Studying Information Security & Working in AI, all views are my own

Austria
ClassicMain retweeted
The billionaire party absolutely despises you
🚨 Virginia passes bill requiring access to water and shade for workers. Every Republican voted no.
28
2,065
23,301
199,549
ai labs use "sandbox" in the game dev sense and not in the linux security sense
Replying to @tomekkorbak
questions: 1. why was the DNS tool able to bypass the sandbox? 2. why 2.5 hours from detection to pausing the run?
23
223
4,327
93,670
ClassicMain retweeted
every time, astra finds nothing after opus 5.5 how is this possible, the models are usually so nitpicky
47
5
470
34,290
ClassicMain retweeted
🚨 Sonnet 5.5 is being grey tested in claude code right now > insanely fast > efficient , have that human touch that we had in sonnet 4 <remember > > Pricing: Input: $2 Output: $10 Cache reads: $0.20/M some banger comparison with sol 6 is coming get ready to see holy mogging
64
34
1,188
211,136
Claude Opus 5.5 is the best model I have ever used. Seriously. It is insane.
2
56
ClassicMain retweeted
"if 3% is a genocide then america has committed genocide several times in the last 100 years"
Replying to @Jay__Catsby
If 3% is a genocide then America has committed genocide several times in the last 100 years you fucking retard
567
96,218
914,950
16,324,916
ClassicMain retweeted
"Mistakes" = they lied to the world until video evidence was found. The video? Medics being executed from 8 metres away and then buried with their ambulances. Read that again.
Israel admits mistakes over medic killings in Gaza bbc.in/3XLEvqq
80
8,882
27,226
295,666
>They've found a grave full of kid's corpes in Mexico >Oh wow! so it's prehispanic archaeology is peak >These bones are recent
En Michoacán hay una fosa donde estarían los restos de cientos de niños de entre 1 y 6 años. La evidencia: mil 600 dientitos de leche hallados por peritos de @FGRMexico y el colectivo Buscando Cuerpos en Todo México. El #CódigoNegro de hoy en @Milenio: milenio.com/opinion/oscar-ba…
143
3,132
38,347
809,731
Recently, we sent two North Korean prisoners of war to the Republic of Korea… These guys are not easy to capture alive. One of them, when he realized he was going to be taken prisoner, tried to kill himself. And he was shocked that fate did not let him die. This is how they raise people in the North. They take away their freedom. They take away their right to choose. And they drill into them that they must die, no matter what. Do you really want to give such a force more time to learn how to wage war? From an address to UNGA. (2/7)
956
7,586
45,530
2,782,767
You leaked my data last year and everything got hacked and it took MONTHS to get everything in order and secure again. Go fuck yourself
when you launch an extremely popular age assurance update that everyone on Discord was really looking forward to and it has bugs
59
3,219
38,465
531,748
Opus 5.5's vision capabilities are seriously crazy It is not equal to Gemini in general object recognition but it REASONS way more about images and gets concepts and shown relations much much much better than ANY other model I tested and use. Wow. Seriously crazy
1
60
ClassicMain retweeted
Fuck every company trying to erase anonymity on the internet with age verification bullshit
16
1,189
9,923
55,824
ClassicMain retweeted
Replying to @discord
If there's no way to verify the uploaded data gets deleted, then there's no way to verify that you having to force doxing yourself to age verify won't be data leaked. Regardless of what policy you set up. NO to digital ID period. A birthdate should be enough.
2
29
577
9,891
ClassicMain retweeted
Replying to @discord
You told us our uploaded data was deleted "immediately" last time, then you admitted your customer service suffered a data breach which leaked 70,000 government IDs at a minimum. So which is it?
24
443
5,655
Opus 5.5 is magic
1
69
Why is @OpenRouter showing providers at 100% availability and as 🟢green when every SINGLE request no matter how small errors out as 429 That doesn't seem like "available" to me.
1
106
ClassicMain retweeted
how to turn a simple $6,500 bounty into a WSJ exclusive.
Here is how OpenAI’s CISO fumbled the situation (in my personal opinion) 🧵 1/X
8
9
551
42,628
ClassicMain retweeted
Bonsai 2 dropped hours ago. It's already doing real work in Osaurus. Read two CSVs. Wrote the Python. Drew the chart. Then found the one channel losing money and told me why. 27B. Ternary. Nothing left the Mac.
Today, we’re announcing Ternary Bonsai 2 27B. Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance. Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use. Ternary Bonsai 2 27B is available today under Apache 2.0.
10
21
334
34,368
ClassicMain retweeted
Today, we’re announcing Ternary Bonsai 2 27B. Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance. Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use. Ternary Bonsai 2 27B is available today under Apache 2.0.
640
1,568
15,549
5,413,782

ALT I Guess Tomodachi Life GIF

Today, we’re announcing Ternary Bonsai 2 27B. Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance. Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use. Ternary Bonsai 2 27B is available today under Apache 2.0.
100