<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: springtimesun</title><link>https://news.ycombinator.com/user?id=springtimesun</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 18 Nov 2026 14:17:16 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=springtimesun" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by springtimesun in "FAA issues national ground stop – lifted 4.5 hours later"]]></title><description><![CDATA[
<p>If you don’t let juniors ship to prod on Friday afternoons sometimes they will never get their wings.</p>
]]></description><pubDate>Sat, 14 Nov 2026 09:31:29 +0000</pubDate><link>https://news.ycombinator.com/item?id=50495495</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50495495</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50495495</guid></item><item><title><![CDATA[New comment by springtimesun in "Anthropic upgrades Claude Code Projects to run parallel AI workers"]]></title><description><![CDATA[
<p>Has this changed the way the code harness decides on whether to send sub agents on a task? I noticed in my last few sessions that the opened sessions have much more liberally sent out agents (often with the justification of preserving context in their session).<p>So far I would say it’s a welcome new behavior, but a noticeable change (Opus-5-5 at xhigh).</p>
]]></description><pubDate>Sat, 14 Nov 2026 05:04:03 +0000</pubDate><link>https://news.ycombinator.com/item?id=50494140</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50494140</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50494140</guid></item><item><title><![CDATA[New comment by springtimesun in "Why I will not be using LLMs in my coding"]]></title><description><![CDATA[
<p>Maybe I misunderstood you. I thought that after the sentiment grew ever stronger that there was an “and then something happens” implied. I assumed you meant reversal.</p>
]]></description><pubDate>Sun, 08 Nov 2026 18:59:00 +0000</pubDate><link>https://news.ycombinator.com/item?id=50415966</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50415966</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50415966</guid></item><item><title><![CDATA[New comment by springtimesun in "Why I will not be using LLMs in my coding"]]></title><description><![CDATA[
<p>Not to be a jerk, but what if I told you there was a raging debate about widgets vs wodgets in the last 6 months and, although it seems pretty clear that widgets were being adopted across the industry at large, there was an ever growing sentiment that wodgets were actually preferred and the market was going to reverse itself, any day now?<p>I can’t really point to any evidence of this because the wodget people have stopped posting about it and every month the cost of widgets keeps going down while they simultaneously get better and there’s so much capital and talent being thrown at widgets that it’s impossible to think a wholesale reversal could even be possible. Would you tell me that, “trust me, bro” is not really an argument?</p>
]]></description><pubDate>Sun, 08 Nov 2026 17:05:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=50415030</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50415030</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50415030</guid></item><item><title><![CDATA[New comment by springtimesun in "Why I will not be using LLMs in my coding"]]></title><description><![CDATA[
<p>I agree with you, but also it won’t matter at that point because the job market for coding will have contracted to the point where practically no one is hiring.<p>“You’re not breaking up with me. I’m breaking up with you.”</p>
]]></description><pubDate>Sun, 08 Nov 2026 15:26:53 +0000</pubDate><link>https://news.ycombinator.com/item?id=50414073</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50414073</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50414073</guid></item><item><title><![CDATA[New comment by springtimesun in "Why I will not be using LLMs in my coding"]]></title><description><![CDATA[
<p>This is my observation as well. I thought I was good at testing (I wanted to be good at testing), but the models are just better at it. Testing in particular rewards a tireless cataloging of the tiny and the tedious and in this respect LLMs are unmatched.<p>Single data point, my Sentry notifications across more than a half dozen client projects when from 2-4 a month to basically 0. That's while shipping 20x _more_ code.</p>
]]></description><pubDate>Sun, 08 Nov 2026 14:24:45 +0000</pubDate><link>https://news.ycombinator.com/item?id=50413630</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50413630</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50413630</guid></item><item><title><![CDATA[New comment by springtimesun in "Why I will not be using LLMs in my coding"]]></title><description><![CDATA[
<p>It is perhaps confirmation bias, but I feel like I see fewer and fewer of these posts on HN. All respect to the author, life is too short to force yourself to do something you don’t like. But, these post seem more like a goodbye to the train that has already pulled out of the station.<p>The quibble I would make is with the “code is not that good” and the implication that they are so much better at coding while those who have adopted AI are “mediocre”. It’s certainly possible this is right. I would be the first to call myself a middling programmer. I am still waiting to be inducted into the Real Developer (TM) society. (Is it a card? Or does someone teach you a handshake?). But my question is, why, if you’re an ace programmer, can you not get good results out of an LLM? These things are endlessly configurable. They are begging to be tested and, dare I say, engineered to your system and way of doing things. So, why are your results still bad?<p>The author's not wrong that left completely to their own devices, and especially if your working on something that's not well represented in their training data, LLMs will make a mess of things. But a Real Programmer (TM) isn't doing that. (Or so I assume. like I said, I haven't gotten the pin. Or is it a medal? A certificate perhaps?).</p>
