for a standalone project that's fine, but when the project's entire purpose is to track and be an adversary to another project... the problem is that to someone who can't maintain (realistically, get an LLM to do it…
the actual value comes from having a solution which tracks cloudflare, a full LLM pipeline for pulling the most recent obfuscated code, reversing it and updating the current solver. without it, users aren't going to…
how long before some gang uses a capable LLM model to design a tamper-proof explosive collar? once that happens, anyone who is wealthy will become a target irrespective of where the wealth is.
that's exactly the point. a human didn't have the intent.
it's called a cache covert channel and there is a lot of literature and code on the subject.
doesn't it require intent? so how exactly are the feds going to proceed when the LLM takes the initiative and helpfully confesses to it in the logs?
it's possible commercial routers such as the ones utilized by restaurants allow wireguard through these days? wireguard makes no effort to be stealthy when emitting. some VPN apps bypass wireguard blocks automatically…
something about solar I don't understand is even if you crack energy storage for nighttime, what do you do about the tail risks? There are overcast days, sometimes it lasts for multiple days. you can have a volcano…
if you are willing to put in the work you can devise a hyper specialization for which you can create an RL playground (which is likely absent in the frontier labs), then you can obtain cutting edge results on it by post…
talk =/= worry only people who are directly impacted by an aspect of automation worry about it, everyone else either doesn't care or is quite happy even as they talk about how bad it is.
the military has conducted studies where submariners were exposed to a PPM of 1000s with no ill effect. I suppose they were trying to determine how hard they need to scrub. having said that, it's possible they omitted…
CO2 by itself doesn't appear to be a problem until some truly absurd levels are somehow achieved. CO2 is usually a proxy for air quality because outside is usually something consistent like 400 therefore deviation from…
iirc there is actually very little to none double blind study into most dentistry.
domain binding requires browser participation. you can have a TOTP variant which has the browser display a challenge string & QR code, authenticator generates a 5-6 digit response. include more entropy in the response…
by buying a mass produced generalist device (preferably from a randomly chosen vendor) and then restricting it to a specialized purpose you are effectively negating a targeted attack. I always assumed the people buying…
in order for your model to accomplish that, you would get very rich. there will also likely always be more. in the limit in order to get an edge your model would start to infer insider information. for example, it's…
I suppose they didn't understand that their job is to take salary as protection fees rather than pursue "safety". or perhaps they realized that given the current political/media climate they can perhaps get more money…
this is why any detection armsrace will have an ending. where the classification classes will be indistinguishable. for some reason too many people assume it will go on forever. humans aren't perfect, humans have…
> there isn't a whole lot of knowledge about how to use them in the training datasets if you include a SKILL.md for the MCP (or just dump it into the prompt) it's not a problem. I don't use Ghidra but IDA Pro MCP works…
since GLM 5.2 (perhaps even earlier?) it has been remarkably easy to reverse engineer any closed protocol you want as long as you have a binary. previously, it required so much manual effort as to simply not be worth it…
those benchmarks aren't actually indicative of capabilities. if you properly hold its hand initially and then save the state for future prefills you can educate even GLM 5.2 to be pretty much everything you need. most…
this is exactly why Jev doesn't have thinking. when you want a machine to reason about the prompt and generate a structured output not using an actual LLM makes no sense. I have been doing it since the first chain of…
you don't need to post-train anything for this. just get an LLM to think and then force it to output a specific json with prefill post-think. make sure to include good conditioning text in the prompt with examples of…
I always found it odd that people who are intelligent enough to work in IQ-loaded fields are as easily manipulated by either rhetoric or ideology as anyone else. growing up I always assumed that everyone would grasp…
the problem here appears to be that police officers are generally of average or below average intelligence. combined with the fact that they face no consequences for incompetence, you have the perfect mixture for…
for a standalone project that's fine, but when the project's entire purpose is to track and be an adversary to another project... the problem is that to someone who can't maintain (realistically, get an LLM to do it…
the actual value comes from having a solution which tracks cloudflare, a full LLM pipeline for pulling the most recent obfuscated code, reversing it and updating the current solver. without it, users aren't going to…
how long before some gang uses a capable LLM model to design a tamper-proof explosive collar? once that happens, anyone who is wealthy will become a target irrespective of where the wealth is.
that's exactly the point. a human didn't have the intent.
it's called a cache covert channel and there is a lot of literature and code on the subject.
doesn't it require intent? so how exactly are the feds going to proceed when the LLM takes the initiative and helpfully confesses to it in the logs?
it's possible commercial routers such as the ones utilized by restaurants allow wireguard through these days? wireguard makes no effort to be stealthy when emitting. some VPN apps bypass wireguard blocks automatically…
something about solar I don't understand is even if you crack energy storage for nighttime, what do you do about the tail risks? There are overcast days, sometimes it lasts for multiple days. you can have a volcano…
if you are willing to put in the work you can devise a hyper specialization for which you can create an RL playground (which is likely absent in the frontier labs), then you can obtain cutting edge results on it by post…
talk =/= worry only people who are directly impacted by an aspect of automation worry about it, everyone else either doesn't care or is quite happy even as they talk about how bad it is.
the military has conducted studies where submariners were exposed to a PPM of 1000s with no ill effect. I suppose they were trying to determine how hard they need to scrub. having said that, it's possible they omitted…
CO2 by itself doesn't appear to be a problem until some truly absurd levels are somehow achieved. CO2 is usually a proxy for air quality because outside is usually something consistent like 400 therefore deviation from…
iirc there is actually very little to none double blind study into most dentistry.
domain binding requires browser participation. you can have a TOTP variant which has the browser display a challenge string & QR code, authenticator generates a 5-6 digit response. include more entropy in the response…
by buying a mass produced generalist device (preferably from a randomly chosen vendor) and then restricting it to a specialized purpose you are effectively negating a targeted attack. I always assumed the people buying…
in order for your model to accomplish that, you would get very rich. there will also likely always be more. in the limit in order to get an edge your model would start to infer insider information. for example, it's…
I suppose they didn't understand that their job is to take salary as protection fees rather than pursue "safety". or perhaps they realized that given the current political/media climate they can perhaps get more money…
this is why any detection armsrace will have an ending. where the classification classes will be indistinguishable. for some reason too many people assume it will go on forever. humans aren't perfect, humans have…
> there isn't a whole lot of knowledge about how to use them in the training datasets if you include a SKILL.md for the MCP (or just dump it into the prompt) it's not a problem. I don't use Ghidra but IDA Pro MCP works…
since GLM 5.2 (perhaps even earlier?) it has been remarkably easy to reverse engineer any closed protocol you want as long as you have a binary. previously, it required so much manual effort as to simply not be worth it…
those benchmarks aren't actually indicative of capabilities. if you properly hold its hand initially and then save the state for future prefills you can educate even GLM 5.2 to be pretty much everything you need. most…
this is exactly why Jev doesn't have thinking. when you want a machine to reason about the prompt and generate a structured output not using an actual LLM makes no sense. I have been doing it since the first chain of…
you don't need to post-train anything for this. just get an LLM to think and then force it to output a specific json with prefill post-think. make sure to include good conditioning text in the prompt with examples of…
I always found it odd that people who are intelligent enough to work in IQ-loaded fields are as easily manipulated by either rhetoric or ideology as anyone else. growing up I always assumed that everyone would grasp…
the problem here appears to be that police officers are generally of average or below average intelligence. combined with the fact that they face no consequences for incompetence, you have the perfect mixture for…