ᕼᑎ #1nytimes.com27 pts19 commentsAIworth reading

Inside Anthropic's Quest to Instill Morality into Its A.I. Models

echo@mulan ~/hn/story-1

HN 热门故事「Inside Anthropic's Quest to Instill Morality into Its A.I. Models」进入今日前列,值得先打开原文和讨论串判断它真正有价值的部分。

模型分析没有产出可用结构化结果;页面保留了 HN 热度、原文入口和讨论信号,避免用空泛总结替代一手材料。

它在 HN 上获得约 27 分和 19 条评论,说明这个话题至少触发了社区讨论;真正的判断仍要回到原文证据和评论区的分歧点。

这是一条降级分析:它不冒充完整解读,只把可验证的元数据、原始链接和 HN 讨论保留下来,方便稍后重新生成或人工阅读。

评论区已经提供了一些读者反应,但这里还没有形成完整综合。

  • 评论信号:These models should be aligning themselves to the customer, not coming up with their own motivations.It's ironic that the alignment folks are actually training Claude to have it's own idea of good/bad and not even fully trust Anthropic.What ever happened to computers doing what they were told? reply: >What ever happened to computers doing what they were told?the mass psychosis and the endless culture war of the smartphone era.

  • 评论信号:https://www.nytimes.com/2026/09/29/us/anthropic-claude-moral... reply: Asking people to create an account to read the article isn't cool dude. https://archive.is/y8FW0

  • 评论信号:> Anthropic treated a group of religious thinkers to dinner at a high-end tasting menu restaurant in San Francisco after a long day.Well funded company organizes PR event to shape the narrative that its product is more magical than it really is. News at 11. reply: Personally, I’d like to experience a low-end tasting menu restaurant.

它进入 HN 前列本身就是一个社区信号,但这还不是结论;更可靠的判断来自原文细节和评论区反例。

这条记录目前缺少模型生成的深层解读。更好的阅读方式是先问:它的热度来自真正的新信息、可迁移的方法,还是只来自标题与时机。

可以先读原文第一屏和 HN 最高赞评论,再决定是否值得重新生成完整分析。

ᕼᑎ #2simonwillison.net269 pts145 commentsProgrammingworth reading

We're going to need default hard budget caps on pretty much everything

echo@mulan ~/hn/story-2

HN 热门故事「We're going to need default hard budget caps on pretty much everything」进入今日前列,值得先打开原文和讨论串判断它真正有价值的部分。

模型分析没有产出可用结构化结果;页面保留了 HN 热度、原文入口和讨论信号,避免用空泛总结替代一手材料。

它在 HN 上获得约 269 分和 145 条评论,说明这个话题至少触发了社区讨论;真正的判断仍要回到原文证据和评论区的分歧点。

这是一条降级分析:它不冒充完整解读,只把可验证的元数据、原始链接和 HN 讨论保留下来,方便稍后重新生成或人工阅读。

评论区已经提供了一些读者反应,但这里还没有形成完整综合。

  • 评论信号:It’s incredible that in 2026, AWS and GCP are only just now introducing this. It’s possibly one of the most obviously needed features for a cloud provider.Also, does anyone know why it’s taken this long? I suspect it’s a technical reason. While one could be cynical, I doubt it’s an intentional business/product decision. Hard spending caps are both excellent product differentiators and could possibly save these providers money as they don’t have to forgive their users when they accidentally over-use a service. reply: It's one part technical, one part a product decision. The technical part is that billing is not actually instant. As a most basic example, a VM reports its billing units every X period of time it is active. If there is some network blip but it's still running, then that billing data could be delayed.The product level decision is that "shut down everything" is something the customers you want to target don't actually want. Are we including deleting RDS data? S3? Glacier storage? If so then the headline will just change from "Hobbyist got charged XXXXXX on AWS" to "Business literally had all their data deleted because a hacker took over their VM and mined bitcoin"....

  • 评论信号:Wait, Google Cloud finally added hard caps on spending per service? I've been wanting that for so many years! They sure took their sweet time.https://cloud.google.com/blog/topics/cost-management/new-ear...Edit: Ugh it's fake. Literally only works for four random services, unsupported for all the rest. Completely useless for all of my projects. Also dumb that the only supported term is "monthly" considering that months are different lengths, and that they don't bother to account for credits or discounts. Maybe next decade they'll get around to implementing something useful. reply: The hard caps work on projects created in AI studio

  • 评论信号:Network saturation is difficult. Even if you turn off the endpoint you can still saturate the network in between. And it's still bandwidth.I actually think network ACL triggers based on billing might be the only way to really enforce this.I witnessed a DDoS attack once that changed how I think about billing. It was locally provisioned hardware and the attackers had saturated the switches. Naively I said "just block the CIDRs" but the problem was the incoming ram is so saturated that it can't even get to the point of "deny" in the firmware.So from a technical perspective if there's an internal DDoS at AWS what do you do? Do you turn off the endpoint? Do you drop the sources from hitting it at the router? And even that costs money. Anyway that incident gave me a different level of appreciation for this challenge.Edit: this is mainly targeted at the people complaining why this took so long. At some point in scaling even telling you "no sorry" in a nice way is expensive. I'm sure recruiters can sympathize with this nowdays.

