This hurts, because, it's true: "Silicon Valley will spend millions developing a glove that translates sign language into speech (badly), but they won't spend ten minutes adding Alt-Text to their images. They will build a headset that describes the room to me (poorly), but they won't use semantic HTML headings so I can actually navigate a webpage."
Stop trying to upgrade disabled people's body when the problem is your design!
@WeirdWriter argues that the tech industry needs to stop trying to sell useless expensive gadgets no disabled person wants, or can afford, and start fixing actual accessibility barriers, like bad code, and poorly designed infrastructures.
https://sightlessscribbles.com/posts/disability-dongle/
So my employer finally decided to roll out controls on browser extensions. I'm kinda surprised they didn't do that sooner, given the shitshows I've seen elsewhere.
Thankfully, unlike when they banned firefox, they pulled the stick out of their arse and actually allowed UBO as "the one true adblock." (Though they do appear to be blocking Consent-o-matic and Don't Fuck with Paste, which is annoying, but at least it's not "actively sabotaging their own security efforts" like banning UBO would be...)
And for some reason they let the rank-and-file see the list of every extension that's installed on a company machine (as an anonymous deduplicated list of "what we're not going to allowlist") so I'm over here with my jaw somewhere down in hell's boiler room with just how much absolute shit people have installed on their work machines...
From release notes of LibreOffice 26.8:
”[It] contains no generative AI features. Documents are not transmitted to remote services for processing (…). This is a deliberate position.”
https://blog.documentfoundation.org/blog/2026/08/26/libreoffice-26-8/
"Why should we prioritize making this accessible?"
"The short answer is: because people have civil rights. The long answer is: because people have this thing called 'civil rights', you cunt."
Weh. My desire for an ARM64 laptop for non-work purposes has gotten strong enough I've actually started looking. Anyone have recommendations for ones that aren't MacBooks and still have decent mainline Linux support? (Macs are the fallback option, but I already know all about those as I've been following the Asahi project since they first got m1n1 running, much less the full kernel)
Dear developers,
Please respect the reduced-motion preferences of your users.
I'm tired of your websites making me nauseous.
Motion is fun until it makes your users sick.
Here you go: https://developer.mozilla.org/en-US/docs/Web/CSS/Reference/At-rules/@media/prefers-reduced-motion
https://developer.mozilla.org/en-US/docs/Web/CSS/Reference/At-rules/@media/prefers-reduced-motion
RE: https://tech.lgbt/@solonovamax/116049115040950367
If I'm reading the disclosure correctly, the issue is:
- Windows Notepad is more than just a plain text editor now.
- In particular, it has a markdown preview feature, including clickable links.
- But, it doesn't have full web browser security processes for what to do if you click on a link with a protocol that triggers a local application. It gets treated as if the user was directly running that application.
So, don't open strange files & then click links. (And update Windows regularly.)
JFC. There's exactly one thing this 'course' lists as a use for LLMs that I can actually see them being a reasonable use not loaded totally full of footguns: "synthetic sample data", and even then there are much more efficient ways to generate that... (You really don't need the monster tens-of-GB models to get "sample-shaped data" out of them...)
(Though I would still never actually use one for that, given how easy it is for them to end up giving you real data in the pile... They're just predictive autocomplete, after all, so if you accidentally give it an input that lines up close enough to data it was trained on enough times, it might just give you verbatim output of that training data. See the recent "extracting famous books from LLMs" research that blows a long-lance sized hole in the slopster's "it's not memorizing, it's learning patterns" argument... Not that that held much water to begin with, given that sufficiently overfitted patterns basically are memorization...)
Dear G-d... [Redacted] is making me take an "AI" course for sunk-cost FOMO reasons... Half the "optimizations" they're talking about aren't optimizations at all. "Boilerplate" is either due to an insufficient level of abstraction, or can be written much more efficiently (and accurately...) via traditional code generation techniques. As for "expressing the problem as natural language", I don't think about the problem-space in natural language, so it's just as much (if not more) work to serialize it out as natural language as it is to just write the damned code...
