Mastodon Feed: Posts

Boosted by glyph ("Glyph"):
pythonbynight@hachyderm.io ("Mario Munoz") wrote:

This weekend, I had the opportunity to give a difficult talk. It's not much of a surprise if you've followed my writing this year, and it's certainly a heavy topic.

And to be honest, I don't think there is too much new info there, particularly for individuals on this platform.

But connecting the dots into a full picture can sometimes have a deeper effect, and that is partly what I was attempting to do here.

The video has now been posted over here:

https://www.youtube.com/watch?v=bf2b1CBz-wc

#NBPy #LLM

1/n

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
trevorflowers@hachyderm.io ("T🌸🌸: dustmouth enthusiast") wrote:

Computers are still cool. We just have to work with hard constraints like those provided by a 9¢ chip.
https://chrisdell.info/using-rust-to-build-a-1-dollar-handheld-gaming-console/

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
grickle@mstdn.social ("Grickle") wrote:

Special. #grickledoodle #love #family #witch #wizard #robot #cartoon #art #drawing #funny #humor

baldur@toot.cafe ("Baldur Bjarnason") wrote:

I've always been on the outskirts of tech, living on the border between web and other media, like publishing, so I've always known that I wasn't fully a part of the industry, but I always thought I belonged to the field of practice and felt a kinship to the people working in it

That's largely gone

baldur@toot.cafe ("Baldur Bjarnason") wrote:

It had already faded quite a bit because of the rise of React but now my sense of kinship with the rest of web dev is largely gone.

baldur@toot.cafe ("Baldur Bjarnason") wrote:

I came up in web dev in the aftermath of the dot-com bust. Most of us believed in the web as a medium that would do good

Now I'm thinking this permanently skewed my idea of the character of those who do the actual work in the industry

(No illusions about the execs/managers, though)

baldur@toot.cafe ("Baldur Bjarnason") wrote:

I've tended to work with idealists: accessibility, education, publishing, etc so in hindsight I think I never updated my internalised assumption about the industry

The people who run it have always been mostly scumbags, but I basically always assumed that most of those doing the work actually cared

Boosted by glyph ("Glyph"):
mttaggart@infosec.exchange ("Taggart :ifin:") wrote:

RE: https://infosec.exchange/@mttaggart/116478503783177069

This is from July, but I missed it at the time. If anyone says "Oh but streaming video takes energy too," this is the response. Netflix is not planning 50GW of data center energy usage by 2028. They are not advocating for the use of federal lands for data centers and power generation—including natural gas. This document outlines a terrifying future for power usage and generation in the United States.

It's a threat in ways other data center usage is not.

https://www.anthropic.com/news/build-ai-in-america

Boosted by glyph ("Glyph"):
pikesley@mastodon.me.uk ("Flounder Mode") wrote:

@glyph

https://mastodon.me.uk/@pikesley/112013892778162041

adele@social.pollux.casa ("Adële 🐁!") wrote:

Is there someone using #SmolFedi with a #Mastodon version 4.5+ ?

Could you try to "Quote" a post (this one for example), using SmolFedi, to see if your post is rendered correctly, including the quoted post?

Thanks for your help ^^

https://smolfedi.pollux.casa/

#smolweb #fediverse

glyph ("Glyph") wrote:

like the fact that LLMs "really can" do some parts of programming reads to me not as a demonstration of their economic utility but as a brutal indictment of our tools *for* programming. it feels productive to have these things emit mountains of boilerplate because *we constantly need mountains of boilerplate* and gosh I wish we could address that problem

glyph ("Glyph") wrote:

for example, long-term, I think one of the uses of LLMs might be as a sort of cognitive red-team. right now the economic orientation towards them is "look at all the work these things can do!" but in a just world, you might use an LLM to automate part of a task, and *the fact that an LLM can automate it* could be seen as evidence that the task itself is unnecessarily adversarial, or duplicative make-work. if spam can solve your problem maybe you're solving the wrong problem

glyph ("Glyph") wrote:

relatedly my personal emotional orientation towards the AI chatbots themselves has been evolving a little bit. I needed to do small bits of experimentation, mostly to inform my skepticism. doing that in the past largely had the valence of constant frustration and anger. now I feel more… mournful. LLMs are doing something really fascinating, and it would be so great if we could actually study it to discover what it *is*, instead of pretending that whatever that thing is is "thinking" or "work"

glyph ("Glyph") wrote:

this might sound depressing but it's actually hopeful. if we could get everyone to recognize this basic reality there's a very different course of action we could take to address it, which is not "constantly freaking out about robots taking our jobs"

glyph ("Glyph") wrote:

you will not lose your job because an AI will replace you. you *will* lose your job because the white-collar criminals driving AI's fraudulent circular financing are going to cause a global financial crisis when the bubble pops though

denschub@schub.social ("Dennis Schubert") wrote:

can the web please stop breaking all the time? I'm tired.

