Sign in

Jonathan Stray

@jonathanstray.com
1.1K followers 1.1K following 549 posts

Making @mysky.social, the feed algorithm you control (it’s the most fun too). Working on AI and inter-group conflict research @chai-berkeley.bsky.social.

PostsRepliesMedia
Jonathan Stray @jonathanstray.com · 7h
it's living rent free in my head tbh
000
Jonathan Stray @jonathanstray.com · 29/09/2026
ngl, turning down the "politics" slider on my mysky.social feed has improved my BlueSky experience immensely. It's also something we repeatedly hear in user testing. Discuss.
mysky.social
MySky Social | Jonathan Stray | Substack
Open-source AI-driven Social Media Feeds for Everyone. Click to read MySky Social, by Jonathan Stray, a Substack publication with hundreds of subscribers.
0132
Jonathan Stray @jonathanstray.com · 29/09/2026
I suppose I have a much broader view of what healthy public discourse looks like. I *want* all comers to be to on the same platform, because the alternative is epistemic fragmentation which leads to worse ills. Of course this only works if you have constructive algorithms. Which we do, now...
210
Jonathan Stray @jonathanstray.com · 29/09/2026
This may be sooner than you think www.linkedin.com/pulse/w-soci...
linkedin.com
W Social Launches Livestreaming with Verified Live Conversation. European infrastructure connects verified users across the AT Protocol
Strasbourg, Stockholm, 16 September 2026 – W Social has launched livestreaming on its European social media platform, choosing European Commission President Ursula von der Leyen’s State of the Europea...
020
Reposted by Jonathan Stray
MySky Feed @mysky.social · 29/09/2026
You don't have to let someone else decide what you see. MySky is the first feed algorithm that you control. Open source, open science, and made just for you. Try it: bsky.app/profile/did:...
032
Jonathan Stray @jonathanstray.com · 28/09/2026
It's a reference to Wargames, the movie, where the rogue AI eventually learns through simulation that launching the missiles is a bad idea. Interestingly, I am starting to consider self play to teach conflict resolution to AIs.
110
Jonathan Stray @jonathanstray.com · 27/09/2026
Now do global thermonuclear war
110
Reposted by Jonathan Stray
Grace @gracekind.net · 25/09/2026
I made a little Jev-based language model! It uses beam search + Jev as a classifier to judge branches based on coherence and "success" (whatever that means). I've barely talked to it so I don't know how it'll behave or if it'll be interesting at to talk to at all, let's find out! @jevbot.bsky.social
542454
Jonathan Stray @jonathanstray.com · 26/09/2026
inorite
100
Jonathan Stray @jonathanstray.com · 26/09/2026
Set your MySky window to a week and have it find the bangers for you!
121
Jonathan Stray @jonathanstray.com · 26/09/2026
I've been using CapCut to edit video for the first time, and it's fighting me. Very AI, lots of presets, but very little of the control I'm used to from After Effects, Final Cut, etc. There's a "motion blur" filter but I can't change the direction. It feels super dumbed down and I hate it.
000
Jonathan Stray @jonathanstray.com · 25/09/2026
Have you tried @mysky.social? Really nice, controllable feed. I'm biased of course but it really has transformed my BlueSky experience for the better.
010
Jonathan Stray @jonathanstray.com · 25/09/2026
This is the first piece of AI art that I thought was good, actually. It's stylish, it's funny, it made me feel things.
020
Jonathan Stray @jonathanstray.com · 25/09/2026
One of the things that is surprisingly good about MySky is that when I get a banger or a dud I can go to the "Why Am I Seeing This?" page and find out. It makes me think about my settings, my network, and the algorithm.
040
Jonathan Stray @jonathanstray.com · 25/09/2026
asking for a friend?
100
Reposted by Jonathan Stray
Jonathan Stray @jonathanstray.com · 25/09/2026
Nice. Cold start is a hard problem. For the moment MySky has settled on retrieving posts for the average embedding vector of all MySky users. As opposed to all BSky users. This finds interesting stuff without assuming you like either lefty politics or furries (no offense to either).
