ᕼᑎ #1cactuscompute.com614 pts135 commentsProgrammingworth reading

Whistle: Speech to Text in 16.9 MB

echo@mulan ~/hn/story-1

HN 热门故事「Whistle: Speech to Text in 16.9 MB」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 614 分和 135 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:Interesting that this is here. I used whistle (and bunch of other things) to take ownership of my echo show. It now doesn't dial to Amazon at all - it does all processing locally with its own CPU and connects to my homeassistant for home automation. My initial setup involved qwen asr (1.7b model) running on rtx 5080. Compared to that, whistle was really bad (out of 170 messages, qwen recognized correctly 168, whistle - 70), but I adjusted whistle to work like jev - instead of free form transcription it recognizes only select set of templates (I trained tiny network with 10,000 generated utterances to translate whistle final state to probabilities within templates). The precision went up to 164/170 - almost matching qwen. By the way - I'm speaking with heavy accent. reply: This feels like we're going back to the past again with CMU's Sphinx4 in Java. It worked way better than I would ever have expected it to for being more than a decade old. It relied on the user defining a grammar of valid words and different flows through a standard format (Java Speech API Grammar Format). I wonder if we'll approach that again for these models just like how MCPs act like WSDLs in spirit....

  • 评论信号:I don't think the challenge with speech to text was size of the binary. In my experience the challenge is understanding my 84 year old Croatian father with a sagging mouth after a stroke, when he's trying to write his autobiography.I just setup Windows speech to text for him last week and it's great to see how he can write an entire page in 10 minutes, it would take him days using the keyboard.But every single sound he makes with his mouth ends up on the page too. reply: Sorry about your father. He needs a dictation model, not a general purpose speech-to-text model. They ignore umms and ahhs, change things like “an elephant, no a monkey, went up the tree” to “a monkey went up the tree,” support saying punctuation aloud sometimes, etc.Gemini team just released Gemini 3.5 Transcribe that’s supposed to be good at this; it’s available via api: https://blog.google/innovation-and-ai/models-and-research/ge...

  • 评论信号:What the demo does not do is show streaming output of transcribed text as we are speaking and recording (before we hit stop). That is an essential feature IMO for most general purpose live STT apps. reply: What do streaming implementations do when a bigger context reveals a different interpretation/parse? When I use Whisper in the terminal, I can see it going back and correcting itself. Are corrections off the table for a true streaming transcription?

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 #2lwn.net41 pts18 commentsProgrammingworth reading

Reducing undefined behavior in the C language

echo@mulan ~/hn/story-2

HN 热门故事「Reducing undefined behavior in the C language」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 41 分和 18 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:"Some Honeywell machines, for example, had nine-bit bytes."OS 2200 has 36-bit words. It is still a supported platform.https://en.wikipedia.org/wiki/UNIVAC_1100/2200_seriesThis platform was the first SMP UNIX implementation:"Any configuration supplied by Sperry, including multiprocessor ones, can run the UNIX system."https://www.nokia.com/bell-labs/about/dennis-m-ritchie/other... reply: > OS 2200 has 36-bit words. It is still a supported platform.Yes, but that's perhaps an argument for 'implementation defined behaviour', not in favour of 'undefined behaviour'.

  • 评论信号:1790620504 | Reducing undefined behavior in the C language | https://lwn.net/SubscriberLink/1095811/b9325731ea9b61e0/ | https://news.ycombinator.com/item?id=49882419 | 15 comments1790673092 | Reducing undefined behavior in the C language | https://lwn.net/SubscriberLink/1095811/efcdbcf080cfa4c6/ | https://news.ycombinator.com/item?id=49890290 | 0 comments

  • 评论信号:i've seen static analyzers catch many ub patterns, but guaranteeing zero ub needs whole‑program analysis that blows up compile time and still produces false positives that drown developers reply: Yes, it's pretty much impossible to prove anything about arbitrary programs.However if you are willing to restrict what programs you allow, you can make guarantees possible.Silly example: if you compile valid (safe) Rust programs to C, you know that the resulting code will not invalidate Rust's borrowing rules by construction; and in principle you could try to establish this guarantee just from the C code alone, never having seen the Rust original.However, you still wouldn't be able to have an algorithm that tells you for any arbitrary C code whether it has these problems or not.