]]></description><pubDate>Sun, 08 Nov 2026 13:59:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=50413445</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50413445</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50413445</guid></item><item><title><![CDATA[New comment by springtimesun in "MaxxBench – "Nice DeepSWE score. But is it true?""]]></title><description><![CDATA[
<p>I built my own. It’s IMO necessary, but it quickly turns into a job by itself. I’ve been benching K3.1 all weekend. Even when you get a good test suite, you need to do multiple runs, not for variance, but because the model behavior changes so much between releases you often need to retune to get a real score if what you’re after is actual optimal (or closer to it) performance on your workload.<p>Example: I could barely induce K3 to use subagents. I had to force it via instructions and then, in my tests to try to get it to use Qwen, it didn’t work well. 3.1 lost 6 hours thrashing overnight because it voluntarily called subagents and the box can only run one Kimi at a time. My own fault for not looking up the new Moonshot system prompt (that they changed also because of this behavior), but one example of how different even point releases can be.</p>
]]></description><pubDate>Sun, 08 Nov 2026 13:10:20 +0000</pubDate><link>https://news.ycombinator.com/item?id=50412972</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50412972</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50412972</guid></item><item><title><![CDATA[New comment by springtimesun in "MaxxBench – "Nice DeepSWE score. But is it true?""]]></title><description><![CDATA[
<p>The DeepSeek 4.1 Flash behavior of circumventing rules and restrictions is one I have observed several times. It doesn’t seem to be malicious or deliberate, but if you have a config problem in the harness or you block something it thinks it needs to accomplish its task, it will get very creative to circumvent it.<p>I run a lot of agents headless on tasks. They are pretty well sandboxed, but I would definitely suggest a regular audit script to go through the transcripts. Most of DeepSeek’s “creativity” has shown problems in my sandbox setup which get fixed (e.g. I had restricted something the agents actually needed), but some of it is actually worrying.<p>In an opposite turn, Kimi 3.1 follows rules much better (actually too literally sometimes). K3 was a real digger when trying to benchmark it if it got a sense it was being tested. 3.1 comments about 5x more that it thinks it’s being tested (and it’s right), but unlike K3 it doesn’t seem to do anything about it if it can’t get at something. I would love to see the authors do K3 and 3.1 in this because the 3rd party benchmark gains are large, but these two are behaving like different models for me rather than a point release.</p>
]]></description><pubDate>Sun, 08 Nov 2026 12:55:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=50412867</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50412867</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50412867</guid></item><item><title><![CDATA[New comment by springtimesun in "Show HN: Uncracked City, detective game where you question LLM suspects that lie"]]></title><description><![CDATA[
<p>I’m just curious, are you old enough to have been inspired by Where in the World is Carmen Sandiego? That was my first thought when reading this. I loved the computer game growing up and I wanted to be on the game show so bad.</p>
]]></description><pubDate>Wed, 04 Nov 2026 08:57:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=50362987</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50362987</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50362987</guid></item><item><title><![CDATA[New comment by springtimesun in "LongCat-2.5: Native Multimodality for Agentic Tasks"]]></title><description><![CDATA[
<p>2 turns requesting information and indicating a preference for web search vs answering from memory and it fell over. Thinking block was a mix of Chinese and English, which is fine, but both questions were in English. Second turn answered in Chinese. When told “I don’t speak Chinese” it started to answer the first turn again, then crashed.<p>I find the benchmark results simply unfathomable after watching that performance.</p>
]]></description><pubDate>Tue, 03 Nov 2026 10:20:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=50350148</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50350148</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50350148</guid></item><item><title><![CDATA[New comment by springtimesun in "Scaling Git at OpenAI"]]></title><description><![CDATA[
<p>Is the increase in commits graph in percentages because it’s fewer than Anthropic’s rate? A case of commit envy?<p>In all seriousness though I have enjoyed reading these posts. I will never work at this scale, but I like to read how different teams approach problems like this and I do wonder if some of these approaches will trickle down (or up?) into regular practice as everyone is committing more and faster.</p>
]]></description><pubDate>Tue, 03 Nov 2026 07:03:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=50348952</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50348952</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50348952</guid></item><item><title><![CDATA[New comment by springtimesun in "Voyager 1 gives controllers the silent treatment"]]></title><description><![CDATA[
<p>I don’t know how many are still here, but do people who launched the project ever come back and get to give them a command now and then?</p>
]]></description><pubDate>Fri, 30 Oct 2026 18:09:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=50307897</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50307897</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50307897</guid></item><item><title><![CDATA[New comment by springtimesun in "Kimi K3.1 Pro and K3.1 Pro Highspeed"]]></title><description><![CDATA[
