There's a long list of content-based signals we might want to rank on. We are limited by the classifiers we actually have. BUT we are working on LLM scoring with arbitrary prompts! The challenge is performance, but we have a cunning plan docs.google.com/document/d/1...

How does GreenEarth determine "constructiveness"? We use the Google Perspective API but not the classic "toxicity" score. Instead we combine many of their "bridging" attributes. The weights are from our previous experiment. github.com/greenearth-s... rankingchallenge.substack.com/p/its-possib...

Fusion power is gonna go a lot like AI. Empty promises for decades, then suddenly here faster than anyone can adjust to.

GreenEarth (@greenearth-social.bsky.social) will be the first feed to explain the algorithm, tell you why you got each post, and let you adjust all the parameters. Here's an early version running on our staging server. We're ~2 weeks away from public beta.

I love love love Lee's essay on "Cultural Pessimism Technology Criticism." As a technology designer, I understand the fundamental importance of criticism -- I just want better critics. (Yes, this is a subtweet about AI commentary on this platform. But also, the essay is great, you should read it.)

Quantized model

We've added a new design to the library, rated likely: 👍 Adding Balanced News In a 6-month experiment from @jonathanstray.bsky.social et al. inserting news from credible, ideologically balanced outlets into feeds reduced partisan animosity. www.prosocialdesign.org/library/addi...

I'm not gonna say he did this because of us, but I know for a fact that his boss read my essay, "Making Social Media Algorithms Social" Competition is good, actually! www.greenearth.social/p/making-soc...

Cough cough something something knowledge workers AI

OH at CHAI: "This week I'm working on an essay on free will." "What made you choose that?" I love my lab.

The app team has been giving me a hard time that not enough atproto knowers have signed up for this community research project. Fill out the form, tell the team what you want from communities.

I sometimes hear people express concern that bridging-based ranking will homogenize discussions, flattening minority perspectives. Many answers to this, but the first one is: not all minority perspectives are worth preserving. I'm sure you can think of a few. We need a criterion besides "minority."

This clever map shows us what time it is across the world, which regions are awake, working or sleeping, all rotating in sync with the Earth. Go to the source where the image only rotates by 3.75 degrees every 15 minutes to remain up to date. Source: buff.ly/tyQ2Z3g

Honestly my main use for Claude Cowork is that it can drive my browser to fetch things that are blocked from robot fetching 🤷‍♂️

The funniest thing about the Random feed is that you definitely get furries. www.greenearth.social/p/welcome-to...

The best discussions about how feed transparency should work, starting with this diagram, are happening over on X. C'mon folks we can do better here.

GreenEarth will be the first fully transparent, controllable, open-source feed. Does this diagram help you understand how the core algorithm works? Try the pre-release version now! survey.qualtrics.com/jfe6/form/SV...

You cannot convince me that a technology where I can type this into a text box and expect to get an interesting, appropriate, and working output is not absolutely astonishing.

I've discovered my "Following" tab is not a great experience for me on BSky, probably because I populated it by porting over my Twitter network. But the overlap ends up being mostly people who are mad about politics... which is ok... but... GreenEarth feed is a much more balanced experience for me.

@faz.ms tell me about your "productive disagreement" app? We've just built a productive disagreement feed!

gradually replacing my colleagues' unix tools with claude until they notice echo 'man() { claude -p "how do I use $*"; }' >> ~/.bashrc

Humans are so creative! Look at how much control over her body this girl has! Also BANANA BREAD HELL YEAH

IT'S WORKING! After months of engineering and tuning (thanks team!) the new @greenearth-social.bsky.social feed is reliably serving me an excellent social media experience. It's already engaging and constructive. Transparency and controls coming soon! Try it! survey.qualtrics.com/jfe6/form/SV...

Tools from operations research solve so many problems and the fact they’re not super widely deployed really tells you that many bottlenecks are not technical

You may remember we did a 10,000 person experiment testing AI-powered healthier feed algorithms. Well, now we’ve built it as a product for BlueSky/ATProto. We are currently recruiting pre-release users to test it out and give us feedback. Want to try it? survey.qualtrics.com/jfe/form/SV_...

