NL

Detection Engineer | Exploring AI in defensive security (⁠^⁠^⁠)

[NL @NLx64]

@SwtCicero @buaaxhm Daybreak?

[NL @NLx64]

@mdp_sec "I have authorization" loop

[NL @NLx64]

@mdp_sec @CommandCodeAI Mimo flash looks interesting >.>

[NL @NLx64]

@0xSero upgrading from the 3090 to the 4090 worth it? looking these numbers idk

[NL @NLx64]

@yoavalon @S1r1u5_ @rootxharsh @iamnoooob Very interested in knowing how to deal with the hallucination rate of these models, it's really wild.

[NL @NLx64]

@TihanyiNorbert It looks very good, let's get to the tests.

[NL @NLx64]

@_jensec in my personal tests 5.6 sol it's like 25-20% better and less refusals too.

[NL @NLx64]

@mdp_sec Yeees

[NL @NLx64]

@louszbd i have reported some vulnerabilities of github version, double check it when you can idk if is the right way to report. (security advisories)

[NL @NLx64]

@eyalsela @polygonben Niceee

[NL @NLx64]

@mdp_sec which harness do you have used to "bypass" the refusals?

[NL @NLx64]

@h4x0r_dz If you say so, I believe it.

[NL @NLx64]

@TheAhmadOsman @vishalsingh2972 I'll give it a chance.

[NL @NLx64]

@NorthGuard69 @realmt99 @Sentdex the lion?

[NL @NLx64]

@allinasecond @TheAhmadOsman Try open code go, it's 10 bucks and you can try the GLM and Deepseek and other Chinese models.

[NL @NLx64]

@Sentdex How many tk/s are you getting on your setup?

[NL @NLx64]

@Highlylrregular @PawelHuryn cache hit price no?

Cache hit

GLM 5.3 flash = US$ 0,03

US$ 0,007 DeepSeek v4 flash, 4,3× more cheaper

[NL @NLx64]

@LxlxIxlxlxL Gold in the form of a post

[NL @NLx64]

@psiloycne @Kostastsale hahahaha perfect

[NL @NLx64]

@psiloycne @Kostastsale Temporal correlation rules ;/