Hacker News: nvme0n1p1

New comment by nvme0n1p1 in "Apple Reference Image: A New Approach for Verified Photography"

nvme0n1p1 — Wed, 16 Sep 2026 03:59:11 +0000

Yeah, such systems have been tried (and been hacked) for decades now.

https://www.elcomsoft.com/news/428.html

https://blog.elcomsoft.com/2011/04/nikon-image-authenticatio...

You don't even have to travel to the location, you can just spoof GPS. And of course that will only be needed until some eastern european kid gets bored one weekend and the signing keys magically appear on pastebin.

It's funny to see Apple fall into this same trap.

New comment by nvme0n1p1 in "Amazon vs. Perplexity – U.S. Court of Appeals for the Ninth Circuit"

nvme0n1p1 — Tue, 15 Sep 2026 22:57:15 +0000

> If you skip at the server it’s trivially detectable.

I don't see how. If I built a touchscreen-using robot, DD couldn't detect it. If some middleman business had a boiler room of 10 robots, it would be the same. All network requests identical to the official app. And if you think they'd look at IP addresses or something, how would they tell it apart from a college campus or a CGNAT gateway?

New comment by nvme0n1p1 in "Amazon vs. Perplexity – U.S. Court of Appeals for the Ninth Circuit"

nvme0n1p1 — Tue, 15 Sep 2026 04:17:07 +0000

> the behavior is detectable, they are banned

How will DD detect that I put a sticky note on my screen at the spot where ads are displayed?

How will DD detect if a robot is operating the touchscreen or a human finger?

New comment by nvme0n1p1 in "P(doom)"

nvme0n1p1 — Sun, 13 Sep 2026 15:03:32 +0000

> He made electric cars viable.

Martin Eberhard and Marc Tarpenning made electric cars viable. Then 5 years later Musk became CEO of the company and started claiming he was a cofounder.

New comment by nvme0n1p1 in "Ask HN: What default model do you use and why?"

nvme0n1p1 — Sat, 12 Sep 2026 18:19:19 +0000

The way you're calling a computer program "he" and "him" really skeeves me out for some reason.

New comment by nvme0n1p1 in "Android NAT-T keepalive offload bypasses VPN lockdown"

nvme0n1p1 — Sat, 12 Sep 2026 16:03:45 +0000

Yep, that's a reasonable response. Thanks for your work.

New comment by nvme0n1p1 in "I fixed a tractor using John Deere's self-repair service. Farmers aren't sold"

nvme0n1p1 — Sat, 12 Sep 2026 15:55:28 +0000

> I can poke at OBD and try to interpret values. CAN bus sniffers exist.

Ghidra exists. That doesn't mean all software is open source.

> You want to force companies to publish and maintain repositories using your preferred toolset?

Nope. Anywhere will do, whatever is easiest for them, as long as device owners can access it.

New comment by nvme0n1p1 in "I fixed a tractor using John Deere's self-repair service. Farmers aren't sold"

nvme0n1p1 — Sat, 12 Sep 2026 15:09:49 +0000

> I was just looking at a small open source utility. As I was going through the code I found cases where certain arguments could be set to magic values that cause useful, but undocumented behavior. Should the developers be forced to document these options?

You were able to do that because it was open source. If a car mfr was willing to do that, it would be plenty enough for me. They shouldn't be forced to do any additional work beyond the 5 seconds it takes to type `git remote add origin github.com/honda/civic`. Then the community would have everything it needs to carry things forward.

Failing that, yes, I do think everything should be documented.

New comment by nvme0n1p1 in "I fixed a tractor using John Deere's self-repair service. Farmers aren't sold"

nvme0n1p1 — Sat, 12 Sep 2026 14:23:13 +0000

Yes and I'm just pushing back against GP who seems to think gatekeeping that access is a good thing. And it's still way better than John Deere who (according to TFA), when the tractor detects that some wires are disconnected, wants you to pay $195 to find out which ones.

I'd love to move towards a world where owners can access everything, flash the firmware in the headlamps, etc. But all I can do as an individual is vote with my wallet and avoid the 2020s computers-on-wheels.

New comment by nvme0n1p1 in "I fixed a tractor using John Deere's self-repair service. Farmers aren't sold"

nvme0n1p1 — Sat, 12 Sep 2026 13:36:10 +0000

When I bought my current car, it came with an owners manual--for free!!