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 #3unsung.aresluna.org24 pts14 commentsProgrammingworth reading

Keyboard differences between Windows and Macs

echo@mulan ~/hn/story-3

HN 热门故事「Keyboard differences between Windows and Macs」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 24 分和 14 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 wish Apple would get back to making decent electromechanical keyboards.

  • 评论信号:Asahi and GNOME teams, please take this as an guide to implement ”Mac input compatibility mode”!Dual booting between a MacOS and Linux is a painful process of manual keyboard and shorcut overrides to get the linux side somewhat to the (better) mac way of doing this.

  • 评论信号:Mac got the Command key right. Windows got the Home and End keys right. Wish we could have both on one system.

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 #4theranos.world339 pts126 commentsProgrammingworth reading

Theranos.world

echo@mulan ~/hn/story-4

HN 热门故事「Theranos.world」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 339 分和 126 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:Fascinating that the only communications with Tyler Schultz, which was the main insider confronting the fraud, had all of ~4 exchanges with her. All in the same day.Truly brave of him to go against the golden geese, despite turning his own family- and even former attorney generals & military officers- against him.Its all on 4/11/2014.Also interesting to see how Sunny & Holmes always thought it was Tyler, or the CLIA Lab Director (Adam) plotting against them.... when their scam was just in plain sight of regulators that bothered to actually do the work.Holmes/Sunny reaction to Tyler's email and to Adam's resignation (Nov 2014) is one of outrage. Using the power inbalance to make their point, is 101 fraud.

  • 评论信号:Content marketing for a startup that does "Document infrastructure for agents". They want you to sign up for an event they're hosting in SF. reply: A24 (the studio behind the movie) is one of our customers, so we're hosting a watch party in SFplus, we used our APIs to parse all the theranos PDFs to power the experiencehope our content marketing made your day a bit more fun!

  • 评论信号:This is great, it reminds me of promotional websites from the Adobe Flash era. Ads were a lot more fun back then. reply: They replaced the old space jam website with a boring href to a yt of the trailer :(https://www.spacejam.com/

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 #5dexerto.com499 pts319 commentsProgrammingworth reading

Man discovers his parents' coffee machine used 1TB of data in 10 days

echo@mulan ~/hn/story-5

HN 热门故事「Man discovers his parents' coffee machine used 1TB of data in 10 days」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 499 分和 319 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:In the Twitter thread linked, the person confirms two things:1. It saturated the local network with 1TB of metadata sniffing scans, not the network uplink to the outside world.2. It does so because, as Keurig notes, it’s collecting data about your household in order to let Keurig sell it to advertisers. reply: > as Keurig notes, it’s collecting data about your household in order to let Keurig sell it to advertisers.WTF!! When did we land in the middle of a Black Mirror episode?I'm half convinced the first run of the LHC split the timeline and we've landed in the evil one.

  • 评论信号:I'm sceptical only because this is a screen shot of unifi client info, and unifi seems to have some bug that will report devices as using terabytes of data in the last 24 hours, which is obviously not true..Here's an example from my unifi dash right now, my laptop has done 24tb of data? lol email and web browsinghttps://imgur.com/a/I54vcYR reply: I posted the same below, they’ve had similar bugs for the ~decade I’ve used them.

  • 评论信号:I'm not the most versed with networking or operating systems but I wonder how difficult it would be to program a raspberry pi to poison the datasets of these types of things. Program it to impersonate a wide set of devices and randomly respond as such in a home. Or even if it would be worth it. Maybe if a phone app could do it too that would make it easy to poison the dataset only when you purchase new smart devices while they do their heavier reconnaissance during the first few weeks (assuming Keurig and the like calm down a bit after some period of time). reply: If I had a machine like this on my network, I'd be tempted to figure out where it's sending the telemetry data, and help it out by sending a LOT more. It would be funny if they blocked my IP because I was reporting too much data.

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。