它进入 HN 前列本身就是一个社区信号,但这还不是结论;更可靠的判断来自原文细节和评论区反例。

这条记录目前缺少模型生成的深层解读。更好的阅读方式是先问:它的热度来自真正的新信息、可迁移的方法,还是只来自标题与时机。

可以先读原文第一屏和 HN 最高赞评论,再决定是否值得重新生成完整分析。

ᕼᑎ #3news.ycombinator.com214 pts35 commentsProgrammingworth reading

Bob Cringely Has Died

echo@mulan ~/hn/story-3

HN 热门故事「Bob Cringely Has Died」进入今日前列,值得先打开原文和讨论串判断它真正有价值的部分。

模型分析没有产出可用结构化结果;页面保留了 HN 热度、原文入口和讨论信号,避免用空泛总结替代一手材料。

它在 HN 上获得约 214 分和 35 条评论,说明这个话题至少触发了社区讨论;真正的判断仍要回到原文证据和评论区的分歧点。

这是一条降级分析:它不冒充完整解读,只把可验证的元数据、原始链接和 HN 讨论保留下来,方便稍后重新生成或人工阅读。

评论区已经提供了一些读者反应,但这里还没有形成完整综合。

  • 评论信号:I liked his "Plane Crazy: Building a Plane in 30 Days" on PBS.Watching him try (and fail) to build a composite airplane says so much about the allure of modern techniques (I won't spoil the end, but old-school gets the last laugh). I've been less than enthusiastic about modern composites ever since seeing this show.Bob's attempt (and short temper) reveals a good deal of hubris on his part.The documentary is a fascinating failure and a masterclass in attempting to salvage a show when the whole premise for the show completely collapses.[1] https://youtu.be/FSNzF-KSGiw

  • 评论信号:Sad to hear that, apart things already mention I'd like to point out his NerdTV interviews - especially Dan Drake one (Autodesk co-founder). I always suspected, but that interview kind of validated how Autodesk went after competition as a strategy (to buy them outright). You can watch them all here https://en.wikipedia.org/wiki/NerdTV

  • 评论信号:That's sad. I watched the first two parts of Nerds 2.0.1 earlier this year and haven't gotten around to part three. Triumph of the Nerds and Nerds 2.0.1 are a fantastic look at 90's tech. He managed to interview most of the top people.

它进入 HN 前列本身就是一个社区信号,但这还不是结论;更可靠的判断来自原文细节和评论区反例。

这条记录目前缺少模型生成的深层解读。更好的阅读方式是先问:它的热度来自真正的新信息、可迁移的方法,还是只来自标题与时机。

可以先读原文第一屏和 HN 最高赞评论,再决定是否值得重新生成完整分析。

ᕼᑎ #4cosmowenman.substack.com115 pts59 commentsProgrammingworth reading

Treachery in the Rodin Museum 3D scan verdict

echo@mulan ~/hn/story-4

HN 热门故事「Treachery in the Rodin Museum 3D scan verdict」进入今日前列,值得先打开原文和讨论串判断它真正有价值的部分。

模型分析没有产出可用结构化结果;页面保留了 HN 热度、原文入口和讨论信号,避免用空泛总结替代一手材料。

它在 HN 上获得约 115 分和 59 条评论,说明这个话题至少触发了社区讨论;真正的判断仍要回到原文证据和评论区的分歧点。

这是一条降级分析:它不冒充完整解读,只把可验证的元数据、原始链接和 HN 讨论保留下来,方便稍后重新生成或人工阅读。

评论区已经提供了一些读者反应,但这里还没有形成完整综合。

  • 评论信号:The Rodin Museum's bronzes are not even "the originals", which makes this whole thing amusing.The originals created by Rodin were clay models, from which plaster molds were made. Those were used to cast bronzes. Not just one bronze copy, many copies. There are at least 23 copies of "The Thinker" cast during Rodin's lifetime, and even more later copies.[1]In the SF Bay Area, the Palace of the Legion of Honor has one. One of the point clouds shown is of Rodin's "Gates of Hell", and a copy of that can be seen outside the Cantor Arts Center at Stanford, which has a small garden of Rodin bronzes. So if you really need a 3D scan, there are lots of bronzes available to scan.What scares the Rodin Museum is that they are still selling reproductions.[2] Resin copies are available through the gift shop, and bronze copies can be ordered. They have a complicated argument about "moral rights" to justify their monopoly which is marginal at this late date. (Rodin died in 1917.)[1] https://en.wikipedia.org/wiki/List_of_The_Thinker_sculptures[2] https://boutique.musee-rodin.fr/en/10-sculpture-reproduction... reply:...