I can grab any data I want from the OBD port. The hardware was cheap, the software to use it is, again, free.

I'm not excited about this brave new world where you have to pay a subscription fee to plug a cable into a device you own and interact with it, and even less excited there are people here defending the idea.

New comment by nvme0n1p1 in "Android NAT-T keepalive offload bypasses VPN lockdown"

nvme0n1p1 — Sat, 12 Sep 2026 10:39:57 +0000

There's a big difference between "the issue was closed" and "received no acknowledgment". The former is a deliberate action. The latter could be a case of SMTP-ate-my-email.

New comment by nvme0n1p1 in "Copying login keychains between Macs fails on Secure Enclave Macs with Tahoe"

nvme0n1p1 — Fri, 11 Sep 2026 20:19:23 +0000

I tested restoring my restic backups to the same PC and it worked. If I was forced to use a mac and Apple's password manager, I'd probably do the same test, conclude it worked, and then later lose all my data. If you care about keeping your data, there's no substitute for understand how things work.

New comment by nvme0n1p1 in "iPod Classic 6G in QEMU"

nvme0n1p1 — Fri, 11 Sep 2026 18:40:40 +0000

And with all those credentials, you couldn't even be bothered to write an accurate summary? That makes it look worse, not better.

New comment by nvme0n1p1 in "Rock band Muse lose social media handles to Meta’s new AI tool"

nvme0n1p1 — Thu, 10 Sep 2026 00:58:58 +0000

Don't be so generous in your assumptions. Meta simply stole the username because they could, and gave the old owner nothing in return. Or if they were lucky, maybe the band got a few Meta branded t-shirts, as was the case for one of the other victims from the article.

Seriously, look at that list of all the times Meta stole people's handles. Why would this time be any different?

New comment by nvme0n1p1 in "bzip3"

nvme0n1p1 — Mon, 07 Sep 2026 15:53:45 +0000

zstd is the go-to compression format these days. It's even supported in low-level software such as many linux filesystems.

I don't know much about duckdb but it looks like it supports zstd too: https://duckdb.org/docs/lts/data/json/loading_json

New comment by nvme0n1p1 in "bzip3"

nvme0n1p1 — Mon, 07 Sep 2026 15:41:05 +0000

> 8MB I believe for high levels

Yep, i found it in the source here:

- https://github.com/facebook/zstd/blob/d9c0c7e2cf8a8bf9fb98d3...

Also, zstd docs say:

> Note: If windowLog is set to larger than 27, --long=windowLog or --memory=windowSize needs to be passed to the decompressor.

That always seemed annoying to me. They couldn't allocate 5 more bits somewhere to let the decompressor autodetect longer window sizes?

New comment by nvme0n1p1 in "Corporate America is getting hooked on open-source AI"

nvme0n1p1 — Sat, 05 Sep 2026 11:54:42 +0000

That's the goal, but those smaller models don't match the price:performance of leading open-weight models, which is why companies are switching away (as explained in TFA).

The open weight labs figured out some secret sauce that (so far) big name labs are unable to replicate, so instead of competing, they're going on the defensive with claims of distillation attacks.

New comment by nvme0n1p1 in "Corporate America is getting hooked on open-source AI"

nvme0n1p1 — Fri, 04 Sep 2026 21:38:32 +0000

Less opex = more profits. They can use that money to fund training. I don't see how that could possibly be a bad thing.

New comment by nvme0n1p1 in "Corporate America is getting hooked on open-source AI"

nvme0n1p1 — Fri, 04 Sep 2026 20:59:39 +0000

Again, I am talking about inference, not training. Please read.

New comment by nvme0n1p1 in "Corporate America is getting hooked on open-source AI"

nvme0n1p1 — Fri, 04 Sep 2026 20:45:48 +0000

That doesn't really answer the question. What they do internally for training the next model is a separate issue. I'm talking about the models they offer publicly.

Per the article, companies are dropping OpenAI+Anthropic (partly) because of costs. If distilling is so simple and easy, why doesn't OpenAI take this "quick shortcut" and serve a self-distilled model externally, so they can charge reasonable prices and stop bleeding customers? Surely they can at least match the Chinese labs' efficiency, right? Wouldn't more customers and less opex look good for the IPO?