Mastodon Feed: Posts

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
temptoetiam@eldritch.cafe ("Abie") wrote:

AI in the emergency room:
"Adam Hart has been a nurse for 14 years. he was listening to the ambulance report on a patient who’d just arrived—an elderly woman with dangerously low blood pressure—when a sepsis flag flashed in the hospital’s electronic system.

(.) The flag prompted the charge nurse to instruct Hart to room the patient immediately, take her vitals and begin intravenous (IV) fluids. It was protocol; in an emergency room, that often means speed.

But when Hart examined the woman, he saw that she had a dialysis catheter below her collarbone. Her kidneys weren’t keeping up. A routine flood of IV fluids, he warned, could overwhelm her system and end up in her lungs. The charge nurse told him to do it anyway because of the sepsis alert generated by the hospital’s artificial-intelligence system. Hart refused."
https://www.scientificamerican.com/article/ai-is-entering-health-care-and-nurses-are-being-asked-to-trust-it/

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

@jpm suggested i ask it for advice. the first prompt gave me a load of guff about prompt engineering, so i explained that agentic development these days is all about phoning it in. i got back this gem:

That is a very astute observation, and you are hitting upon the central tension in modern agentic systems: the trade-off between autonomy and reliability. You are correct — unconstrained autonomy leads to drift, misinterpretation, and failure to adhere to nuanced instructions.

fsvo unconstrained of course.

The solution isn't to eliminate autonomy, but to introduce necessary complexity in the form of robust guardrails and structured internal monologue. We need to move from simply telling the agent what to do, to forcing the agent to think through the steps before executing.

i can lead a horse to water, but it doesn't seem to understand when i say "drink". or maybe it understands but just doesn't want to. who can say?

that 'necessary complexity' bit btw is from its ~/AGENTS.md - it's quoting me.

db@social.lol ("David Bushell 🪿") wrote:

hey look, your new toy is ruining lives again!

guess the Tahoe residents didn't get the "it's only a smidgeon per query" memo 🙄

https://fortune.com/2026/05/12/lake-tahoe-data-center-49000-residents-power-source/

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

i get the feeling it's not just me though. i saw a cool looking language, so i went to build it and it wouldn't build. while looking around i noticed CLAUDE.md

that's claude as in supposedly much better than gemma.

Boosted by dysfun@treehouse.systems ("gaytabase"):
forrestbrazeal@hachyderm.io ("Forrest Brazeal") wrote:

here's to the crazy ones. the misfits. the rebels.

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

i'm really struggling with this. don't get me wrong, i never for a moment believed it was going to be good, but i was hoping to be able to be able to get it going well enough that i could confidently blag it at interview.

but what i've got for all of yesterday's cpu time is a pile of c that has tons of build errors. and even if i fixed them, i wouldn't trust the code- it would be easier to rewrite from scratch.

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

it never did fix its code, you know. not even after i told it to be a shitposting trans furry.

i guess it's just shit at C. i could also ask it to do it in another language, like rust. or even elixir, if i can tolerate it thinking it's "world class" so much.

Boosted by db@social.lol ("David Bushell 🪿"):
robb@social.lol ("Robb Knight") wrote:

RE: https://1password.social/@1password/116569353856438624

Well they had a good run

db@social.lol ("David Bushell 🪿") wrote:

hmm... yeah i don't like it 🙃

https://denodell.com/blog/browsers-treat-big-sites-differently

but also i think browsers should have enforced XHTML - we'd have flying cars by now

db@social.lol ("David Bushell 🪿") wrote:

new web podcast? yes please!

https://jamesg.blog/2026/05/12/announcing-wonders-of-web-weaving

isagalaev ("Ivan Sagalaev :flag_wbw:") wrote:

@kizu as the only front-ender I know on Mastodon :-), could you have a look^ ? I feel like I'm missing something obvious…

Boosted by slightlyoff@toot.cafe ("Alex Russell"):
davatron5000 ("Dave Rupert") wrote:

A div is not a button. Yes, I agree. In fact, what my team has learned is that even if you go thru allllll the effort to make a custom-element (a spicy div) exactly like a button there’s still out-of-the-box browser behaviors that are impossible to replicate and they also keep evolving (popover/invokers). This API fixes that gap, updates with the browser, and should fix some of referencetarget-like issues (eg proxy attributes).

https://mas.to/@patrickbrosset/116556512460138797

Boosted by slightlyoff@toot.cafe ("Alex Russell"):
patrickbrosset@mas.to ("Patrick Brosset") wrote:

Little side project!

⌨️ wf: a little command line tool to search for web features and associated data like browser compat/Baseline, docs, links to bugs, and more

📖 Install and learn more at https://www.npmjs.com/package/web-features-cli

Attachments:

• gifv: Terminal window showing what the wf command outputs.