And "implementing features": if you can accurately express what needs to be done such that a computer can do the rest, what's that called? Code. Congrats, you just made a shitty compiler that's both nondeterministic and with frequently wildly incorrect output...
When we design for disabilities, we make things better for everyone. This is called the Curb-Cut Effect. The term was coined by disability students and activists in the 70s, who added curb cuts to the Berkeley sidewalks to make access easier for those in wheelchairs. They discovered those also helped people with strollers, using trolleys for deliveries, etc.
I’m surprised I haven’t heard more responses from web developers about the change to Safari on iOS and iPadOS regarding web apps. Now every website can be a web app (saved to the Home Screen and opened as a stand-alone app) — not just sites that have been configured a certain way by the developers.
It’s a big difference for users. Every site gets the same experience. No more mysterious sometimes-it-works-one-way, sometimes-another.
Lovely. I need to buy new pants, and the pants-buying website I usually go to (because they'll happily let me buy the exact same pants I'm currently wearing, just without a hole in them) is just straight up broken with a javascript error somewhere in the giant mess of tracker blobs they've haphazardly stacked on top of each other. (I tried with tracker blocking on, and the site broke even worse)
I just wanted to buy pants, but apparently the place selling pants doesn't feel like actually selling any pants for some strange reason...
(This advice doesn't come from any particular incident, I just had a random thought about backups)
Was just (randomly) inspired for more Random Admin Advice: Keep your backup storage's management plane completely separate from any directory services (AD, etc) you may be using. You can connect it to authorize clients for both backup and restore purposes, but all management of the system and retention periods should be done with a distinct set of accounts managed locally on-device, and ideally (although not always possible) require physical hands-on-keyboard for certain operations.
The true key to good backups is don't let AD dictate who can delete backups. In the event of infrastructure compromise, step one of ransomware after getting an idea of what's going on is "destroy backups." If there's no trivial way for them to pivot from domain admin to the backup server, that task becomes a hell of a lot harder.
(Also, this is your periodic reminder to A: check your backups, and B: make sure they're on a different disk/host than the live copy, though with some exceptions for oopsie-rollback snapshots)
I'm growing tired of people blaming accessibility guidelines, for stripping away the creativity from the web.
Most websites are not accessible. Many designers don't know (yet I hope), or don't care about accessibility. Or care in a performative way.
So, stop blaming accessibility for this. Thanks.
Study: When experienced open-source developers used current AI tools, work took 19% longer to complete. In spite of which, the developers *thought* they were working *faster.*
It’s just one study, and they don’t claim anything beyond what they observed. Their results are not predictive. Still....
(There must be a name for the psychological trick we play on ourselves to believe we’re being more productive even when we’re considerably less so.)
Hat tip: Dave Martin
https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/
Lazyweb question: After the latest windows update, my work laptop absolutely refuses to wake from suspend. (power light "breathing") Keyboard does nothing, and the power button does a 'soft reboot' that appears to be a hibernation resume instead of a 'wake from suspend'. And because the system was never actually hibernated (it's not configured to when plugged in, only when on battery) it boots to a bitlocker recovery prompt, which can only be 'bypassed' by a hard-reboot to reset the TPM state to a clean-boot ("The secureboot policy has changed" is the bitlocker error)
Has anyone else encountered this before?
Stupid idea: "Emulating TPMs in/with an HSM." Specifically for VMs, because snapshots and other stateful stuff breaks the 'normal' TPM behavior: one-way counters may need to roll back if a snapshot is restored, but anti-hammering shouldn't allow that. So an "emulator" of sorts is still needed.
Normally TPM emulators are using swtpm, but that's a pure-software solution: if the host/dom0 is compromised, there goes all your encrypty-bits, even at rest.
Another idea would be to software-emulate the TPM "as normal", but instead of keeping the key material unprotected in RAM, protecting it with the real TPM (under a per-VM key, and in a completely decoupled manner from the emulated one, aside from potentially the key sealing hierarchy) So you'd lose out on things like end-to-end parameter encryption and true hardware attestation, but you'd still end up with hardware-protected secrets.
She's my new hero.