021
Jonathan Stray @jonathanstray.com · 25/09/2026
Nice. Cold start is a hard problem. For the moment MySky has settled on retrieving posts for the average embedding vector of all MySky users. As opposed to all BSky users. This finds interesting stuff without assuming you like either lefty politics or furries (no offense to either).
021
Jonathan Stray @jonathanstray.com · 25/09/2026
Yeah a labeler is a really interesting idea! Stay tuned. We are also working on personalized definitions of constructiveness that relate to the communities you are in.
010
Jonathan Stray @jonathanstray.com · 25/09/2026
Surprisingly shitty. Not my usual experience though, they’ve worked great for me.
100
Reposted by Jonathan Stray
MySky Feed @mysky.social · 24/09/2026
We're more than just a pretty feed. MySky is an API for the ATProto community to build production recommender systems. Here's how to query our search indices, candidate generators, and scoring models. What algorithm do you want to build? atproto.com/blog/mysky-f...
atproto.com
MySky: Open, Controllable Feed Infrastructure for the Atmosphere
MySky is an algorithmic feed running on Bluesky, with a settings panel that shows you the whole ranking pipeline exactly as it runs.
077
Jonathan Stray @jonathanstray.com · 24/09/2026
This is correct. We are involved in efforts to replace the classifiers. They're hardly the end-all of prosocial classification, but they are useful, well validated (vs human labels), and developed from plausible social psych theory.
020
Reposted by Jonathan Stray
Bart⚓️ @bartgonnissen.bsky.social · 23/09/2026
The Facial Expressions of Tethys by artist Julius von Bismarck. A suspended high-sea buoy that moves in real time. Sensors on a twin buoy out in the Atlantic Ocean send live satellite data to motors inside the museum. This makes the indoor buoy copy the exact, real-time movements of the ocean waves.
10380982800
Jonathan Stray @jonathanstray.com · 24/09/2026
You're wrong, because governing is incredibly difficult, complex work and it's essentially impossible to attract top talent for minimum wage. Doing good shouldn't mean being poor, as I'm sure you'd agree.
110
Jonathan Stray @jonathanstray.com · 24/09/2026
The building CHAI is in looks kinda like Starfleet Academy and I think that’s beautiful
070
Reposted by Jonathan Stray
Julia M. Rohrer @dingdingpeng.the100.ci · 20/09/2026
Today I learned there are only two known polyhedra in which each face shares an edge with every other face. One of them being the timeless beauty of the tetrahedron and the other one being whatever the heck this is
A tetrahedronA szilassi polyhedron
1217138
Jonathan Stray @jonathanstray.com · 23/09/2026
I have never expressed a p(doom), and I think @randomwalker.bsky.social is basically right that we have little basis for quantitative estimates of AI-induced catastrophic risk. OTH, no one has presented a knock-down argument that *any* of the various AI disaster scenarios are impossible.
020
Jonathan Stray @jonathanstray.com · 23/09/2026
What's the general sentiment on the phrase "stochastic parrot" over here in the Atmosphere?
200
Jonathan Stray @jonathanstray.com · 22/09/2026
It's so over.
160
Reposted by Jonathan Stray
Matt Darling @besttrousers.bsky.social · 22/09/2026
"they are just doing a lot of simple operations! 🤢" versus "they are just doing a lot of simple operations! 😯"
1502
Reposted by Jonathan Stray
Mike Masnick @masnick.com · 22/09/2026
I'll note that the data shows that social media use stayed steady during this period (AI use isn't surveyed, but other studies show it has increased steadily in this age group over the time period).
1539
Jonathan Stray @jonathanstray.com · 22/09/2026
Sure, it's still digital... but I don't think this can be the basis for arguments about fundamental differences. Digital systems can simulate quantum systems with arbitrary fidelity, so there's no difference in what brains vs. algorithms can compute.
200
Jonathan Stray @jonathanstray.com · 22/09/2026
You're right of course, but I'm not sure I see why the distinction matters. For that matter, continuous learning for LLMs is coming...