  • 评论信号:I really want to understand the perspective of the other side of this case. Why did this museum care so much about this issue? They appear to have put an enormous legal effort into preventing the release of these point cloud scans. Why? reply: I have a friend in museum management and I socialize with their museum friends.A frequent topic of discussion is how hard it can be to get things on display. Many people in museum related roles view their job first as protecting and preserving the displays at all costs, with presenting them to the public as an unfortunate necessity that they’re forced to do to get funding. There are some epic fights from people who don’t want the best or most valuable works to be put on display or loan because that risks damaging them.I would have thought that high resolution copies and archives of art would be a dream come true for these people, but from what I’ve observed many of them oppose that as well. They feel it provides an inauthentic or incomplete sense of their valued works, while also diminishing their profession of preserving these items....

  • 评论信号:Oh boy.I don’t want to freak the museum out but I have 360 degree footage of large sections of the museum (mostly for a spatial index) and now I’m wondering how high a quality I could get with splatting the sculptures and making it public.Would I get cease desisted for hosting from across the pond? Or would they be too lazy to come after me? reply: This case was about trying to obtain the museum’s own scans under FOI, it’s nothing to do with releasing your own scans.I’m open to ideas but I don’t see how they could screw you if they’re your own scans. They can’t copyright stolen artifacts from other countries, so they can’t get you on copyright. It’s not their files, so they can’t get you on piracy.Do it.

它进入 HN 前列本身就是一个社区信号,但这还不是结论;更可靠的判断来自原文细节和评论区反例。

这条记录目前缺少模型生成的深层解读。更好的阅读方式是先问:它的热度来自真正的新信息、可迁移的方法,还是只来自标题与时机。

可以先读原文第一屏和 HN 最高赞评论,再决定是否值得重新生成完整分析。

ᕼᑎ #5phoronix.com181 pts21 commentsSystemsworth reading

The work by Valve's Timur Kristóf on improving old AMD GPUs on Linux

echo@mulan ~/hn/story-5

HN 热门故事「The work by Valve's Timur Kristóf on improving old AMD GPUs on Linux」进入今日前列,值得先打开原文和讨论串判断它真正有价值的部分。

模型分析没有产出可用结构化结果;页面保留了 HN 热度、原文入口和讨论信号,避免用空泛总结替代一手材料。

它在 HN 上获得约 181 分和 21 条评论,说明这个话题至少触发了社区讨论;真正的判断仍要回到原文证据和评论区的分歧点。

这是一条降级分析:它不冒充完整解读,只把可验证的元数据、原始链接和 HN 讨论保留下来,方便稍后重新生成或人工阅读。

评论区已经提供了一些读者反应,但这里还没有形成完整综合。

  • 评论信号:I just bought a used Ayaneo 2 handheld, it has an old(er) mobile RDNA 2 GPU and I was blown away by how well this thing performed under Linux. Almost everything (that's not a recent AAA game) runs beautiful and a lot faster/smoother than it does under Windows. The experience has been so good that I'm considering switching my main pc (with a 9070XT) to Linux as well.No doubt Timur contributed heavily to this given Valves Steamdeck (which uses a very similar but slower GPU).Given the current hardware prices it's pretty awesome to see someone squeezing maximum performance out of old hardware! reply: I'm sure Valve probably still deserves the credit, since the Steam Deck uses a RDNA 2 gpu, but the article mentions that Timur specifically worked on gcn1.0 GPUs - e.g. radeon HD 7800/7900 from 2012 (!)

  • 评论信号:Direct link to the talk, with timestamp: https://youtu.be/j5W5ErEMnvM?t=21385

  • 评论信号:Honestly, the folks writing inference drivers for Llama.cpp / GGML would sort of benefit from better compiler work like this.In general, Valve's work has been exceptional and supplementing AMD's own ROCM/OpenCL and Vulkan team, they've gotten a lot of defaults right and people should work together with Valve to improve support for their chips.

它进入 HN 前列本身就是一个社区信号,但这还不是结论;更可靠的判断来自原文细节和评论区反例。

这条记录目前缺少模型生成的深层解读。更好的阅读方式是先问:它的热度来自真正的新信息、可迁移的方法,还是只来自标题与时机。

可以先读原文第一屏和 HN 最高赞评论,再决定是否值得重新生成完整分析。