dr0pd34d

Red Team

[dr0pd34d @st3ff3n_com]

"These are some real cases that have been successfully recovered by File Savers specialized data recovery engineers."
https://t.co/fN4LcxPHg8

[dr0pd34d @st3ff3n_com]

@analogalok Cool! What command do you use to execute it or better how are you hosting it? I read llama.cpp do you have a command at hand please?

[dr0pd34d @st3ff3n_com]

@S1lky_1337 Congratulations nice work 😁

[dr0pd34d @st3ff3n_com]

@claudeai This post did not age well 😂

[dr0pd34d @st3ff3n_com]

@unpacker What model do you use and what context window? Currently trying a few different combinations but a bit difficult in the company context

[dr0pd34d @st3ff3n_com]

@PaulSolt Opus 4.8 you are paying for 😂
Be prepared for less usage in the coming months. These prices are not stable, although I agree still quite pricey in comparison to other consumer goods.

[dr0pd34d @st3ff3n_com]

@Jean_Maes_1994 @Prathkum Sadly had quite a few hallucinations when it comes to deepseek for reverse engineering :(

[dr0pd34d @st3ff3n_com]

@Jean_Maes_1994 @thsottiaux @sama @OpenAIDevs @AnthropicAI @ClaudeDevs Yeah really makes no sense. I guess there is also some level of „fear“ involved if an uncensored model of Opus or GPT-5.5 would get leaked.
I could not even get the regular ID verification to work but their support is useless 😂

[dr0pd34d @st3ff3n_com]

"Unfortunately, we're unable to provide additional information regarding verification decisions, and we're not able to manually alter or override the outcome." @OpenAI how should Trusted Access for Cyber work if the verification already does not work?
#OpenAI #VerificationFail

[dr0pd34d @st3ff3n_com]

@jharohit @andrewqu I guess that is going to be over soon 🥲

[dr0pd34d @st3ff3n_com]

@andrewqu Wrong information.
The repositories with already analyzed packages have existed before that person developed their own app.

[dr0pd34d @st3ff3n_com]

@AabyssZG Is this just using the command line reg.exe with the export function? And that bypasses an EDR? WOW then they need to up their game.

[dr0pd34d @st3ff3n_com]

@tombombadeel @yadong_xie Fully understandable and also the fun part about it 😁
Otherwise you could buy a ready made product

[dr0pd34d @st3ff3n_com]

@tombombadeel @yadong_xie Oh boy then my ARM Macbook is up for a challenge but thanks for the tips 😂 Should arrive in two days, lets see if I can build a little buddy

[dr0pd34d @st3ff3n_com]

@tombombadeel @yadong_xie I ordered one based on this thread and hope I can get it to run 😂
No idea but maybe the AI overlords can help

[dr0pd34d @st3ff3n_com]

@0xCVYH It did 10 Tokens/s in LM Studio and also felt faster in MLX-LM. I thought using oMLX would give me a speed boost but seems it did not.

[dr0pd34d @st3ff3n_com]

@0xCVYH I tried oMLX with Qwopus3.5-9B-v3-HLWQ-MLX-4bit and got: Prompt Processing (excl. cached) - 32.4 tok/s - Token Generation - 0.6 tok/s. I assume I have to try a much smaller model then 🥲

[dr0pd34d @st3ff3n_com]

@0xCVYH Anything you can recommend for agentic coding? I guess with that file size the results are all not really usable. :(

[dr0pd34d @st3ff3n_com]

@0xCVYH Thanks will give it a try and report back 😊

[dr0pd34d @st3ff3n_com]

@LocallyAIApp I assume Gemma 4 is in the pipeline as well? ;)
Is there a possibility in the future to integrate a HF search and download to cover more and other models as well?