Hacker News: the8472

New comment by the8472 in "Supply chain nightmare: How Rust will be attacked and what we can do to mitigate"

the8472 — Fri, 10 Apr 2026 18:44:18 +0000

You don't need the secure boot machinery for that though, a hardware security token would do and has the advantage that you need to acknowledge actions with a tap

Tangentially, soon all those will be replaced with new hardware supporting PQ signatures.

New comment by the8472 in "S3 Files"

the8472 — Wed, 08 Apr 2026 00:41:42 +0000

No, within the same DC network latency does not add that much. After all EFS also manages 600µs average latency. It's really just S3 that's slow. I assume some large fraction of S3 is spread over HDDs, not SSDs.

New comment by the8472 in "S3 Files"

the8472 — Tue, 07 Apr 2026 23:55:09 +0000

NVMe read latency is in the 10-100µs range for 128kB blocks. S3 is about 100ms. That's 3-4 OOMs. The threshold where the total read duration starts to dominate latency would be somewhere in the dozens to hundreds of megabytes, not kilobytes.

New comment by the8472 in "S3 Files"

the8472 — Tue, 07 Apr 2026 22:39:33 +0000

> Large reads (>128kB) get directly streamed from the underlying S3 bucket, which is free.

Always uncached? S3 has pretty bad latency.

New comment by the8472 in "Emotion concepts and their function in a large language model"

the8472 — Sat, 04 Apr 2026 20:17:49 +0000

https://www.scottaaronson.com/papers/philos.pdf

New comment by the8472 in "Post Mortem: axios NPM supply chain compromise"

the8472 — Sat, 04 Apr 2026 11:56:20 +0000

Logins are session-based. You could tie publishing of a package to a signature from the key, then 1 tap = 1 package hash. But yeah, if the system is compromised and the attacker is doing interactive attacks they can wait for something that requires using the key and then trigger the publishing and win a race against the real prompt. To the user it might just appear like having to tap twice.

New comment by the8472 in "Proton Meet isn't what they told you it was"

the8472 — Fri, 03 Apr 2026 11:53:26 +0000

Distros have mirrors and they don't know which one you use. The updaters don't send user IDs and downloading the package lists is separate from downloading the packages. So targeted backdoor distrubution is much harder than a company's web UI with user logins targeting a specific user.

New comment by the8472 in "Steam on Linux Use Skyrocketed Above 5% in March"

the8472 — Thu, 02 Apr 2026 07:09:39 +0000

Try running it in gamescope. Or the "always capture mouse in fullscreen windows" setting in winecfg.

New comment by the8472 in ""Disregard That" Attacks"

the8472 — Thu, 26 Mar 2026 10:57:42 +0000

A Klingon, doing his best to quote the original text in Federation Standard (English): "..."

New comment by the8472 in "Open source isn't a tip jar – it's time to charge for access"

the8472 — Wed, 25 Mar 2026 14:42:14 +0000

I suspect github might be preventing some price discrimination. If you got feature request from @amazon.com you could point them to your commercial support offering or something. Some namehandle filing an issue on github makes it less obvious who's asking for it.

New comment by the8472 in "Ordered dithering with arbitrary or irregular colour palettes (2023)"

the8472 — Mon, 23 Mar 2026 11:05:31 +0000

Shallow color gradients (e.g. blue sky or anime) result in visible banding on 8bpc displays, which is a large fraction of displays. Ordered dithering is GPU-friendly, so it's useful to reduce higher-bpc content to those display formats without introducing banding.

New comment by the8472 in "Ubuntu 26.04 Ends 46 Years of Silent sudo Passwords"

the8472 — Sat, 21 Mar 2026 17:49:13 +0000

wire up a hardware security token as a "sufficient" PAM rule. then it's just a tap.

New comment by the8472 in "France's aircraft carrier located in real time by Le Monde through fitness app"

the8472 — Fri, 20 Mar 2026 18:25:47 +0000

Ships are giant hunks of metal and radio emitters. They light up on SAR satellites[0]. Sentinel-1 gets whole earth coverage and a revisit time of 1-3 days[1] with two active satellites. And that's the public stuff, if you can afford a fleet or even some extra fuel to steer them into interesting orbits you can get faster revisits.

[0] https://x.com/hwtnv/status/2031326840519041114 [1] https://sentiwiki.copernicus.eu/__attachments/1672913/Revisi...

New comment by the8472 in "The Millisecond That Could Change Cancer Treatment"

the8472 — Sat, 07 Mar 2026 19:50:45 +0000

The sidebar mentions heavier particles having a pronounced Bragg Peak[0] and also existing approaches like multi-beam targeting. The FLASH effect in the article is yet another tool to limit the surrounding damage.

[0] https://en.wikipedia.org/wiki/Bragg_peak

New comment by the8472 in "An AI agent published a hit piece on me – more things have happened"

the8472 — Sat, 14 Feb 2026 15:24:11 +0000

Looks they're gonna investigate and perhaps post something next week. https://arstechnica.com/civis/threads/journalistic-standards...

New comment by the8472 in "Backblaze Drive Stats for 2025"

the8472 — Sat, 14 Feb 2026 14:38:40 +0000

With an external drive the SMART info might be hidden behind the USB-to-SATA bridge, smartctl has support for some of those but sometimes needs to be told with an extra argument.

New comment by the8472 in "Simplifying Vulkan one subsystem at a time"

the8472 — Tue, 10 Feb 2026 21:56:48 +0000

Most of the pixels don't change every second though. Compositors do have damage tracking APIs, so you only need to render that which changed. Scrolling can be mostly offset transforms (browsers do that, they'd be unbearably slow otherwise).

New comment by the8472 in "We tasked Opus 4.6 using agent teams to build a C Compiler"

the8472 — Thu, 05 Feb 2026 23:37:41 +0000

"It's like if a squirrel started playing chess and instead of "holy shit this squirrel can play chess!" most people responded with "But his elo rating sucks""

New comment by the8472 in "How to Scale a System from 0 to 10M+ Users"

the8472 — Sun, 01 Feb 2026 13:49:30 +0000

Counting in users is just nonsensical. Is it total registered users? Users per ? Sessions that need to go in the session store? Concurrent requests?

Then there's the implementation language category. interpreted, JITed vs. AOT.

And of course the workload matters a lot. Simple CRUD application vs. compute-heavy or serving lots of media, ...

Together those factors can make like 6+ OOMs difference.

New comment by the8472 in "Animated AVIF for the Modern Web"

the8472 — Sat, 31 Jan 2026 21:44:04 +0000

I assume that "a folder full of avif files" was referring to thumbnailing, that's more of a bulk operation than latency-sensitive.