Boosted by glyph ("Glyph"):
grimalkina ("Cat Hicks") wrote:

- joined new patient community
- immediately recruited to the Evidence War Against Insurance because I can read papers
- now providing lit reviews to fellow patients

I am grateful for the skills I have. However, it would be nice if I could spend some living and breathing moments not Doing Work. The world is so critically cut off from evidence that this is impossible for now.

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

I really wish we'd used the term 'confabulate' instead because that's what's going on, mixing things together to make plausible stories and explanations.

That mostly works, but doesn't change its fundamental nature of _making shit up_.

EmilyEnough@hachyderm.io ("Emily 🏳️‍🌈🏳️‍⚧️") wrote:

Ooooh, so close. Maybe next time.

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

EVEN WHEN THEY ARE WORKING RIGHT, "AI" systems are hallucinating. They hallucinate in line with reality most of the time, and sometimes they diverge. That's the whole thing. They're _useful_ but when they're _right_ that's either a function of guardrails and checks that've put in place, or an accident.

Usually it's an accident. Plan accordingly.

Boosted by aredridel@kolektiva.social ("Mx. Aria Stewart"):
bridget@gts.sharitt.com ("Bridget") wrote:

The untrustworthy AI agent you plugged into production systems without any safeguards that deleted important data didn’t “go rogue”, you’re just a dumbass

EmilyEnough@hachyderm.io ("Emily 🏳️‍🌈🏳️‍⚧️") wrote:

Only here for less than a day so I did a quick speed-run around midtown last night trying to see what I can. I wish I had a couple of weeks free to sight-see instead of a couple hours.

adele@social.pollux.casa ("Adële 🐁!") wrote:

RE: https://snac.mro.name/marcus/p/1777379671.110468

Let me know what you think about it ;-)

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
PavelASamsonov ("Pavel A. Samsonov") wrote:

Tools like Claude Design have an assumption baked into them: that "productivity" is fungible, and more "productivity" of artifacts leads to more value.

But if anything, the problem is that there are too MANY artifacts — and decision-making within your company begins to take on a "garbage can" model.

The solution for designers is to step out of clock time work, and think on calendar time.

https://productpicnic.beehiiv.com/p/claude-design-is-a-clock-time-solution-for-a-calendar-time-problem

#UX #UXDesign #AI #LLM #tech #softwaredevelopment #design

pzmyers@freethought.online ("pzmyers 🕷") wrote:

True Love.

https://freethoughtblogs.com/pharyngula/2026/04/28/shes-giving-me-signals/

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
hrheingold ("Howard Rheingold") wrote:

I think this is one of the most important things I've ever written.

Thinking about thinking about #thinking
cognitive hygiene for 5 year olds

#ai #learning #llms

https://www.patreon.com/posts/156285478

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
cmconseils ("Lady Laura :bongoCat:") wrote:

Dear restaurant owners:
We ALL hate the QR code menu.
Stop.
-everyone

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
MikeDunnAuthor@kolektiva.social wrote:

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
augieray ("Augie Ray") wrote:

The dot-com and #AI bubbles aren't identical, but they sure look similar to me:

https://www.linkedin.com/posts/augieray%5Fopenai-misses-key-revenue-user-targets-in-share-7454885241468424192-lsiU?utm%5Fsource=share&utm%5Fmedium=member%5Fdesktop&rcm=ACoAAAARAO4BAPx0-DzizrQdFWXScUx1J1Sa0xc

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
thejapantimes ("The Japan Times") wrote:

North Korea dramatically increased executions during the COVID-19 pandemic, particularly for consuming South Korean dramas, K-pop and other foreign culture and political offenses, a report has shown. https://www.japantimes.co.jp/news/2026/04/28/asia-pacific/north-korea-executions-covid-19/?utm%5Fmedium=Social&utm%5Fsource=mastodon #asiapacific #northkorea #capitalpunishment #covid19