA[New comment by urbsgpw in "Mozilla: The state of open source AI"]]></title><description><![CDATA[
<p>Haven't been following the articles and snippets we get from these labs about training their models for a while. But I'm guessing the latest chinese models are way less based on distilling? If not, then your speed of progress is still limited by the two labs (which we are collectively, in various forms subsidizing).</p>
]]></description><pubDate>Fri, 17 Jul 2026 15:31:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=48948595</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48948595</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48948595</guid></item><item><title><![CDATA[New comment by urbsgpw in "Mozilla: The state of open source AI"]]></title><description><![CDATA[
<p>I use hermes only ever saw their repo. Atrocious. I was sure you were exaggerating.</p>
]]></description><pubDate>Fri, 17 Jul 2026 15:29:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=48948566</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48948566</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48948566</guid></item><item><title><![CDATA[New comment by urbsgpw in "A Beautiful Theory Falls to Ugly Data"]]></title><description><![CDATA[
<p>In other words, theory does not conform to reality, i.e. ugly data. Being an academically trained economist, not surprised.</p>
]]></description><pubDate>Thu, 16 Jul 2026 14:49:51 +0000</pubDate><link>https://news.ycombinator.com/item?id=48935391</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48935391</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48935391</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok Build is open source"]]></title><description><![CDATA[
<p>This.<p>When I was less into tech (2010-2015), I hated how everyone fawned about Elon (remember Hollywood casting him in iron man?). As I started to transition into tech I remember being impressed by the design principles of his companies (simplify, remove complexity).<p>But I am absolutely baffled how his detractors don't see that exactly what you mentioned, that they are part of subset of society, with very strong opinions (think race, economics, religion) and can't fathom somebody having a different opinion without that person being immoral. And the worst part is what you mention at the very end: the mental policing of this group in the past 10-15 years. I liken it to a religious sect (ironically, even though they hate Christianity, probably closest to a new age christian sect).<p>I've had this discussion up here last week pointing out the vitriol Elon gets is, to my mind, for the wrong reasons (and the reasons are exactly what you mention).</p>
]]></description><pubDate>Thu, 16 Jul 2026 06:49:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=48931133</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48931133</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48931133</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>While a valid point, I don't buy HN doesn't have a non-neutral ideological preference.<p>I completely agree his foray was a fiasco (and his shareholders paid for it) and your colourful portrayal is more accurate than my shorthand, but I would argue that the comments would have been fewer (on this portal, maybe not on foxnews.com) if he had done the same thing in the Biden admin.</p>
]]></description><pubDate>Thu, 09 Jul 2026 13:42:03 +0000</pubDate><link>https://news.ycombinator.com/item?id=48845704</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48845704</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48845704</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>No. I'm saying people make this whole charade about being rational where in reality we are much more preference-driven. And usually the algorithm is: 1. do i agree with this person and 2. lets build arguments that support the affiliation and make it sound objective.<p>I will admit this is not always the case. But humans weren't built for consistency. And my point is merely (was making it to another commentator in this thread), that Elon gets more flak up here not because he is inherently less moral than, say Larry Page, but because more people on HN are ideologically closer to the other side of the political spectrum. Which, I will again reiterate, is fine. But then I would expect (or actually, see first paragraph - I wouldn't) that the vitriol would be consistenly dished. But it isn't. Now to be sure, partly this is due to Elon's move into politics and his personality, but I doubt he would have got the same amount of hate if he went into politics in the Biden administration.</p>
]]></description><pubDate>Thu, 09 Jul 2026 13:38:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=48845663</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48845663</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48845663</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>I agree, it was whataboutism. And i completely agree with the rest of your statement as well. My point was just that Elon gets more flak here than the rest of them (And it's his fault for going into politics, to be sure) and that this is due to the general ideological leaning of HN being to the dems side. That's all. Which is also completely valid and to some degree expected - but irksome whenever there is a Tesla/SpaceX topic.<p>As for the rest, especially in the US (im from the EU), we should invest in everything you mentioned and I wouldnt mind taxing Elon and his gang more either.</p>
]]></description><pubDate>Thu, 09 Jul 2026 12:35:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844898</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844898</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844898</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>Well u can see other people from the democratic party doing similar salutes and the press reported it differently.<p>As to the first point: maybe it should. But look at how the aid actually works, what it funds (it was not politically neutral - btw nothing in the realm of society is) etc. I mean, not just USAID, a lot of these schemes are at the expense of people who pay the most taxes, to fund corporations that send aid to countries that otherwise wouldn't have been able to afford the commodities at the market price. Surely you can see that as a taxpayer you might support the dismantling of these orgs?<p>Now, I do agree that the ramp-down could have been 1-2 years and that the theatre around it shows the worst of Elon. And the current administration.</p>
]]></description><pubDate>Thu, 09 Jul 2026 12:11:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844615</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844615</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844615</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>Directly? I would love to see that study.<p>I'm saying a lot of the populace is against foreign aid. And that populace has the right to shut it off. And we live in nation states - the state giving the aid can always shut it off (for whatever reason). Granted, I see no reason to do it SO abruptly (and I agree that was an infantile show), but I am not convinced aid as such is a net benefit for humanity. At the very least this is something you can do an econometric analysis and discuss different policy choices.<p>Now, that being said, I do agree it would have probably made more sense, from an austerity point of view, to cut the military aid to Israel.</p>