If you've ever wanted to mainline the Internet, try our new Random feed. www.greenearth.social/p/welcome-to...

FINALLY managed to load my M4 Macbook enough to get Spotify to stutter. Occasionally. Modern hardware is very powerful, hard to max out.

After playing Cyberpunk 2077 for a bit, I agree with Gibson's famous quip: it feels more like a reskinned GTA than the NeoTokyo of my cyberpunk dreams. This vision is less chrome and more gangsta. But there's history here: the Cyberpunk tabletop RPG was created by Mike Pondsmith, from Oakland CA.

Claude Cowork sure is introducing a lot of people to the problems of multi-process synchronization for the first time. I have no idea how mortals are handling this.

Is this a good logo for the GreenEarth feed? It's a healthier, user-controllable, open-source, LLM-powered, transparent feed we're building -- now in alpha testing. Try it? Tell us what you think! bsky.app/profile/did:...

Lots to think about when we ask AI for relational advice -- what *I* usually think about is how it will affect human conflict. Will AI make us smarter about our differences, or will it start wars?

Some radical experimentation happening in what a social media client can be. This one has a poetry mode.

This articles made me stop and think. As someone who enjoys discussing ideas, it was a good reminder that most arguments don't change minds. Building things, listening when people ask, and staying open to feedback usually has a much bigger impact. wangcong.org/2026-06-30-w...

X head of product says the most monetized poasters are a bad user experience, actually.

I want to make sure AI doesn't incite human conflict. It's sometimes hard to explain what I do, but that's the core of it -- it won't happen automatically. And we're making progress! Both theoretically, and in field experiments that test how AI alters human relationships.

Do you have OPINIONS about ALGORITHMS on social media? Here at Greenearth.social we're building healthier feeds and we want your opinions about what to build. Do you have a few minutes to fill out our survey? survey.qualtrics.com/jfe/form/SV_...

Really good example of negativity bias today on BlueSky: Lots of liberal anger over SCOTUS rulings on trans rights, campaign finance. Very little discussion of SCOTUS ruling upholding birthright citizenship.

Hello I am in Montreal for the week! Anyone interesting in AI safety I should meet here?

I love the reaction of cat lovers upon seeing a cat. Every single time the level of excitement is like they've read about cats for decades but never actually seen one in real life and they're SO EXCITED. Every. Single Time. Even if it's the hundredth cat they've seen that day.

Curmudgeon take: the reflecting pool saga is irrelevant. It's not a lot of money and very little is at stake. Discussion of it is mere liberal snark. Most Americans do not know and do not care, and would be put off by the obsessive discussion of it here.

One idea we have for GreenEarth feed: a slider that goes between "Community" at one end and "Serendipity" at the other, controlling how far out of the social network we go to look for interesting posts.

Do you have OPINIONS about algorithmic feeds in BSky? We're building them, and we want them to be awesome. Who wants to talk to me about this? This is your chance to be heard by algorithm designers (me! and the rest of the GreenEarth.social team)

Does anybody here use a tool that allows simultaneous posting to BlueSky and X?

Why have conservatives come to reject the word "democracy"? "Their question is not 'should America be a democracy?' but 'has American democracy remained faithful to what makes it legitimate?'" conservativestudy.redassociates.com

Conservatives no longer trust the word "democracy." But they still believe in a functioning, citizen-led government. Fantastic new research unpacks what it is that they believe in, and why they think Trump is restoring the nation. www.theatlantic.com/ideas/2026/0...

AI safety typically assumes one well-meaning user. I'm working on the case where two of them are at war.

Zero-calorie sweeteners are just reward hacking, amirite?

Treating people with dignity and fairness is the right thing to do. But what if applying that fairness to how rule violations get handled could also reduce harm and foster healthier engagement? A primer on due process in content moderation ⤵️ www.prosocialdesign.org/blog/due-pro...

Humanity's ability to know, reason, judge, and act well is the foundation of science, democracy, crisis response, & management of AI itself. AI poses serious risks to that foundation. New paper on epistemic risks by 30 experts calls for attention and proposes solutions. Link in thread.