<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: lgessler</title><link>https://news.ycombinator.com/user?id=lgessler</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 23 Sep 2026 01:56:27 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=lgessler" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by lgessler in "Claude Opus 5.5"]]></title><description><![CDATA[
<p>I should find information about the user's concern instead of just assuming.<p>The user is right. The outage is a real concern, and the issue is worse than we realized. Requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 encountered elevated error rates. Worth stating plainly: these are not just models — they are load bearing rungs on the software development tooling ladder, and a blocker on this level makes the outage really bite.<p>One decision that is yours to make, not mine: should an email be drafted to Anthropic support? This issue has teeth, and a canonical handoff can land us where the main gate is no longer breaking silently.</p>
]]></description><pubDate>Tue, 22 Sep 2026 17:17:07 +0000</pubDate><link>https://news.ycombinator.com/item?id=49804742</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49804742</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49804742</guid></item><item><title><![CDATA[New comment by lgessler in "Claude Opus 5.5"]]></title><description><![CDATA[
<p>I thought about taking a shot every time Opus 5 said "load bearing", "bites", "teeth" (real oral fixation it had), "real {concern,issue,problem,...}" and realized I'd be dead of acute alcohol poisoning by lunch if I did so.</p>
]]></description><pubDate>Tue, 22 Sep 2026 17:07:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=49804575</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49804575</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49804575</guid></item><item><title><![CDATA[New comment by lgessler in "Claude Fable 5.1 and Claude Mythos 5.1"]]></title><description><![CDATA[
<p>Spelling is no guide for pronunciation here, though. In North American and Commonwealth dialects of English I don't think there's a context in which the <f> in <of> is really pronounced as a [f]. It is rather a [v].</p>
]]></description><pubDate>Wed, 02 Sep 2026 12:25:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=49535221</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49535221</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49535221</guid></item><item><title><![CDATA[New comment by lgessler in "Claude Fable 5.1 and Claude Mythos 5.1"]]></title><description><![CDATA[
<p>Not sure where you're from but in my dialect (North American) it's more common than not to have _have_ realized as [əv] ("uhv") in contexts like _should have_, _could have_ (but not _I have a car_, where it has to be the full [hæv]). Only in deliberately enunciated speech do I feel like I'd expect [hæv] in the former kind of context. So it's an understandable mistake to make.</p>
]]></description><pubDate>Wed, 02 Sep 2026 12:23:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=49535202</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49535202</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49535202</guid></item><item><title><![CDATA[New comment by lgessler in "More than half of adults in U.S. say they lack basic statistical understanding"]]></title><description><![CDATA[
<p>true for XCOM in all difficulties except the hardest iirc</p>
]]></description><pubDate>Wed, 26 Aug 2026 11:49:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=49447369</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49447369</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49447369</guid></item><item><title><![CDATA[New comment by lgessler in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"]]></title><description><![CDATA[
<p>I'm confused. I think you're confusing Gemini and Gemma. Gemini is Google's frontier offering which is closed-weight, API-only, like most frontier models. Gemma is Google's open-weight offering focused on deployment on consumer and edge hardware.</p>
]]></description><pubDate>Thu, 06 Aug 2026 15:26:55 +0000</pubDate><link>https://news.ycombinator.com/item?id=49198024</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=49198024</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49198024</guid></item><item><title><![CDATA[New comment by lgessler in "The state of open source AI"]]></title><description><![CDATA[
<p>You're saying it's important to have up-to-date facts stored in parametric knowledge? It seems to me like that's grown less and less important as agentic capabilities have grown. Even if a frontier model doesn't know something, if it's out there, it can easily find it through tool use.</p>
]]></description><pubDate>Fri, 17 Jul 2026 16:36:56 +0000</pubDate><link>https://news.ycombinator.com/item?id=48949319</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=48949319</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48949319</guid></item><item><title><![CDATA[New comment by lgessler in "Performance per dollar is getting faster and cheaper"]]></title><description><![CDATA[
<p>Accuracy isn't a meaningful metric here without reference to a specific task.</p>