]]></description><pubDate>Thu, 09 Jul 2026 12:07:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844580</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844580</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844580</guid></item><item><title><![CDATA[New comment by urbsgpw in "Bonnie Tyler, singer of Total Eclipse of the Heart, dies aged 75"]]></title><description><![CDATA[
<p>My mum had a cassette with some of her songs. We'd have it on for long trips. I loved the long version of Faster than the speed of night. it's basically just "carpe diem" in a different format, but i loved her voice and the slight melancholy and almost call to action that the song brought with it. Also, the video (of the shorter version) is peak 80's: <a href="https://www.youtube.com/watch?v=jm4CgwRxw3Y" rel="nofollow">https://www.youtube.com/watch?v=jm4CgwRxw3Y</a></p>
]]></description><pubDate>Thu, 09 Jul 2026 11:56:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844455</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844455</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844455</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>Ya maybe. But gemini didn't want to make all white families (or even white nazi soldiers). Isn't that just as racist and more importantly, counterfactual? Or is that good, because we should all strive to make our personal lives look like The Odyssey?<p>The world is more nuanced (or should be). But up till Trump (who is a loathsome cheap crook, so Im not saying this in support, but stating a fact) the whole Silicon valley other than Karp and Thiel was basically one hivemind. And btw that's ok, they/you are allowed to have your worldviews. But don't mistake morality for preference similarity. Fine, you like your elves black, your models chinese, your religion from the Arab peninsula and your sexual preferences lean towards the rainbow (the cliche right wing characterization of a "lefty" in 2026), you have every right to have that view. And also every right to say Elon suck (and yes, he is marketing over matter, I agree, but he is the only serious westerd large scale industrialist). But then let's not pretend that the reason why Dario or Sam are "ok" isn't because you're lifestyle and worldviews are more aligned with them. And not because of an objective real metric which makes Elon bad and Altman better (example, pick any tech CEO/founder other than Karp or Luckey).</p>
]]></description><pubDate>Thu, 09 Jul 2026 11:48:38 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844359</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844359</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844359</guid></item><item><title><![CDATA[New comment by urbsgpw in "Grok 4.5"]]></title><description><![CDATA[
<p>I mean ok, starlink could be an issue. But the USAID defunding is an ideological stance with which you can, based on your own stance, agree or disagree. You could also argue that without USAID in the first place, there wouldn't be incentives for overpopulation?<p>My point being, it seems incredibly simplistic to just assume that aid is good and no aid is bad. This is just first order effects that you're looking at. Also, if you want to look at how real world systems operate (u like USAID, for example, but NATO, a western imperialist project, operates much the same), institutions rarely respond to slow change (i.e. evolution). That's the surest route to keep the status quo (once a government agency has been set up, it will rarely be wound down). Now maybe the status quo is preferable! But maybe it isn't. And maybe - most likely - it depends on subjective preferences and your Weltanschauung.<p>And I would argue that for all his marketing and missing deadlines, the man has changed the world for the better (my view). And I would also argue if he were a Biden and Harris supporter, this comment section would look completely different. And that tells us just as much about the HN crowd as it does about Elon. Now I won't go into his political views, but he obviously isn't alone with those views (for which I think HN believes are influenced by Elon, which is incredibly patronizing) and maybe, just maybe, there are valid reasons for those views.<p>As for the model itself - seems like a similar set of metrics we always get with SOTA and near SOTA models, they compare themselves to anthropic and they are usually way cheaper. But the combination of the harness (claude code) and the models makes their end product noticeably better than the competition (admittedly haven't tried codex). I'll definitely give it a try with pi if its on openrouter (currently using GLM 5.2 there mainly).</p>
]]></description><pubDate>Thu, 09 Jul 2026 11:34:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=48844211</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48844211</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48844211</guid></item><item><title><![CDATA[New comment by urbsgpw in "Car touchscreens are cheap, not good"]]></title><description><![CDATA[
<p>Uhm, I like Teslas a lot. The design and the philosophy of just minimizing the amount of components. But where I would draw the line is AC via touch screen (again, if Grok can now do it reliably, fine, otherwise, buttons please). Also, depending on the model, I know some models had turn signals in some weird places (above the head? or am I hallucinating?). Also to make your point: if it hadnt been for carplay and its android equivalent, legacy auto would have been cooked long ago. I never use my VW native apps.<p>My point being, I think we havent found the sweet spot yet.</p>
]]></description><pubDate>Mon, 06 Jul 2026 17:16:34 +0000</pubDate><link>https://news.ycombinator.com/item?id=48807650</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48807650</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48807650</guid></item><item><title><![CDATA[New comment by urbsgpw in "Car touchscreens are cheap, not good"]]></title><description><![CDATA[
<p>I gotta say, the one redeeming feature of Ferrari Luce, for me at least, was the interior. I don't dislike screens, I just hate the tesla-esque obsession (where, for them with FSD - for all the hate they get about it up here - it might actually make sense since u are gonna have a FSD+Grok car) with no buttons. I know buttons add cost, but going back to the Luce example again: you have a healthy sized screen (so u don't go to the pre-tesla days), but you also have wonderful buttons across the board.<p>Now, I know it's not a very representative car. But nobody said the buttons need to be as flashy or as numerous.</p>
]]></description><pubDate>Mon, 06 Jul 2026 16:47:22 +0000</pubDate><link>https://news.ycombinator.com/item?id=48807219</link><dc:creator>urbsgpw</dc:creator><comments>https://news.ycombinator.com/item?id=48807219</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48807219</guid></item></channel></rss>