Mastodon Feed
pzmyers@freethought.online ("pzmyers 🕷") wrote:
Since when do unqualified business turds like Jerry Cirino get to determine what knowledge may be taught?
Mastodon FeedMastodon Feedbrib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:") wrote:
one of the things I noticed when coding yestersay is the thought process to "I want this feature" to the actual code is very nonlinear and often includes just writing the actual code.
I was trying to keep a devlog to remember my progress and this often meant going back after I had coded something to do the writeup. which future self will very much thank me for, but it was very much a process of natural language reconstruction of code (and nonlinear thought processes) rather than coming up with the natural language first and the code later.
that is to say: no wonder I prefer actual coding to whatever you do with LLMs, because I very rarely find "express the thing you want in English but in a way an LLM can understand" to be any sort of productivity improvement. I already find natural language an incredibly unnatural way to work with computers. I'd rather just express what I want in code and get the advantage of determinism.
Mastodon FeedBoosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
infinitybear@mastodon.infinitybeardesigns.com ("Manda | Infinity Bear Designs") wrote:
This little chick always finds the most cozy spot.
Both of these cute cross stitch patterns are freebies!
https://www.infinitybeardesigns.com/freebies
#CrossStitch #Embroidery #Art #FediArt #MastoArt #FreePattern #Cats
Mastodon FeedChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕") wrote:
We talked about the film for longer than the actual film! 5/10
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
That’s the real “twatted cache” scenario.
😂 😂 😂 😂 😂
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
So yes: if the I/O path is wrong or races with reuse, your cache gets “twatted” no matter how elegant the SIMD lookup is.
😂
Mastodon FeedBoosted by ChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕"):
FediVideo@social.growyourown.services ("Fedi.Video") wrote:
Chris Were is one of the longest-running vloggers on the Fediverse, posting chilled-out videos on many topics including Linux, FOSS, gaming, food, life etc since 2018. You can follow at:
➡️ @chris
Don't worry if it looks blank to you, that just means no one from your server follows it yet. Follow and its videos will gradually start showing up on your server too.
You can also follow Chris's non-video account at @ChrisWere
Mastodon FeedBoosted by ChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕"):
hamishtpb@mewblog.thepolarbear.co.uk ("Hamish The PolarBear") wrote:
Me, @ChrisWere and @uoou watched it so you don't have to!
video.thepolarbear.co.uk/w/5xb7apjc5a5XxWeYwvSpFJ
Space Virgin's latest looks at "Megalopolis" from F.F. Coppola...
Also features loads of us talking about unrelated nonsense but for once that all seems to be after the review.
Mastodon Feedbrib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:") wrote:
@glyph Yeah I get your pain.
Deciphering LLM security reports back when a new vuln was publicised each week was a world of pain in terms of the report's readability, this seems to be somewhat better (although I still spend a lot of time trying to piece together "why did it say that there?). Seeing the LLM-isms also makes me wonder how thoroughly it was human-checked; what sort of issues sneaked through because LLMs are good at hitting the LGTM part of our brains? But you mention it looks well-reviewed, and perhaps some people are better at reviewing LLM output than I am; my ADHD makes it better to do it all manually even when autocompleting something as simple as a function (I say with experience!).
That's not to say the report is necessarily inaccurate; but I guess I wish the writer would use their own words rather than relying so heavily on an LLM's, even if LLMs did the bulk of the analysis. The reasoning behind wanting the writer's own words is that it forces the writer to actually process the LLM output with their brain rather than risking a LGTM moment, and it saves the end user from parsing slop. If I was receiving an issue like this, I would probably encourage the writer to do this unless it was a super-urgent security thing.
When you're not in a position to make that sort of request...? I guess you can't un-see LLM content so if it's needed you might as well use the information. If it's not needed then I'd probably ignore it (and I generally lean towards muting people who post obvious LLM output, although that's usually in the context of Mastodon accounts rather than technical write-ups)
Mastodon FeedBoosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
gotosocial@superseriousbusiness.org ("GoToSocial") wrote:
Some updates on relays:
If you're running #AodeRelay and you want it to be compatible with users of #GoToSocial, you can update to the latest version of the AodeRelay software, which contains fixes and workarounds for various issues which were preventing it from working properly with GoToSocial:
https://codeberg.org/asonix/relay
We already know that any relay using the ActivityRelay software does not work with GoToSocial. This includes large relays like relay.toot.io. We're not sure yet why it doesn't work, but it's been almost a year since ActivityRelay was last updated, so we're not sure if it will ever work. We'll investigate this more from GtS's side and see if there's anything we can do.
Relays that we know work right now are FediBuzz, anything using the most up-to-date version of AodeRelay, and rel.re. There might be more, not sure.
@ gotosocial admins: If you have a relay stuck in "approved: not yet" state and you want to retry connecting to it, you can do this by removing all connections to the relay (pushes and subscriptions), and recreating them.
(How this works: when the last connection to a relay is removed, the Follow targeting the relay gets undone. And when you create a new connection to a relay with no prior connections, a Follow is sent targeting the relay.)
We may in future add something that lets you re-send a Follow to a relay without having to remove + recreate all connections, but for now that's how to do it :)
Hope this helps everyone battling to get their relays connecting properly.
Mastodon Feedget used to accepting slop, even finely-filtered slop, and eventually you relax and let the machine consume you, albeit indirectly. I don’t know if I am even equipped to address the wider implications here. it makes me want to quit open source forever. but if I quit, that just means the people who remain are that much less interested in even thinking about the problem, and the world gets just a bit worse.
this sucks.
