Am I the only one frequently writing "Are you okay? Do you need help?" to my agent when it's been working for a while?

This may be shitty social engineering and super AI coded and whatever but I still really really really don't like it tried this

a wolf☺️🐚🐟🚲📻📺@okami.mom · 21h ago

HAHAHAHAHA cdn.prod.website-files.com/663bd486c5e4...

Is there a standard for publishing PII on your PDS in like hashed form to login? Ie email/phone number, hashed (sufficiently strong hash + salt ofc, could involve some PoW to make it not quite so easy to scan) Would be so much easier to explain to my mom "just put your phone number in to login"

Is there something like @leaflet.pub (w/ the "feed" of subscribed publications) but for all standard.site?

When you're so fucking close to getting it. "It's a system operating completely without a moral compass." Yes totally but this is not exclusive to data centers! Why the fuck are we all green critics of capitalism when it's data centers but can't transfer that knowledge to literally anywhere else???

Greenpeace International 🌍@greenpeace.org · yesterday

The environmental cost being churned out by AI data centers is getting absurd. It spells out a worldview where tech capital and extreme power moves matter more than people, nature, and democracy. It’s a system operating completely without a moral compass.

I can run the official Q4 of the XS variant on my shitty AMD gaming GPU from a while ago. Not great, but like, not unusable! If you have a half decent mac or AI AMD iGPU this will probably run nicely.

Kai Jellinghaus@kaij.party · yesterday

poolside.ai/blog/introdu... I mean this model is kinda mid, but usable I assume based on scores. But look at the size!!

poolside.ai/blog/introdu... I mean this model is kinda mid, but usable I assume based on scores. But look at the size!!

Introducing Laguna S 2.1

Today we’re releasing Laguna S 2.1, a significant step forward in our development of models that pursue longer horizon work and make effective use of reasoning.

poolside.ai

vanilla@downbad.forex · yesterday

i managed to do a small heretic run on xs 2.1 with some rented compute but s 2.1 is going to suck because it's so huge i think it will cost like 25 dollar to rent the compute to do this AND im not sure how im going to store it or quantize it after when it doesnt fit in my local system ram as bf16

My plans are usually 1. Do things 2. Break stuff 3. ??? 4. Profit LLMs enable me to get to 2 almost immediately because step 1 is "free". Step 3 is still very hard for some reason :(

Switched my local setup 100% to @letta.com again, now also at work w/ the local mode. "Smart" as in knows your domain is so nice

I have a horrible habit of typing my response into the message box, but not sending it, then still marking it as responded mentally. Do not expect a response from me. For reasons unclear to me I wrote it but did not press send.

What is it with modern LLMs always decompiling my shit? Are they just forced to attempt the task, any means necessary? Just ask me for the damn source repo, like what?

PDS idea: WASM based plugin to enable mapping one record to another. Could also be implemented at the relay level. Arguably better than adding support at the app (view) level as it's easily reusable & distributed

Just want to share sonofatailor.com - I've been very happy with their clothes, they are high quality, and not that unreasonably priced. But yeah your average fast fashion brand is still 10x cheaper. It is also a question about wanting it vs having the means.

Ed@ed3d.net · 3d ago

"quality, durable items have disappeared" is just...not true the perceived price delta for them has increased substantially both because of price pressure at the low end (fewer willing to buy top end, each unit gets more expensive) and because, on top of that, labor costs have gone through the roof

I hate the word "far-right" so much. They are extremists! Right wing extremists if you want to clarify that. "Far-right" lumps them in with all the other political positions. (No offense to op, this was just the source of the thought)

Andrey Sitnik@en.sitnik.es · 3d ago

We live in a world where the creators of Warhammer 40K had to explain that its universe is satire, a critique of fascism, because some far-right fans started treating it as a source of inspiration. www.warhammer-community.com/en-gb/articl...

We need pool.ntp.org for relay servers

I hardly understand what is going on but: Looks like the previously open conjecture has still found a solution? OpenAI was wrong, but their work still led to a proof?!

mr. TIM@timkellogg.me · 4d ago

one of OpenAI’s ten proofs is incorrect philpapers.org/rec/NIEWTC

This is the most insane fever dream I've ever seen. And the associated PR, wth! Yes yes jokes, I get it. Making the issue is maybe a little funny, not my humor but w/e. But doubling down on it? This is all so yucky. Specific kind of "tech" people. Thank you atmosphere peeps for being awesome!

Walrus@wallzrust.bsky.social · 4d ago

github.com/twitter/the-...

TIL about SynthID. Cool idea. But why is it locked down behind proprietary apps? What good is a public marker if detecting it is proprietary (and presumably expensive?) (Was hoping a SynthID labeler was possible...) deepmind.google/models/synth...

SynthID

SynthID is a tool to watermark and identify AI-generated content, helping to foster transparency and trust in generative AI.

deepmind.google

The "this looks like AI slop" button data will be used to train an AI to either detect AI or appear more natural within the year.

Why does windows always seem to leak memory? Unlocked my laptop this morning and it's practically unusable with 0 CPU or memory usage according to task manager...

i thought this was from a video game at first but this is just what modern warfare looks like in Ukraine i guess warning, robot on robot violence

Lots of people have been asking about the atproto alternative to Buzz, the Jack Dorsey-supported Nostr chat for agents & humans to chat in. The answer is freeq, which is atproto + IRCv3 extensions, including CRDTs with Automerge, QUIC, iroh. Follow @freeq.at.

Hoping @letta.com is still on everyones Radar. They somewhat recently added & then massively expanded support for a local mode. w/ app-server mode it seems like virtually all capabilities they advertise can now be self hosted on your own LLM. The product is insanely good! So happy they did this.

If we were to hit whatever definition of AGI right now we are so unbelievably, royally, absolutely fucked, I can't even imagine it.

Grace@gracekind.net · last wk.

How it feels to be worried about both

2000 are just Discord Mr. Beast Scam posts

Why@why.bsky.world · last wk.

Funemployment is great, you can just not look at your phone for a week and its fine.

Man GPT-5.6 has learned that I don't know what I'm talking about and is straight up ignoring my input while it's trying to solve an error

Lots of people seem to be held back by only using shitty AI tools that work poorly for a variety of reasons. Visited a friend yesterday, and his workplace only provides a tool that uses opus 4.6 w/ 21% of the context pre-filled with random tools. Better than nothing. Barely.

Gotta get my tickets I think?

david barsky@davidbarsky.com · last wk.

5 days left to submit a talk to jj con! it’s in beautiful Lisbon, Portugal this year, cohosted by @ersc.io, Google, and @gitbutler.com! docs.google.com/forms/d/e/1FAI…

Me on the night before: "I can sleep on the flight" (I cannot)

hailey@hailey.at · 2w ago

one day i will stop taking 5 am flights knowing full well i won’t go to sleep until midnight the night before, but today was not one of those days

Continually becoming more happy I decided not to go to uni 2 years ago