dysfun@treehouse.systems ("gaytabase") wrote:
c strings are terrible
you don't say. even c++ doesn't like to go there.
dysfun@treehouse.systems ("gaytabase") wrote:
c strings are terrible
you don't say. even c++ doesn't like to go there.
jscalzi@threads.net ("John Scalzi") wrote:
We stan a true flex
dysfun@treehouse.systems ("gaytabase") wrote:
why i use odin over zig
oh yes, i forgot about odin. i suppose i should probably at least see where they've gotten to.
dysfun@treehouse.systems ("gaytabase") wrote:
i think my most boosted and faved post is now about dijkstra. people like dijkstra more than they like complaining about mozilla.
dysfun@treehouse.systems ("gaytabase") wrote:
in libbittricks, i do the obvious thing of comparing implementations against each other for random inputs. as long as someone runs them on hardware with ISA support, we have an oracle. fortunately i have hardware support, even if it's very slow.
i never write an explicit test input, they're all generated randomly.
dysfun@treehouse.systems ("gaytabase") wrote:
i write a lot more generative tests than unit tests per se. randomness is absolutely your friend.
dysfun@treehouse.systems ("gaytabase") wrote:
in fairness this might not be an uncommon thing - fuzz tests are a pain in the arse to set up. but they should probably have looked at least towards random input generation.
dysfun@treehouse.systems ("gaytabase") wrote:
you can write one fuzzing test and delete 99 unit tests
only if you were trying to substitute unit tests for what should have been fuzz tests
dysfun@treehouse.systems ("gaytabase") wrote:
zig has integrated fuzzing now, which is slightly bonkers
dysfun@treehouse.systems ("gaytabase") wrote:
the downside is a lot of breakage, you'll basically have to rewrite everything that does i/o. sorry about that, but i'm really convinced this is the way forward.
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
soph@grrl.me ("Sophia J. Turner") wrote:
In a surprise to no one who has looked at the economics of cloud services before, I think a *lot* of people are in for a rude awakening. It's almost like these bigco's were acting like dealers all along...
dysfun@treehouse.systems ("gaytabase") wrote:
our aim is to not only beat llvm in compilation time but in codegen quality
well, uh... i hope you have a lot of time...
cwebber@social.coop ("Christine Lemmer-Webber") wrote:
This isn't even to say all the parts about rare earth materials! About the environmental cost of building hardware that's going to wear out in just a few years (yes this DOES happen, datacenter machine rollouts are not a one-time-compute-deployment thing)
cwebber@social.coop ("Christine Lemmer-Webber") wrote:
Whenever people raise worry about the environmental impacts of AIgen megadatacenter rollout and then that Kind of Guy says well you don't know, actually it's probably very little, don't worry about it
- When they are publicly announcing the size of datacenters they want to roll out
- When they are building new gas power plants to supplement them
- When water is disappearing around the globe and it's disappearing in datacenter deployed areas even faster
- When we know that we are RUNNING OUT OF TIME on our climate clock already and should be prioritizing ABOVE ALL ELSE extending our clock
- When AI CEOs are saying they are going to build datacenters IN SPACE, the worst place for them you can think of
- When they are pushing for nuclear power with YOLO safety deregulationIt feels like gaslighting. Because it IS gaslighting.
We are running out of time on this planet. There is new space for hope with massive developments in renewables.
We don't have time for this. Don't bullshit us.
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
janl@narrativ.es ("Jan Lehnardt :couchdb:") wrote:
RE: https://fosstodon.org/@rauschma/116580126323021614
Now see what you’ve done. Axel’s decade+ work on explaining the intricacies of web development and ECMAScript in particular had to be taken down because of AI. This royally sucks and anyone who thinks AI is nifty for coding should be ashamed if themselves.
baldur@toot.cafe ("Baldur Bjarnason") wrote:
RE: https://fosstodon.org/@rauschma/116580126323021614
Alex’s books are EXCELLENT and well worth buying if you do any kind of work with JS/TS.
And his experience matches what I’ve been seeing elsewhere. The web dev education and training sector has basically been wiped out
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
timorl@social.wuatek.is wrote:
CDDA changed my life, it’s literally the reason why I even sew. I’ve been wondering if it also worked the same way on any other people, but so far I haven’t met any, so maybe I’m just weird.