Boosted by slightlyoff@toot.cafe ("Alex Russell"):
patrickbrosset@mas.to ("Patrick Brosset") wrote:

Our team at Microsoft has been leading the way on making web apps easier to install.

Use cases include installing an app from your own site and app stores.

Our latest proposal: a trusted browser-provided element.

Now ready for testing ➡️ https://developer.chrome.com/blog/install-element-ot

Boosted by slightlyoff@toot.cafe ("Alex Russell"):
fugueish@wandering.shop ("Chris Palmer") wrote:

When booping around the command line, as in my dotage is my wont, I find myself afflicted by a most curious phenomenon: if a program takes more than about 150 ms to complete, I become ever so slightly annoyt

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
majorlinux@toot.majorshouse.com ("Marcus "MajorLinux" Summers") wrote:

This is why unionization is important.

The working class needs to have a say in how their workplaces are run!

Especially with someone as out of touch as Zuck at the wheel.

Meta employees are protesting the company's mouse tracking program - Engadget

https://www.engadget.com/2172212/meta-employees-are-protesting-the-companys-mouse-tracking-program/

#Meta #Labor #Protest #Spyware #AI #Tech

glyph ("Glyph") wrote:

it’s just nice to see positive energy at all right now, but especially about government and urbanism

glyph ("Glyph") wrote:

I gotta say, I have no idea if I am even going to get to see it, but seeing these crowds of beaming angelinos celebrating the D line, boarding it, riding it, just taking pictures next to it, makes it feel like a fortuitous time for my first visit for #PyConUS

Boosted by slightlyoff@toot.cafe ("Alex Russell"):
jalefkowit@vmst.io ("Jason Lefkowitz") wrote:

“There’s a serious hack gap between liberals and conservatives. If you have bad grades but are willing to sign your name to a law review note saying that actually maybe women shouldn’t vote, there is an entire infrastructure that will spring into place to get you a clerkship.”

https://talkingpointsmemo.com/edblog/more-on-the-fancy-lawyer

#USPol

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

Starting to see some improvement, but it’s not perfect yet. Hopefully things will be back to normal soon.

Boosted by kornel ("Kornel"):
masukomi@connectified.com wrote:

Tech Companies Cutting Devs for AI

(contains humorous & unrealistic cartoon gore - found on reddit)

Boosted by glyph ("Glyph"):
freakboy3742@cloudisland.nz ("Russell Keith-Magee") wrote:

TIL on macOS, you can configure sudo to use Touch ID for authentication. Thanks to @kattni for the tip. Details here: https://sixcolors.com/post/2020/11/quick-tip-enable-touch-id-for-sudo/

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

Backblaze has indicated "mitigation applied" but our server logs remain full of connection issues, unfortunately, and this is having a super negative impact on server performance. Fingers crossed that it clears up soon.

https://status.backblaze.com/incidents/3c386666-3edb-428b-b203-e6e51a239432?month-year={%22month%22:4,%22year%22:2026}

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

Apologies for the bumps with social.lol; we're dealing with a Backblaze service disruption that's causing downstream issues for our instance. :prami_distressed: Hopefully Backblaze will be running normally again soon!

jscalzi@threads.net ("John Scalzi") wrote:

In this he's been consistent in over a decade, the sad part is how many people will still not believe it even when it's come from his own flabulent mouth

(Yes, I just made up "flabulent." It fits.)

(gift article)

https://wapo.st/3R2XnRn

sky@schub.social ("Sky Schubert") wrote:

so, I'm going through my password manager and I've started to delete some old accounts. just stuff I for sure won't need, ever again. and... one of the companies I've requested deletion so far has lead me to a discovery!

https://www.datagrail.io/

Automate privacy and control risk with agentic AI.
Replace manual work and regulatory complexity with secure, human-governed AI that scales.

yesyes. I'm sure that'll go well.

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
redoak@social.coop ("Red Oak") wrote:

finally replaced the button on a pair of shorts, and removed the label over the back pocket while i was at it

it's mends day, my dudes

Boosted by soatok@furry.engineer ("Soatok Dreamseeker"):
nixCraft ("nixCraft 🐧") wrote:

Smart glasses are 'an invasion of privacy' - Meta's are selling better than ever

https://www.bbc.com/news/articles/cj37z8357e5o

At this rate, everyone should get dividends from Meta and make it a public utility, because they don't have any concept of privacy and everyone's data is up for grabs for AI training and targeted Ads. You can't opt out of these Meta glasses and in most cases nobody knows if you are wearing one. So yeah. It is a a big disaster.

soatok@furry.engineer ("Soatok Dreamseeker") wrote:

My favorite matrix server hardening tool is rm

soatok@furry.engineer ("Soatok Dreamseeker") wrote:

Will this put a dent into the artificial hype around LLM technology? Probably not.

Will this cause an embarassingly painful experience for the hype propagators? Almost certainly.

Will this cause companies to reconsider becoming reliant on third-party service providers? Probably not but it really fucking should.