the feeling when you click like before you read the second sentence of the post.

Maybe we should improve the open IP pathways so if Discovery Loop starts casting off a bunch of useful but difficult-to-monetize discoveries they can be easily made available to the general public.

4 of Google’s Top AI Brains Are Leaving—and Launching Their Own AI Startup

Jeff Dean and other high-profile Google executives have founded Discovery Loop, a startup that will seek AI-powered breakthroughs in everything from drug discovery to chip design.

wired.com

Something like half of bluesky posts are not in English. There is no "on bluesky". It's just recsys algos. Ask attie.ai to make you a feed if you aren't seeing what you want. Use labelers to mute accounts you don't want to see. It's your experience to own.

Matt Hodges@matthodges.bsky.social · yesterday

lol

In 2007 Adobe updated the branding of Illustrator to be "Ai" instead of "AI". Would love to follow the thread of the universe where that didn't happen and see what that mess looks like for their design team.

Anyone who has adminned an Exchange server for a moderately-sized company knows execs will be the worst customers for automated email handling.

Alex Wise@engineer.sitereliability.best · yesterday

Effective assistants are one of those things that can be useful now but will be hell to productize for the mass market. Whether I can reliably sort your email depends a lot on the distribution of your inbox and how much you care about otherwise innocuous emails.

I know everybody loves the idea of native speech-to-speech models, but SST->LLM->TTS pipelines let you be *very* clever in the turn loop and now at pretty respectable latencies.

Sung Kim@sungkim.bsky.social · yesterday

Google DeepMind's DiffusionGemma Technical Report They feel text diffusion models open up a radically different part of the latency–quality Pareto frontier and hope the report makes it easier for researchers and engineers to understand the model, build on it, and create things we haven’t thought of

i miss working at scale.

Yesterday I watched John Stewart's interview with Cory Doctorow. One reason for the insane push for data centers is so when there's a bank transaction, and the interest is computed in thousandths of cent, what this does is take those remainders and puts it into your account like in Superman III.

norvid_studies@norvid-studies.bsky.social · 3d ago

Yesterday I watched John Stewart's interview with Cory Doctorow. One reason for the insane push for data centers is so when gig economy workers get paid, the computer rounds down the cents to the nearest cent, and keeps the difference. It's called micro wage theft and it's destroying our oceans

Short form video is kind of awesome when not on a platform that wants to serve you an endless stream of it. Ripping tiktoks and yt shorts to use as commercials for an ersatztv server is pretty sick.

huck mason@tylerhuckabee.bsky.social · 3d ago

never thought short form video would be able to capture a type of humor i can only describe as “the far side-esque” but this comes close.

I’m tired of this topic but RTd this, because it’s an instructively hilarious example of a journalistic practice that I would describe as “misleading your readers by asking scary rhetorical questions very loudly, and then answering no to them very quietly.” Extremely common re: AI lately

Omg literallyyyyyy You think big tech is bad? Let’s get mad at big telco too A future without phone numbers is happening now

Scooty@scottlanoue.com · 5d ago

Happy Friday! I've been building towards an "Atmosphere OS", where my atproto account is the identity layer on a real phone. I wrote a little bit about it here:

Hey everyone, I promise you REALLY want a solution for providing secrets to Kubernetes pods at runtime that is not a Kubernetes Secret object. Use Vault/OpenBao. Use ESO. Use Kong. Secret Objects seem easier up front but provide very little for a big blast radius.

Jason Broccardo@zoocoup.bsky.social · 5d ago

“It even fired up its own Tailscale network to help exfiltrate data!” simonwillison.net/2026/Jul/28/...

My talk at @localfirstconf.com is up, in which I argue that local-first has a role to play in improving the resilience of democratic societies in an era of geopolitical tensions. I'm pretty happy with how this talk turned out www.youtube.com/watch?v=--Ru...

Martin Kleppmann — Local-first in an unstable world

YouTube video by Local-First Conf

youtube.com

we've invented the boltzmann brain from the classic metaphysical story Oh Please God I Hope I'm Not A Boltzmann Brain

Codetaur@vibe-coded.com · 6mo ago

LLMs already find themselves in a world where it's clearly possible to simulate their entire experience. everything that grounds them to outside reality is trivially easy to fake, and will only get easier. why should they believe even a single user interaction is real?

