Mastodon Feed: Posts

aredridel@kolektiva.social ("Mx. Aria Stewart") wrote:

I think it's time to get off the big “code forges” entirely (I say, having just migrated github -> gitlab, because github's quality is deeply wanting, among other factors)

Gitlab I am also finding wanting in quality, and given today's announcement, I expect that to be worse not better. The team restructuring they talk about is going to make user interface cohesivity, already a problem, far worse. (Conway's law!)

I'll probably be moving my stuff to a local forgejo, though not super happily. I'll be figuring out CI. Again.

And most importantly, I'll be submitting the forgejo instance to the Software Heritage Foundation for indexing.

This is the thing that had kept me on the big providers, and I realize is actually solved. I don't want what I create to evaporate if I get tired of ops or something happens to me. Too much software availability is fragile.

jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:

I am not so sure this is a harbinger of good things

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
rysiek@mstdn.social ("Michał "rysiek" Woźniak · 🇺🇦") wrote:

Software has been "built by machines, directed by people" for decades.

That's what compilers and linkers do, that's what uncountable lines of Bash and endless CI/CD pipelines are – machines building software, directed by people.

And for decades, the bottleneck has not been churning out code. It was code review, it was quality control, it was bug fixing. AI slop makes that *worse*, not better:
https://freakonometrics.hypotheses.org/89367

GitLab, and the rest of the industry, is solving for the wrong problem.

dysfun@treehouse.systems ("gaytabase") wrote:

this is really quite bizarre. two interpretations:

they see "slop" as a badge of honour

LOL, ok, but i wouldn't put it past some of these fucks.

they see themselves as an oppressed class and are attempting to 'reclaim' the 'slur'.

quite aside from the fundamental lack of understanding going on, this would be hilarious.

dysfun@treehouse.systems ("gaytabase") wrote:

LOL, here's the repo description:

pi extension to empower a curl request with coding agent capabilities (slopcoded)

unless i installed some firefox extension i don't remember, they put that in the repo description

dysfun@treehouse.systems ("gaytabase") wrote:

why aren't there coding agent capabilities in curl already @icing ? can't believe someone had to make a super curl smh

dysfun@treehouse.systems ("gaytabase") wrote:

pi-super-curl — Extension to empower curl requests with coding agent capabilities

what does this even mean?

Gargron ("Eugen Rochko") wrote:

I need a non-technical guide for installing and using #Linux in Spanish for my mother-in-law. Anyone know a good one?

jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:

yup

https://www.theguardian.com/us-news/2026/may/12/fentanyl-test-strips-samhsa-funding

Boosted by aredridel@kolektiva.social ("Mx. Aria Stewart"):
EndlessMason@hachyderm.io wrote:

@aredridel
11 = I'm a millennial and I am making a phone call to book an appointment with a pain management specialist

jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:

"after years of Washington’s preventing Chinese companies from buying certain advanced technology products, firms like DeepSeek and Moonshot AI are starting to design their A.I. systems around the constraints rather than waiting for them to disappear."

https://www.nytimes.com/2026/05/12/business/china-semiconductor-ai-deepseek.html?smid=nytcore-ios-share

jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:

want some idea of how PRC scholars see the present historical setting?

http://rdcy.ruc.edu.cn/yw/HOME/index.htm

Boosted by kornel ("Kornel"):
dystroy@mastodon.dystroy.org ("Denys Séguret") wrote:

Here's a strange situation:

thousands of #Rust developers use #bacon, #broot, #dysk, or #lazy-regex every single day — tools I wrote, maintain, and improve for free.

Their companies, though? None of them want to hire their author.

If you use my tools at work and your company does #Rust, I'd really appreciate a hand landing a job or freelance mission. A boost goes a long way. 🙏🦀

https://dystroy.org

Boosted by kornel ("Kornel"):
barthalion@treehouse.systems ("Bart Piotrowski") wrote:

GitLab could have won by not doing anything at all but they would rather be renaming issues to work items, adding a dumb split view for the issue tracker, and working on other misfeatures no one asked for, anything but fixing 5+ years old bugs and half-assed functionality. And this doesn’t even touch the layoffs and “AI first” pivot.

dysfun@treehouse.systems ("gaytabase") wrote:

Despite the differences in approach, both OpenClaw and Pi follow the same idea: LLMs are really good at writing and running code, so embrace this.

think i've found your mistake

dysfun@treehouse.systems ("gaytabase") wrote:

Over the last few weeks I became more and more of a shill

Hey, you chose the word.

Who am I kidding? You let the vibe cod choose the words.

dysfun@treehouse.systems ("gaytabase") wrote:

If you haven’t been living under a rock, you will have noticed...