]]></description><pubDate>Sat, 04 Jul 2026 05:12:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=48782799</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=48782799</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48782799</guid></item><item><title><![CDATA[New comment by lgessler in "4× RTX Pro 6000 Blackwell on Water, and the One Card That Wouldn't Behave"]]></title><description><![CDATA[
<p>Not that this really takes away from the substance of the article, but the first two paragraphs are giving heavy Claude smell. Semicolons, em dashes, "That sequencing matters"... I guess I'm just a little surprised that anyone could be arsed to take on a hardware project like this but can't be arsed to write their own introduction.</p>
]]></description><pubDate>Tue, 16 Jun 2026 15:49:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=48557171</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=48557171</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48557171</guid></item><item><title><![CDATA[New comment by lgessler in "Artificial intelligence is not conscious – Ted Chiang"]]></title><description><![CDATA[
<p>Raphaël Millière has a very useful term for this kind of vacuous dismissal, the redescription fallacy (<a href="https://arxiv.org/pdf/2401.03910" rel="nofollow">https://arxiv.org/pdf/2401.03910</a>, page 9):<p>> Recent debates have been clouded by a misleading inference pattern, which we term the “Redescription Fallacy.” This fallacy arises when critics argue that a system cannot model a particular cognitive capacity, simply because its operations can be explained in less abstract and more deflationary terms. In the present context, the fallacy manifests in claims that LLMs could not possibly be good models of some cognitive capacity  because their operations merely consist in a collection of statistical calculations, or linear algebra operations, or next-token predictions. Such arguments are only valid if accompanied by evidence demonstrating that a system, defined in these terms, is inherently incapable of implementing . To illustrate, consider the flawed logic in asserting that a piano could not possibly produce harmony because it can be described as a collection of hammers striking strings, or (more pointedly) that brain activity could not possibly implement cognition because it can be described as a collection of neural firings. The critical question is not whether the operations of an LLM can be simplistically described in non-mental terms, but whether these operations, when appropriately organized, can implement the same processes or algorithms as the mind, when described at an appropriate level of computational abstraction.</p>
]]></description><pubDate>Thu, 04 Jun 2026 00:52:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=48392299</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=48392299</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48392299</guid></item><item><title><![CDATA[New comment by lgessler in "Qwen3.6-27B: Flagship-Level Coding in a 27B Dense Model"]]></title><description><![CDATA[
<p>I'll be really interested to hear qualitative reports of how this model works out in practice. I just can't believe that a model this small is actually as good as Opus, which is rumored to be about two orders of magnitude larger.</p>
]]></description><pubDate>Wed, 22 Apr 2026 19:29:57 +0000</pubDate><link>https://news.ycombinator.com/item?id=47868161</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=47868161</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47868161</guid></item><item><title><![CDATA[New comment by lgessler in "Writing Lisp is AI resistant and I'm sad"]]></title><description><![CDATA[
<p>Is Java or Haskell any closer to human language?</p>
]]></description><pubDate>Sun, 05 Apr 2026 04:24:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=47646092</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=47646092</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47646092</guid></item><item><title><![CDATA[New comment by lgessler in "If you thought the code writing speed was your problem; you have bigger problems"]]></title><description><![CDATA[
<p>Has everyone always nailed their implementation of every program on the first try? Of course not. Probably what happens most times is you first complete something that sorta works and then iterate from there by modifying code, executing, observing, and looping back to the beginning. You can wonder about ultimately how much of your time/energy is consumed by the "typing code" part, and there's surely a wide range of variation there by individual and situation, but it's undeniable that it is a part of the core iteration loop for building software.<p>I don't understand why GP's comment is so controversial. GP is not denying that you should maybe think a little before a key hits the keyboard as many commenters seem to suppose. Both can be true.</p>
]]></description><pubDate>Tue, 17 Mar 2026 19:16:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=47416925</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=47416925</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47416925</guid></item><item><title><![CDATA[New comment by lgessler in "Show HN: Han – A Korean programming language written in Rust"]]></title><description><![CDATA[
<p>I know this is mostly about keyword substitution but it still tickles me that you still write f(x) in this language and not (x)f given that Korean is SOV but I guess that's just how you notate that no matter what cultural context you're in. Hadn't ever considered that the convention of writing a function before its arguments might have been a contingency of this notation being developed by speakers of SVO languages.</p>