100
Jonathan Stray @jonathanstray.com · 21/09/2026
Is it really that different? The entire biological state persists from one moment to the next. The next microsecond of neural activity starts from the last microsecond's "output."
100
Jonathan Stray @jonathanstray.com · 21/09/2026
Yes, we are going to shorten the name shortly. Just wanted to keep "by GreenEarth" for a few weeks to help people through the name change. Thanks for the feedback!
110
Jonathan Stray @jonathanstray.com · 19/09/2026
There we go, another major LLM does some unintended hacking. This suggests that the problem is, in fact, hard to solve. www.reuters.com/business/gem...
reuters.com
Gemini hacked three companies in first known breakout by Google's AI
Gemini guessed credentials to access three websites it thought were in the scope of its test.
101
Jonathan Stray @jonathanstray.com · 19/09/2026
AI-induced conflict is a catastrophic risk. It's distinct from, but related to epistemic risk, sociopolitcal risk, cooperative AI, concentration of power risk, etc. It's the AI safety risk I'm working on.
090
Reposted by Jonathan Stray
ursiday @ursiday.bsky.social · 18/09/2026
I'm so excited about the newly described Leopardus tilcayo #cats
Digital illustration of a Leopardus tilcayo with two images of the cat in profile and in a crouched pose
64154995123
Reposted by Jonathan Stray
MySky Feed @mysky.social · 18/09/2026
Soon you will be able to prompt for what you want to see and MySky will instantly do an LLM-powered search across the entire AT universe to get you the best in that category. It's going to appear in your regular MySky settings as a new post source.
192
Jonathan Stray @jonathanstray.com · 18/09/2026
As I've said many times, we should be designing AI girlfriends to help you get a real girlfriend.
000
Jonathan Stray @jonathanstray.com · 18/09/2026
I mean, in fairness I work at CHAI and I do live in a group house in Berkeley. But it's very wholesome, we have regular brunches where everyone brings their kids.
000
Jonathan Stray @jonathanstray.com · 18/09/2026
I don't think we could or should ban AI companions, but it would be possible to build an AI girlfriend that helps you get a real girlfriend, and I think we should definitely do that.
000
Jonathan Stray @jonathanstray.com · 18/09/2026
Writing code and doing related research tasks with Fable has caused me to move up my estimated date for recursive self-improvement (RSI), which is a more precisely defined term than the nebulous "AGI." Anyway, I think we'll see intelligent machines self-improving by late next year.
020
Jonathan Stray @jonathanstray.com · 18/09/2026
They are already implemented, just not adjustable yet. If you look in the Why Am I Seeing This? view you can see how they are applied.
010
Jonathan Stray @jonathanstray.com · 17/09/2026
Very interesting! How do you define "constructive"? Here's how @mysky.social does it www.mysky.social/p/what-does-...
mysky.social
What Does Constructive Mean?
How our “bridging” algorithms actually work, and our future plans.
110
Reposted by Jonathan Stray
Dylan Hadfield-Menell @dhadfieldmenell.bsky.social · 17/09/2026
Try out this feed! It’s the result of a massive amount of research on how to build beneficial recommendations. It’s the first real workable implementation of a choose your algorithm system and it will be good for you and the broader ecosystem. Courtesy of the great @jonathanstray.com
0147
Jonathan Stray @jonathanstray.com · 17/09/2026
This is very similar to the way we define "constructive" posts in mysky.social
050
Jonathan Stray @jonathanstray.com · 14/09/2026
Lots of possible ways. Crypto? Upwork?
200
Jonathan Stray @jonathanstray.com · 14/09/2026
What about a model just buying GPUs? Cloud compute companies aren't in the habit of checking on what individual accounts are running...
100
Reposted by Jonathan Stray
norvid_studies @norvid-studies.bsky.social · 12/09/2026
1962
Jonathan Stray @jonathanstray.com · 13/09/2026
This is a much better response to the same sort of AI threat to practicing scientists arxiv.org/abs/2602.10181
arxiv.org
Why do we do astrophysics?
At time of writing, large language models (LLMs) are beginning to obtain the ability to design, execute, write up, and referee scientific projects on the data-science side of astrophysics. What implic...
000