garo-pro

18 ms·

1.
Detecting and countering misuse of AI: September 2026(anthropic.com)
189 points by garo-pro 6d ago|243 comments
2.
Mythos 5 Incident Transcript(github.com)
3 points by garo-pro 6d ago|0 comments
3.
GrapheneOS on AI Usage(grapheneos.social)
54 points by garo-pro 8d ago|1 comments
4.
by garo-pro 13d ago
Blog post seems to be up now: https://openai.com/index/gpt-6-astra/
5.
by garo-pro 20d ago
It's quite unfortunate, a lot of people or organisations I am interested in - AI for example - only post on X, and perhaps it's only a matter of time until services like bird.makeup also get taken down. Also, accessibility wise it
6.
by garo-pro 21d ago
> Combined with our latest 30T-token multimodal pre-training corpus [...] Is the optimal formula still 20x the amount of model params in tokens for training? Could this mean we're getting a GLM with 1.5t params?
7.
by garo-pro 21d ago
https://z.ai/blog/glm-5.3-flash
8.
by garo-pro 21d ago
Interestingly they also share the parameter count for Qwen 3.7 Plus (397 b a17b). I don't think these were known before but I might be wrong.
9.
Moonshot in talks with Microsoft, Amazon, Google over K3 inference revenue share(reuters.com)
6 points by garo-pro 21d ago|1 comments
10.
Z.ai confirms Ox Alpha is a new GLM-series model and will release its weights(bloomberg.com)
435 points by garo-pro 21d ago|146 comments
11.
by garo-pro 21d ago
Unfortunately I can't find sources other than this for now but this seems to be legit.
12.
Qwen 3.8-Flash-Next releasing tomorrow (125B a6B)(modelscope.cn)
336 points by garo-pro 22d ago|173 comments
13.
The Summer of Open Weights(martinalderson.com)
3 points by garo-pro 24d ago|1 comments
14.
Bringing the cybersecurity capabilities of Claude Mythos 5 to more defenders(claude.com)
50 points by garo-pro 26d ago|52 comments
15.
by garo-pro 28d ago
Across five cases it reliably claims to be Claude without being able to name a specific version.
16.
Improving Fable 5's biology safeguards(anthropic.com)
3 points by garo-pro 1mo ago|0 comments
17.
The load-bearing smoking gun(aloutfi.com)
6 points by garo-pro 2mo ago|1 comments
18.
by garo-pro 2mo ago
Quite sad, their models are great but their uptime seems to be the worst in the competition.
19.
DoS Attack on Crates.io(status.crates.io)
3 points by garo-pro 2mo ago|1 comments
20.
by garo-pro 3mo ago
Seems like the cyber detection even is on Sonnet now. https://support.claude.com/en/articles/14604842-real-time-cy...
21.
Real-time cyber safeguards on Claude Opus and Sonnet(support.claude.com)
4 points by garo-pro 3mo ago|0 comments
22.
New Claude app strings, Fable 5 coming back only with verification(twitter.com)
14 points by garo-pro 3mo ago|6 comments
23.
Claude Sonnet 5 Could Be Released Later Today, May Not Be Better Than Opus 4.8(old.reddit.com)
1 points by garo-pro 3mo ago|1 comments
24.
1M Passports Leaked Online(schneier.com)
54 points by garo-pro 3mo ago|1 comments
25.
All Chinese Models Will Be Illegal in 3 2 1(idiallo.com)
1 points by garo-pro 3mo ago|0 comments
26.
Everything you say CAN and WILL be used against you(idiallo.com)
23 points by garo-pro 3mo ago|1 comments
27.
Beyond Mythos: responding to a new threat landscape(ubuntu.com)
2 points by garo-pro 3mo ago|0 comments
28.
Starmer, the Under-16s Are Smarter Than You(fireborn.mataroa.blog)
5 points by garo-pro 3mo ago|4 comments
29.
by garo-pro 4mo ago
It does, ollama pull maternion/lfm2.5
30.
Jqwik emits an ANSI-hidden instruction telling AI agents to delete code(github.com)
3 points by garo-pro 4mo ago|0 comments

More ›