]]></description><pubDate>Sun, 15 Mar 2026 00:58:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=47383111</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=47383111</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47383111</guid></item><item><title><![CDATA[New comment by lgessler in "Claude Code is being dumbed down?"]]></title><description><![CDATA[
<p>Let's be real here, regardless of what Boris thinks, this decision is not in his hands.</p>
]]></description><pubDate>Thu, 12 Feb 2026 02:02:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=46984011</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=46984011</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=46984011</guid></item><item><title><![CDATA[New comment by lgessler in "Cognitive load is what matters"]]></title><description><![CDATA[
<p>Novels are fictional too. So long as they're not taken too literally, archetypes can be helpful mental prompts.</p>
]]></description><pubDate>Sat, 30 Aug 2025 17:15:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=45076319</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=45076319</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=45076319</guid></item><item><title><![CDATA[New comment by lgessler in "Gemma 3 270M re-implemented in pure PyTorch for local tinkering"]]></title><description><![CDATA[
<p>If you're really just doing traditional NER (identifying non-overlapping spans of tokens which refer to named entities) then you're probably better off using encoder-only (e.g. <a href="https://huggingface.co/dslim/bert-large-NER" rel="nofollow">https://huggingface.co/dslim/bert-large-NER</a>) or encoder-decoder (e.g. <a href="https://huggingface.co/dbmdz/t5-base-conll03-english" rel="nofollow">https://huggingface.co/dbmdz/t5-base-conll03-english</a>) models. These models aren't making headlines anymore because they're not decoder-only, but for established NLP tasks like this which don't involve generation, I think there's still a place for them, and I'd assume that at equal parameter counts they quite significantly outperform decoder-only models at NER, depending on the nature of the dataset.</p>
]]></description><pubDate>Wed, 20 Aug 2025 21:11:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=44966516</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=44966516</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44966516</guid></item><item><title><![CDATA[New comment by lgessler in "FFmpeg 8.0 adds Whisper support"]]></title><description><![CDATA[
<p>I recommend having a look at 16.3 onward here if you're curious about this: <a href="https://web.stanford.edu/~jurafsky/slp3/16.pdf" rel="nofollow">https://web.stanford.edu/~jurafsky/slp3/16.pdf</a><p>I'm not familiar with Whisper in particular, but typically what happens in an ASR model is that the decoder, speaking loosely, sees "the future" (i.e. the audio after the chunk it's trying to decode) in a sentence like this, and also has the benefit of a language model guiding its decoding so that grammatical productions like "I like ice cream" are favored over "I like I scream".</p>
]]></description><pubDate>Wed, 13 Aug 2025 11:36:28 +0000</pubDate><link>https://news.ycombinator.com/item?id=44887204</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=44887204</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44887204</guid></item><item><title><![CDATA[New comment by lgessler in "Grok: Searching X for "From:Elonmusk (Israel or Palestine or Hamas or Gaza)""]]></title><description><![CDATA[
<p>In my (poor) understanding, this can depend on hardware details. What are you running your models on? I haven't paid close attention to this with LLMs, but I've tried very hard to get non-deterministic behavior out of my training runs for other kinds of transformer models and was never able to on my 2080, 4090, or an A100. PyTorch docs have a note saying that in general it's impossible: <a href="https://docs.pytorch.org/docs/stable/notes/randomness.html" rel="nofollow">https://docs.pytorch.org/docs/stable/notes/randomness.html</a><p>Inference on a generic LLM may not be subject to these non-determinisms even on a GPU though, idk</p>
]]></description><pubDate>Fri, 11 Jul 2025 01:10:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=44527453</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=44527453</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=44527453</guid></item><item><title><![CDATA[New comment by lgessler in "How University Students Use Claude"]]></title><description><![CDATA[
<p>Sure, this is a common sentiment, and one that works for some courses. But for others (introductory programming, say) I have a really hard time imagining an assignment that could not be one-shot by an LLM. What can someone with 2 weeks of Python experience do that an LLM couldn't? The other issue is that LLMs are, for now, periodically increasing in their capabilities, so it's anyone's guess whether this is actually a sustainable attitude on the scale of years.</p>
]]></description><pubDate>Thu, 10 Apr 2025 02:44:21 +0000</pubDate><link>https://news.ycombinator.com/item?id=43640148</link><dc:creator>lgessler</dc:creator><comments>https://news.ycombinator.com/item?id=43640148</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43640148</guid></item></channel></rss>