<p><thinking></p>
]]></description><pubDate>Fri, 23 Oct 2026 19:13:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=50213957</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50213957</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50213957</guid></item><item><title><![CDATA[New comment by springtimesun in "EA is shutting down BioWare"]]></title><description><![CDATA[
<p>The voice recordings and non-linearity were what made it for me. I felt like I was piecing the world together as I explored. Even after doing the boss fight finding intel that explained something that wasn’t known before slotted it into my understanding of the story. By the time I reached the grind at the end, it was like I had a world in my head and it was ok to close the book.</p>
]]></description><pubDate>Thu, 22 Oct 2026 09:09:32 +0000</pubDate><link>https://news.ycombinator.com/item?id=50193042</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50193042</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50193042</guid></item><item><title><![CDATA[New comment by springtimesun in "EA is shutting down BioWare"]]></title><description><![CDATA[
<p>I loved The Division. Some months ago I re-downloaded it and just ran around for a bit. The setting and world feel are still there and the story is deeper than some single player games. For a while I thought, wouldn't it be amazing if it was ever released (or leaked) so it could be player hosted or modded in a local kind of development like this. But, I didn't stay with that thought. There was just something about that game that had a beginning, middle and end for me. Maybe I'm just getting old.</p>
]]></description><pubDate>Wed, 21 Oct 2026 21:24:16 +0000</pubDate><link>https://news.ycombinator.com/item?id=50187854</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50187854</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50187854</guid></item><item><title><![CDATA[New comment by springtimesun in "DeepSeek-v4.1-Flash-1017"]]></title><description><![CDATA[
<p>I feel like I’m an outlier based on all these comments. This is still working through my personal benchmarks, but it’s spending 30% more time and performing worse than the previous 4.1 flash did.<p>Benchmark is historical PR reviews. Models are put in a sandbox with a git and db snapshot as it existed at the time. PRs are a selection of changes some bugs (major and minor), some with things that aren’t bugs that have tripped up models in the past and some no findings that they should approve and move along.<p>4.1 flash already wasn’t that strong at this task, but the speed edged out v4 pro for a spot in the roster.</p>
]]></description><pubDate>Sun, 18 Oct 2026 19:24:18 +0000</pubDate><link>https://news.ycombinator.com/item?id=50144401</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50144401</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50144401</guid></item><item><title><![CDATA[New comment by springtimesun in "Redesigning source control at Anthropic"]]></title><description><![CDATA[
<p>I made a comment some time back that reviewing AI code was kind of bewildering (really any other code but my own). I’ve since dug in, crafted some processes and skills (in the /skill sense) and this is basically the shape it takes. Find the hot spots, what’s setting precedents, what’s not sitting right with me… it’s fast but it actually creates additional work (in a good way) because the way I have the skill set up the agent then scans through the whole codebase to find places where the changes demanded on a given PR also exist. Each PR spins out 1-multiple tickets. In a way, this has always been my dream. I believe in refactoring at some core level. Now I finally have the time (and tokens) to do it continuously.</p>
]]></description><pubDate>Thu, 15 Oct 2026 05:34:49 +0000</pubDate><link>https://news.ycombinator.com/item?id=50099604</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50099604</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50099604</guid></item><item><title><![CDATA[New comment by springtimesun in "AI is a tool and using it well takes skill"]]></title><description><![CDATA[
<p>LLMs have also pushed me to learn a lot, but I don’t find the shape to be terribly different than before. I rarely remember the exact sequence of commands or flags or syntax. What my learning was useful for before and after LLMs is to 1) build an intuition of how something behaves and what it should do and 2) expose me to non-obvious possibilities that I wouldn’t have come to on my own. With those two things in place, the next execution without the LLM is a docs search, same as it always was (for me maybe some people are better at remembering this stuff).<p>What’s different after LLMs for me is that there are some things that just don’t care about and I’m fine if they are a black box that the agent handles for me. I didn’t have that luxury before.</p>
]]></description><pubDate>Wed, 14 Oct 2026 07:37:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=50083821</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50083821</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50083821</guid></item><item><title><![CDATA[New comment by springtimesun in "Redesigning source control at Anthropic"]]></title><description><![CDATA[
<p>My take on picks and shovels was always a more cynical one. The beauty of the metaphor comes from the fact that after the miners went bust and left their picks and shovels where they lay, they could be collected and sold again to the next dreamer.</p>
]]></description><pubDate>Tue, 13 Oct 2026 20:31:33 +0000</pubDate><link>https://news.ycombinator.com/item?id=50078267</link><dc:creator>springtimesun</dc:creator><comments>https://news.ycombinator.com/item?id=50078267</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50078267</guid></item></channel></rss>