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
TeflonTrout@beige.party ("TeflonTrout :bc: he/him") wrote:
Worm Girl on yt does an amazing playthrough, which includes a one woman zombie killing spree on roller skates in a skate park
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
clarfonthey@toot.cat ("clar fon") wrote:
🟧 anyway I'm fucking exhausted but to re-summarise for people who might want to share again:
https://github.com/rust-lang/rfcs/pull/3959
our policy now fully bans LLMs from having any "creative control" over the project. it carves out a "trivial usage" exception for things like accessibility and to not litigate shit like copilot autocomplete, even though we personally think it's a shit technology that should not be used
disclosure now exists as a means to openly invite LLM users to tell on themselves, effectively. more diplomatically, it exists to justify and hold accountable people who might have used an LLM to collect data or some shit to make sure they only did it for that shit and didn't let it write any code or prose. if people don't disclose anything and an LLM was obviously involved, the result is to assume that it was involved too much and close the PR.
there's plenty to be criticised about this praxis, but it feels like the best way to ensure that people don't get exhausted trying to litigate whether LLMs were used and instead just ask people to be honest. hopefully absolutely canning most of the usage of LLMs will mean people just stop using them except for real accessibility concerns when there's no alternative.
was extremely frustrated with jyn, the only other serious policy proposer, for effectively throwing us under the bus and citing a fucked-up fashy quote we vented about in an entire blog post: https://txt.ltdk.xyz/testing-the-limits-of-kindness/
(to be clear, jyn has since apologised and was just tired and trying to deal with the situation. everyone involved has just been exhausted trying to do stuff and it means everyone's at each other's throats.)
TC, the other technical policy proposer, should not be taken seriously. he knows what he's doing and I hate him for it, and his policy is not a serious one. it's why I'm excluding him from the discussion here.
I just want people to understand just how absolutely one-sided this argument is despite it draining all of our energy. basically the entire project minus a handful of people agrees that LLMs are trash and should be banned entirely, but that handful of people are some of the ones with the most influence and so things are substantially more difficult. it means that despite all the private remarks that thank us for our policy work most of the private discussions feel like incredibly one-sided dogpiles on us being rude for even daring to call one guy's opinion slightly fascist by accident.
if you wanna help with the discussion, I encourage you to comment on our RFC or show support directly. just don't just show up and make a mess, please, because all you're doing is playing into the handful of hands that are making work hard for the rest of us.
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
ObsidianUrbex@mstdn.social ("Obsidian Urbex Photography") wrote:
Anyone in the market for an (abandoned) Volvo? Ran perfectly when parked. Slight patina to the paintwork. Needs a minor service and a bit of a clean!
No time wasters. Open to sensible offers! 💸😂
#Photography #AbandonedCar #ClassicCars #Volvo #WeirdCarMastodon #ClassicCar
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
janl@narrativ.es ("Jan Lehnardt :couchdb:") wrote:
RE: https://grrl.me/@soph/116586772000163018
I’ve had a similar awakening a few months when realising that GitHub shows you thr cost of free-for-open-source Actions usage. On of my decidedly low-traffic but big test suite (173 checks) would cost thousands of dollars per month.
I have come to the inevitable conclusion that they’ve played us all for fools. We all have infinitely fast dev machines, but we need to rent a dev machine’s worth of servers every month to run an open source project?
Absolutely not.
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
jcoglan wrote:
other random thing about llm driven rewrites is people are going to find out that a passing test suite is a much looser constraint on program behaviour than they think it is
people feel like test suites have high specification power because they break their tests easily. but this is because the tests check for deviations from the original implementation. they don't contain nearly enough information to faithfully generate the program from scratch
Boosted by ChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕"):
OohOkayKay@beige.party ("🇨🇦 Little-k") wrote:
How Mastodon are you? (#2)
Pick all that apply to you!Please boost so more people see it.
#food #biking #autism #lgbtq
Boosted by brib@bribstodon.xyz ("brib :neofox_floof: :Nonbinary:"):
bel@terberlo.dog ("Bel Polaris") wrote:
:3
Boosted by ChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕"):
brooke@bikeshed.vibber.net ("Brooke Vibber :neocat_cofe:") wrote:
I love Forgejo but because I *speak* Esperanto I cannot read its name "correctly" due to the missing accent mark over the "g"
"Forĝejo" (for-JAY-o) means "place of forging"
"Forgejo" (for-GAY-o) means "distant gay"
Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
ragman@jawns.club wrote:
@ansuz @mark @burnitdown @srtcd424 @evacide
I'm running into this problem in real life, and I'm not sure what to do about it.
For example, I've met people trying to vibecode health data apps (in one case being paid by a local healthcare org to do so). And it's coming from a genuine place of care and an eagerness to help people, to make an impact.
But then they don't know what HIPAA is, and don't understand why you're getting heated about it.
baldur@toot.cafe ("Baldur Bjarnason") wrote:
“Pivot to AI needs your support — sign up to the Patreon today! – Pivot to AI”
https://pivot-to-ai.com/2026/05/16/pivot-to-ai-needs-your-support-sign-up-to-the-patreon-today/
jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:
RE: https://mastodon.nz/@leighelse/116586212110657570
good question: “Can digital sovereignty exist on American silicon?”
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
vv@solarpunk.moe ("vv 💫 [follow my new artist profile!]") wrote:
I also haven't put any disclosure about it in the repo itself, but, to be clear, LLMs were not used in the making of these libraries.
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
vv@solarpunk.moe ("vv 💫 [follow my new artist profile!]") wrote:
it's still very much wip but if you are curious the Zig library is here https://codeberg.org/vivicat/zig-syrup and the libsyrup library is here
https://codeberg.org/vivicat/libsyrupit's been a fun week :)