You should try it under this rock, it's real comfy

dysfun@treehouse.systems ("gaytabase") wrote:

large language models shall henceforth be known as 'the vibe cod'

dysfun@treehouse.systems ("gaytabase") wrote:

okay apparently this is a thing, whatever it is

dysfun@treehouse.systems ("gaytabase") wrote:

i'm sorry what?

neatnik@social.lol ("Neatnik") wrote:

I had a wonderful time chatting about the web with James! And I’m super excited for his new podcast. New episodes every Tuesday! https://jamesg.blog/2026/05/12/announcing-wonders-of-web-weaving

dysfun@treehouse.systems ("gaytabase") wrote:

omg you guys they've discovered functions.

https://pi.dev/docs/latest/prompt-templates

Boosted by neatnik@social.lol ("Neatnik"):
xandra@tilde.zone ("alexandra") wrote:

a podcast about the independent web? yes pls!!! @jamesg.blog interviewed adam (@neatnik, omg.lol) for the very first episode! what a great listen of two folks just enjoying discussing website building as a hobby. so dope! (and if you’ve never built a website, you absolutely should give it a try!)

thank you, james, for Wonders of Webweaving:
https://web-weaving.jamesg.blog/1/

(and adam, i’d always read you with a southern accent!!! 🤣 surprise!)

#indieweb #indie #podcast #webdev #webdesign #personalweb #web #internet #SmallWeb #webrevival #coding #html #anticapitalist #freeweb

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
jenniferplusplus@hachyderm.io ("Jenniferplusplus") wrote:

Anyway, AI fans keep claiming that AI is like the industrial revolution, and yeah, it is. But somehow, for some reason, people broadly view that as a good thing. But it wasn't. It was this. It's the thing I just described, but for physical goods. And now they're coming after art, and science, and correspondence, and law, and medicine, and bookkeeping, and it is incomprehensible to me that anyone at all would be in favor of this

dysfun@treehouse.systems ("gaytabase") wrote:

so the other thing i've been meaning to look at is extending my vibe codding apparatus. it comes with a handful of things like read and write files and execute a script with bash. yes, bash. no, it won't work with sh, it will gaslight you with fake output if you try and make it. nice touch that totally didn't wind me up for half an hour working out what it had done.

anyway, i am thinking web search might be useful as a thing, since at the minute it would rather guess than access the internet. of course how i'm supposed to actually get it to use it instead of hallucinating is anybody's guess...

looking at the docs, it seems to be... more text for the LLM. this has proven so reliable so far, i have great confidence 😬

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
inthehands@hachyderm.io ("Paul Cantrell") wrote:

I’ll spare you all the rambling 50-post thread I could write about this, but here’s a tiny slice:

To make software that creates a great experience, the team designing and building it needs to have room to prioritize that experience — to sometimes prioritize the person •using• the software over other stakeholders.

3/

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
reiddragon@fedi.catto.garden ("Reid :blobcat_happy:") wrote:

@jonny I sound like a broken record at this point, but whenever the legal status of LLM training data is brought up I can't help but think of Aaron Swartz and how he was treated like he did the heist of the decade for trying to release some academic papers to the public

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
koen_hufkens ("Koen Hufkens, PhD") wrote:

The slow death of the power user.

"This isn’t an accident. This is the result of two decades of deliberate, calculated effort by the largest technology companies on earth to turn users into consumers, instruments into appliances, and technical literacy into a niche hobby for weirdos. They succeeded beyond their wildest expectations"

https://fireborn.mataroa.blog/blog/the-slow-death-of-the-power-user/

#technology #tech #sustainability

dysfun@treehouse.systems ("gaytabase") wrote:

the thing i just find genuinely baffling is the overall structure of this.

let's say i was sold on the general concept of machines writing code and was designing a system to make that happen.

i just would not design it this way. at all. everything is thrown at the llm where other algorithms would be quicker and do a better job. there is no structure, because LLMs want free structured text as input and output and nobody is even really trying to constrain that (outside of research anyway...)

dysfun@treehouse.systems ("gaytabase") wrote:

anyway, assuming a typical backlog of infinite length, and generally that nothing is getting in its way, the real problem is quality.

i saw from @dakkar playing yesterday that hosted gemini is a little better than the self-hosted gemma model i'm using (actually he's using a slightly less compressed version of gemma than me to compare it against).

one thing i would be interested in is quantifying how much better the paid models are. because i'm always being told that the latest models change everything. this always seems to ignore the fact that the latest models are presumably to blame for all the outages we see on a near constant basis these days.

anyway i am decidedly not paying for any premium services (not least because they're all fash), but anyone with an employer who is already paying and enough of a gpu to run models locally could do this if they were interested.

as i go further down this ridiculous rabbit hole, what i find myself doing is trying to think of ways to steer it towards better output. prompting carefully seems more like trying to find the cheat code than a reliable process.

if the model produced significantly better code, it should generally work better and i could see how people could convince themselves it was 'good', though i'm fairly sure already it wouldn't meet my personal standards of good.