> It has been over a decade since any ordinary off-the-shelf closed-source software was meaningfully obscured from serious adversaries.
Probably goes without saying but the last line of defense is not deploying your software publicly and instead relying on server-client architectures to do anything. Maybe this will be more common as vulnerabilities are more easily detected and exploited. Of course its not always feasible.
It has been annoying seeing my (proguard obfuscated) game client binaries decompiled and published on github many times over the last 11 years. Only the undeployed server code has remained private.
Interestingly I didn't have a problem with adversaries reverse engineering my network protocols until I was updating them less frequently than weekly. LLM assisted adversaries could probably keep up with that now too.
> You would not start with "The car is functional [...]"
Nope, and a human might not respond with "drive". They would want to know why you are asking the question in the first place, since the question implies something hasn't been specified or that you have some motivation beyond a legitimate answer to your question (in this case, it was tricking an LLM).
Why the LLM doesn't respond "drive..?" I can't say for sure, but maybe it's been trained to be polite.
> You said it correctly with "would've been safer to start building a distance buffer", that is the proxy the insurance companies want to use for risk assessment.
Then use it? Mandate reaction speed tests or other driving mechanics competency evaluation (not road sign comprehension) and watch insurance margins explode.
The driver in my example did poorly and scored top marks in the heuristic.
The landing page video's first incident is a car coming from behind and from the right, cutting off the filming car. The filming car didn't react at all when instant (but measured) braking would've been safer to start building a distance buffer.
One thing HPDE taught me is that most people under brake in dangerous situations because they simply don't know the limit of their vehicle nor the sensitivity range of the brake pedal.
The hard braking heuristic makes sense when estimating risk of road segments, but not as a proxy for driver competence.
> I reboot, log into Epic and GOG, and start downloading The Outer Worlds, a game from 2019 I’ve been playing a bit lately. It runs fine with Proton, and I can even sync my saves from the cloud. I play it for a few minutes with my trackball, remember I hate gaming on a trackball, and plug my gaming mouse back in. It works fine as long as I’m in the game, but outside the game, mouse clicks stop working again. It makes sense — the bug is on the desktop, not in games — but it’s very funny to have a gaming mouse that only works for gaming.
What is it with mice and OSes?
Windows is the only OS I can seem to configure to get low latency, high accuracy, linear movement with, and it's not for lack of effort.
I struggled for several years to do SWE work on a Mac and no 3rd party program could get it working the way it does on Windows. I tried Linear Mouse and many others. I eventually gave up, went against the prevailing (90%) culture where I work, and exchanged my mac for a windows laptop. I haven't measured it, but I feel more productive simply because I can click what I want to click marginally faster.
Is something in Mac drivers performing non-linear mapping? Why?
Based on the quote above it seems like Linux hasn't even gotten up to par with Mac for mice.
The best litmus test for an OS for me is whether I could play an RTS or FPS competitively with it, even though I haven't played either for years.
I see the reasoning for accountants keeping future liabilities off of the balance sheet. I do this myself in multiple contexts.
Still, when making decisions about whether to take out or grant a loan (personal or business) I need to consider future "value" and cash flows. To someone running a business this is probably more important than the balance sheet. So I think the interest recording criticism is valid but relatively minor in the context of the whole article.
> One way would be to use a polycentric form of law where each individual could determine what or which form of law to live under that protects them by voluntarily entering some sort of protective group. Or choose none at all and merely protect their natural rights on their own.
I don't have much in the way of critique or judgement to offer on this political philosophy, just an observation: it sounds tribal or even pre-civilization. Out of curiosity I asked an LLM what present day countries most closely implement it. It came back with Somalia and a label: anarcho-libertarianism, with the caveat that it isn't an exact match. Historical examples were also interesting. I'm curious whether you think that's a good example or not.
If the world had more unsettled land I think your ideal would be a lot easier to implement. The U.S. was borne out of people fed up with their current situation (legal or otherwise) deciding to start something new. The fact that it's made up of 50 states, each with their own set of laws and relatively high internal mobility, suggests that its already a mild compromise away from pure democracy and toward your ideal.
To me the purest form of your ideal seems unstable, especially in the face of power imbalances and conflicting choices, and I suspect it would inevitably evolve into something else. As far as I can tell history supports that view.
> Neither was slavery. Was that OK too? And to clarify (though it’s worrying this point needs to be made), I mean morally.
It may well have been morally OK to most people (see: moral relativism), and since you're implying it wouldn't have been OK to you, it's worth pointing out that you probably wouldn't have done anything about it in the relevant time periods.
If you're an American you don't even need to try that hard to make moral relativism visceral: was the displacement (and far worse) of Native American tribes "OK"? I'd say no, but it isn't morally urgent enough to me or the 99%+ of Americans who are unwilling to pack their bags and return the entirety of two continents to the native descendants.
My question was "What's the value of a secret benchmark to anyone but the secret holder?"
The root of this whole discussion was a post about how Gemini 3 outperformed other models on some presumably informal question benchmark (a"vibe test"?). When asked for the benchmark, the response from the op and and someone else was that secrecy was needed to protect the benchmark from contamination. I'm skeptical of the need in the op's cases and I'm skeptical of the effectiveness of the secrecy in general. In a case where secrecy has actual value, why even discuss the benchmark publicly at all?
I learned in another thread there is some work being done to avoid contamination of training data during evaluation of remote models using trusted execution environments (https://arxiv.org/pdf/2403.00393). It requires participation of the model owner.
I see the potential value of private evaluations. They aren't scientific but you can certainly beat a "vibe test".
I don't understand the value of a public post discussing their results beyond maybe entertainment. We have to trust you implicitly and have no way to validate your claims.
> There is no "winning" at benchmarks, it's simply that it is a better and more repeatable evaluation than the old "vibe test" that people did in 2024.
Then you must not be working in an environment where a better benchmark yields a competitive advantage.
It's hard to have any certainty around concealment unless you are only testing local LLMs. As a matter of principle I assume the input and output of any query I run in a remote LLM is permanently public information (same with search queries).
Will someone (or some system) see my query and think "we ought to improve this"? I have no idea since I don't work on these systems. In some instances involving random sampling... probably yes!
This is the second reason I find the idea of publicly discussing secret benchmarks silly.
No eye tracking in this dystopia I guess.