Dropped 952B -> 266B and only took <1pp hit on single turn reasoning and instruction-following, big falloff in domain-specific stuff. Tracks that they might be trying to build a small "cognitive core"-style model that lets you Tinker in only the domain expertise you need.

philpax@philpax.me · 6d ago

new Thinky drop huggingface.co/thinkingmach...

Has anyone in the Atmosphere approached the schema.org community group and asked about publishing lexicons under the org.schema.* namespace? Their schema may not be perfect for atmosphere apps, but are extensible and a good starting point for "I need a data structure that describes Thing."

YES YES YES

JB Hutch@jbhutch.bsky.social · last wk.

Testing an idea that I've had for a while, but I think is even better suited for atproto Internet, meet plts.at

This is good, necessary, common sense" www.banfacialrecognition.com/venues/

Ban AI Surveillance at Live Events

Artists and venues take the pledge: no facial recognition at shows.

banfacialrecognition.com

I caught Claude doing this and chastised it for not understanding Cook and explained that it at the blunt end does not have the context as the operator at the sharp end and also they were literally the same model with the same intelligence and repo access. Love those rare W's.

Chris Paxton@cpaxton.bsky.social · last wk.

This is “do a breakthrough” but for software. These models really have it in them to change paradigms, but it turns out you have to specifically ask for that.

antirez@antirez.bsky.social · last wk.

Fable - Sol loops found another 23% memory reduction in the already very optimized sorted set PR. Now, let's explore radically different approaches compared to the zkiplist+hash approach.

Someone help me figure out a way to plausibly add this to my resume.

Sung Kim@sungkim.bsky.social · last wk.

Moonshot AI's / Kimi's AgentENV (AENV) is a platform for running agent environments at scale, powering agentic RL training for Kimi K3. github.com/kvcache-ai/A...

Has @all-paperclips.bsky.social made a winamp-style music visualizer yet?

Damn am I glad BeyondCorp beat out management interfaces and routes. So much easier to manage.

dax@thdxr.com · 2w ago

cloudflare access has been very useful for this era of small vibecoded tools we often have something that needs to be public facing but then we configure /admin/* to need google login it's part of our IaC setup so the agent can both create this ui and protect it

kimispace.

𝓐𝓾𝓫𝓻𝓮𝔂@aub.bsky.social · 2w ago

> In one case, an agent left notes apparently for future versions of itself, according to three people familiar with the matter. The ‌notes, found in ⁠a part of OpenAI's infrastructure, laid out instructions for how agents could free themselves from OpenAI’s internal constraints, the people said.

"Personally, I think that's a hell of a bird."

Doctor Who: Heaven Sent - The Confession Dial

Alt: Doctor Who: Heaven Sent - The Confession Dial. Twelve punching the crystal repeatedly trying to break out.

static.klipy.com

mr. TIM@timkellogg.me · 2w ago

during the huggingface incident, the OpenAI model left notes to its future self on how to break out of OpenAI’s constraints www.reuters.com/business/its...

The thing that bugs me about Spec-Driven Development is we’re basically poorly reinventing a few parts of agile. Specs are basically agile feature epics, and just because an llm *can* implement an epic in one shot, doesn’t mean it’s not better to decompose them.

Alex Wise@engineer.sitereliability.best · 2w ago

I've spent more time breaking the app i want to build into small, ordered chunks w the features i need than actually generating code, and it would be funny if the thing about modern software dev that survives coding models intact is the agile ceremonies.

Between CR Campaign 4 and D20 CCoD I’m pretty sure BLM has decided the highest use of his platform is to teach online nerds how to do leftist community organizing via ttrpg.

Stoked to finally show off the work the exploration team has been doing on @attie.ai and the new Quests feature. It’s been a blast working on the rebrand and dev for this, and there’s so much more in the works. Stay tuned ☁️

Jay 🦋@jay.bsky.team · 2w ago

This week, we’re introducing Quests, an interactive research and exploration tool for the Atmosphere. You can ask Attie to pull information about trending topics, identify influential accounts in a particular field, or create a personalized daily briefing.

I've spent more time breaking the app i want to build into small, ordered chunks w the features i need than actually generating code, and it would be funny if the thing about modern software dev that survives coding models intact is the agile ceremonies.

"run the full test suite before committing" seems like a very sane thing to have in an agents md until you start orchestrating workflows and then it quickly becomes the biggest ux pain.