Mastodon Feedof course all the problems are still there, indirected one level. this explanation seems plausible to me, but the involvement of an LLM does damage the credibility of the report. on balance I probably believe this one is accurate given my previous knowledge, but this eventually reduces to the vigilance decrement problem again: *this* time I can trust the output, but I would need to apply this same level of skepticism *forever* to every report like this, which is impossible.
Mastodon Feedfromjason ("fromjason.xyz ❤️ 💻 ✍️ 🥐 🇵🇷") wrote:
feeling like optimus in his prime
Mastodon Feedparticularly when you are not getting pray-and-spray output but thoughtfully curated and already-human-reviewed reports, not being forced to check them all yourself?
Mastodon FeedThis is a generalization of the “what do you do with valid LLM-generated security reports” problem; obviously you can’t ignore the fact that your users are at risk. Here, I cannot un-know what the problem is. What do you do when your complex bugs are analyzed with an LLM and correctly provide you with factual information about their causes?
Mastodon FeedThe thing that the LLM-derived artifact has provided me here is knowledge of what a very long-standing problem is. I could double-check all its work, but the author already *did* that; while there are clear hallmarks of slopbot writing style here, there's also quite a bit of evidence that a human had to extensively review the output and actually understand what's going on. Just logistically there is no way for Claude Code to one-shot a multi-platform multi-OS-version problem like this.
Mastodon FeedHere's a concrete example of the kind of "fun" ethics question that is going to be facing open source LLM critics in the coming years.
This is an *analysis* of a years-old bug report and kernel bug, developed using Claude. It's a good analysis, very thorough, from a previously-trusted contributor. It's developed out-of-tree so no LLM output is committed to the downstream projects. (I have my suspicions about issue & PR descriptions.)
https://github.com/graingert/asyncio-macos-loopback-rst-hang
Mastodon FeedBoosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
gotosocial@superseriousbusiness.org ("GoToSocial") wrote:
Hello sloth aficionados!
This is a security announcement for an issue that was fixed in #GoToSocial releases v0.21.3 and v0.22.0. If you are running an earlier version of GoToSocial please find time to upgrade to one of these releases as soon as possible!
The security vulnerability allows a remote attacker crafting (properly signed) ActivityPub activies to send Create, Update and Announce activities to a GtS instance that persist statuses / boosts as if they were written by a remote victim account. This happens due to our (previously) incorrect handling of multi-valued attributedTo and actor in the effected activities. We are not aware of anybody using this vulnerability in the wild, but it remains possible on all GtS versions v0.21.2 and below.
This was kindly disclosed to us by Lain of the Pleroma project, thank you Lain!
Timeline:
- June 23rd, Lain emailed
admin@gotosocial.orgwith details of the issue.- June 24th, a fix was merged.
- June 25th, versions v0.21.3 and v0.22.0-rc4 including the fix were released, along with a pre-disclosure warning of the security issue urging users to update.
- July 6th, security vulnerability publicly disclosed by this post.
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
The question isn't really "what does C say?" (we know: UB)
😂
Mastodon FeedGargron ("Eugen Rochko") wrote:
#Deadlock character interactions are peak and so are fan animations of them
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
To this end, we revisit an old idea: token rings.
Oh no.
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
oh, weewoo
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
okay, i have to hand it to the authors of this paper for a great title
Are Your Epochs Too Epic?
Mastodon Feedbaldur@toot.cafe ("Baldur Bjarnason") wrote:
“it’s not enough to build something, part 1 | by Doc Burford | Jul, 2026 | Medium”
https://docseuss.medium.com/its-not-enough-to-build-something-part-1-cd9442c250b0
> If you just kinda dick around and refuse to learn anything about how art is made, well… you’ll make shit.
Mastodon Feedbaldur@toot.cafe ("Baldur Bjarnason") wrote:
“Write – And the sea, Dent”
https://anticdent.org/and-the-sea-dent.html
> We built the tools that have now enabled extraordinary systems of mediation, centralisation, mediocrity, and mendacity. It sucks.
Mastodon Feeddysfun@treehouse.systems ("gaytabase") wrote:
NO leap second will be introduced at the end of December 2026 #bulletinC
Mastodon Feedsoatok@furry.engineer ("Soatok Dreamseeker") wrote:
Hey, reminder if you're new here:
Mastodon FeedBoosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
lowqualityfacts@mstdn.social ("Low Quality Facts") wrote:
Hey look, it's the first comic we ever made! My apologies if you speak Celsius.
Mastodon FeedBoosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
malteengeler@legal.social ("Malte Engeler") wrote:
RE: https://mastodon.online/@tarakiyee/116864478747652698
"There is no innocent ground for the machine to stand on [...] A cleaner mine, an open licence, a machine built to last twenty years instead of two, all of it can be real and the thing is still not decolonial if the people whose land and water it runs on cannot make it stop."
What a beautiful, strong essay. How much (and what kind of) technology is compatible with egalitarian and post-capitalist societies? The question really is at the heart of all struggle for better (digital) futures.
Mastodon Feeddb@social.lol ("David Bushell 🪿") wrote:
I've got an absolute beast of a blog post in the pipeline, but for now here's something stupid:
https://dbushell.com/2026/07/06/behold-the-perfect-algorithm/