Entro-P

The right of the people to keep and bear cyberweapons shall not be infringed.

[Entro-P @int0x80h]

Need is the wrong framing, a model like Opus 5.5 lets you prompt much more ambitiously and Ambiguously. I love deep seek, at this point I do maybe 10 billion tokens worth of inference on it a month, but it just doesn't understand user intent or follow instructions to nearly the same degree that Opus 5.5 does. So even for tasks where it isn't strictly required, Frontier models still save you a lot of time and attention. I think a world where openweight models get left behind by the closed weight Frontier is a terrible one, Sam and Dario should not be at the helm of humanity, but denying reality doesn't help us here. There's a reason why total token spend on these two labs eclipses open weight usage by at least one order of magnitude.

[Entro-P @int0x80h]

@AirMan1979 @RepThomasMassie @NorthTactics @elonmusk @cradrrat This is satire right?

[Entro-P @int0x80h]

@PablitoR14 @RepThomasMassie @elonmusk @cradrrat Youre just describing Bend

[Entro-P @int0x80h]

@AzFlin Data annotation for AI training pays 15-40/hr usually

[Entro-P @int0x80h]

@mdp_sec 4 months

[Entro-P @int0x80h]

@secureideas @PortSwigger seems like more of a case of mallace than malice.

[Entro-P @int0x80h]

@chrisalbon cve coming soon^tm

[Entro-P @int0x80h]

@0xConda What they dont know wont hurt them

[Entro-P @int0x80h]

@Alisoapi @cshwshd99

[Entro-P @int0x80h]

@samb_tech @rez0__ that is the least crowded part of bug bounty lol. Operating a soldering iron and a flasher is actually a pretty decent moat to defend against the hoardes of hackbot operators.

[Entro-P @int0x80h]

@notjazii @theo Performance != UX

[Entro-P @int0x80h]

@merlindru @theo Is it really unfair to judge a model for that when you're required to use it inside of that harness?

[Entro-P @int0x80h]

@mdp_sec DS 1017 and GLM 5.3 are the ones to test.

[Entro-P @int0x80h]

This is a totally legitimate concern, the transition period is going to be the tricky part, and I can relate to that since I'm in a field where my work will most likely be fully automated well before most other fields. At the same time I don't think the incentives can ever align to where capable AI is both widely available and optional to use in a professional setting. The risk of not making it widely available means that those gains go to an even fewer number of people, and not making it available at all means we're not going to see that post-scarcity future either.

[Entro-P @int0x80h]

@log4j @zebassembly Prod doesn't go down on a friday evening because the product art was ugly and nobody in corporate america likes dealing with creatives(that's not a diss btw, you guys are cooler than them). Tech people can also be insufferable, but they're a lot easier to shove into a basement.

[Entro-P @int0x80h]

@Nerdsauce1 @luckeyfaraday Pump the breaks guys, we need to stop trying to create a post-scarcity world because Nerdsauce and 1000 other people actually enjoy doing their jobs by hand. Surely you realize that you are in the extreme minority here right?

[Entro-P @int0x80h]

@dd_pwnd Yeah when I was working in cyberinsurance(during the height of ransomware popularity) two biggest things we wanted to see were immutable backups and some indication that you were actually testing your ability to restore from them.

[Entro-P @int0x80h]

@HackingDave Good.

[Entro-P @int0x80h]

@GodfatherOrwa So in other words they used you as free labor to validate their tool output.

[Entro-P @int0x80h]

@vxunderground @malwrhunterteam @smica83 Malware author read about LLMs refusing to touch adult content and threw in a random bangbus string lmao