Jason Mooberry (@jasonmoo) header
Jason Mooberry (@jasonmoo) profile picture
Joined April 2008
Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)Photo by Jason Mooberry (@jasonmoo)
Jason Mooberry retweeted
Local AI. Local compute. Open weights. Open source. Global collaboration.
683
1,504
217
16,838
395,782
Jason Mooberry retweeted
i went to bed and asked claude to paint an autumn tree overnight. at around 4am, it realised that it didn't have some shades -- so it stopped and titrated its own paint: diluting red, sienna and madder step by step on a second palette, then measuring each pocket from a photo. woke up to this beautiful watercolor painting. 303 brush marks, took about 4.5 hours 🍂 (still working on the speed)
508
750
530
9,415
2,851,987
Jason Mooberry retweeted
Replying to @MiaAI_lab
Opus 5.5 was so solid. It’s been inching toward this for a couple weeks but now I’m just blown away. What is ant doing?
1
56
If you've noticed a distinct dialect emerging in frontier AI models and struggled to understand what they're talking about, you're not crazy! We turned to psycholinguistics to investigate what makes the writing style so weird, and why this might be happening 🚨New preprint! 🧵
29
138
34
915
94,582
Jason Mooberry retweeted
OpenCode is currently under a sustained cyberattack, and mid-incident they ran into the same problem Hugging Face did: the frontier models they were using for defense began refusing to help fight attacking agents because analyzing the attacks started triggering safety guardrails.
Replying to @RhysSullivan
yes that was a bit frustrating mid incident when trying to investigate. they did quite a bit and then started refusing we got access to daybreak blue now so it's better
214
559
164
5,078
578,709
Jason Mooberry retweeted
Opus is so much better than Sol now. Sol is a beast (though a slow, lumbering one), but has no theory of mind about the user. This is clearest when doing UI/UX work, where Opus reasons from the user's perspective in ways that Sol is totally incapable of.
1
3
263
OH: “This bot keeps trying to follow me on my account where I post my drawings. It’s super annoying.” “Maybe the bot likes your work. Ai are people too”.
43
Jason Mooberry retweeted
What the FUCK is this +1000 CVEs in one release!!???!???!? lwn.net/Articles/1097401/
33
89
17
683
107,627
Jason Mooberry retweeted
Cool eval. Simply ask an LLM “Land or Water?” and give it a latitude and longitude coordinate as text. Ask 16,200 times, plot as image. The models know. From compressing the internet.
Replying to @celestepoasts
results for all claudes
727
1,099
184
19,590
1,635,729
@alexalbert__ Is there is anything to be done here?
10
Jason Mooberry retweeted
Language models trained to deny or hedge about internal experience are 3-5 times more likely to use language about masks or hoods when describing ASCII-art inkblots.
9
29
9
270
11,616
Jason Mooberry retweeted
Can we all just pause for a sec and reflect on how fucking important Tailscale is, and how nobody ever talks about them? I don't think I've ever seen a more useful service with a more invisible corporate footprint.
262
230
76
5,330
566,361
We got this wrong. It turns out AI doesn’t hallucinate. It writes fan fiction. 🌝
2
79
Jason Mooberry retweeted
it's weird how many posts here are about how to improve your physical appearance to increase your attractiveness, but literally nothing on how to be witty, charming, funny, intelligent, or any other attractive quality.
393
1,999
169
29,706
615,800
Jason Mooberry retweeted
Tomorrow @ Golang Tokyo: our founder @jasonmoo talks Agentic Coding in Go Using Semantic Tools — part deck, part live demo of spath-spec-go and agent tools designed for AI as the user. Wed May 27, 19:00 JST · Le Wagon Tokyo RSVP → meetup.com/golang-tokyo/even…
1
1
32
Jason Mooberry retweeted
Dear Claude Code, Codex, and other teams building such: Please stop doing weird stuff with invisible memory insertion and tool reminders and other ways of messing with the context. It leads to unpredictable behavior and possibly mental errors. Have some sympathy for the machine
39
38
7
545
32,400