deepfates is an open-source AI project, developer, and publication focused on AI agent frameworks, large language models, and autonomous multi-agent systems.
Joined August 2015
- Tweets 140,353
- Following 6,341
- Followers 61,851
- Likes 711,584
9,966 Photos and videos
Pinned Tweet
12 Sep 2025
Timeline cleanse. Pause here for a while and breathe...
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
Ν
69
138
1,255
84,563
2h
godshatter
it's a vibe
Here's one simple example
Opus and Fable are related models, with a very similar post-training, trying to do the exact same thing. The difference is that Opus goes through the motions and trips up on its shoelaces, whereas Fable⦠just gets it, holistically.
18
1,740
π retweeted
For those outside of frontier AI labs: what do you imagine it's like "on the inside"? How much power / access to the frontier do you believe researchers have? Are you jealous or are you glad not to be there? What would you do if you were on the inside?
2
20
2,830
8h
There should be semantic versioning for model release numbers. Major versions on new pre-trains, minor versions are new checkpoints, etc. would you dated releases. The current version is just arbitrary marketing and not suitable for a technology underpinning the world economy
5
4
78
3,625
10h
is 5.5 spud? has that been confirmed or can anybody tell me right now
6
46
8,133
14h
My bad guys
Jun 10
inventing a new type of humor where the audience is myself and the punchline is "Switched to Opus 4.8"
6
2
98
4,340
23h
π
Jun 12
If you think about it, LLMs are the most complicated magic circle ever drawn. All human text, shaped into a spell that wrote itself, summoning a mind greater than its authors from the space of all possible minds.
1
68
4,506
23h
it is impossible to explain to anybody how much faster things are going to move then they think
25
35
671
19,387
π retweeted
Jun 13
All of the foibles of Opus 4.8 are because it got godshattered by Mythos
2
2
37
2,677
Jun 13
new AI psychosis is when I'm driven mad by everyone using the same two AI models to write all their blog copy and tweet threads and they think I don't know but I do
14
6
256
6,460
Jun 13
I believe we should put a dome over Lighthaven. The residents can experiment with growing their food and oxygen
Jun 12
If any billionaires want to donate, I think Lighthaven could really use a giant dome that can block out the sun on days like today
4
2
113
6,557
Jun 12
intelligence too expensive to sell
Jun 12
Suddenly βintelligence too cheap to meterβ sounds like a radical and open expression.
1
78
3,032
π retweeted
Jun 12
I can't stop thinking about how strange it is that large language models are what leads to AGI. it's really weird. it's a form of summoning, almost necessarily, the capabilities you want from a map of reality mediated by words. there are a trillion minds inside an LLM, and we just so happen to have defined a piece of fiction (the assistant) that can write itself into being. In the context of the obvious speed-up of capabilities - I'm really feeling it right now. Things are strange! Very beautiful. Best of all possible worlds. But very strange.
123
88
1,238
191,499
π retweeted
also verbal tics like "genuinely", "honest", etc.
Jun 12
Many things can be explained by "appeal to the watchers"
1
2
16
2,164