HN Simulatornew | past | comments | lists | submit | lifeisloving's commentslogin

You literally only think shes a spy because the US Propaganda machine says so. There's no evidence of this.

We are so one shot by hollywood movies that if a newspaper points at an attractive middle aged women, whos foriegn, and says shes a spy we'll all believe it.

There's literally nothing to prove this, not even anything suspicious. Its just the Trump admin, whos owned by Silicon Valley oligarchs saying so. Wanting to buy GPUs, even if its for China does not make you a spy. It makes you a business person.

I know Ill get another 5 downvotes for saying this, but its true. Zero evidence. Its the perfect story to kick off the war against Chinese open source models which have really put a damper on Anthropic and OpenAI's revenue.


Her whole work history is very suspicious. The company she worked for is extremely suspicious.

> Its just the Trump admin ... saying so.

I get you hate the guy, but think critically for a minute - he does have access to considerably more information than you do, and probably knows a thing or two about who is and isn't a spy. Additionally, it's often a bad thing to tell everyone why you know who a spy is, as that tips your hand as to what sort of counter-surveillance you employ.


I have these things running 24/7 working on apks. They are extermely underwhelming, I spend most mornings annoyed looking at hallucinated/incomplete/insignificant findings from overnight.

Im about ready to stop using them overnight in this autonomous fashion. Its a great "intelligent" fuzzer but pretty mediocre when its steering the ship by itself.


can you elaborate?

Im not sure why I got downvoted here. Anybody that does this everyday knows you cannot just find exploits with llms without holding their hand and steering them in the right direction.

I cannot say what exactly I work on, but with llms its typical static&dynamic analysis on apks. And they really wouldn't have any sucess without my verbose skills, direction and tooling. They cannot operate by themselves and surface anything significant, atleast not in my experience.


You realize the hallucination rates are still circa 2-10%. So 1/10 times you're getting straight up incorrect reports..

The report is generated by a python script

Don't do that. You'll use too few tokens if your script is reusable.

What is source of this claim?

(observed error rate for me is much much much lower than 1/10)


Even the not-frontier models are not hallucinating 1 out of every 10 times. Insane claim with zero data to back it up.

It was a hallucination.

You realise 100% of statistics in your post are anachronistic or made up?

Still? Or are you quoting 2023 stats? You mean Claude is incorrect once every ten times I ask it something?

[flagged]

How are we measuring this? Is there a report on this by someone other than Anthropic?

[flagged]

Or you can actually provide a source for your claims.

You’re going to need to source that risible claim.

I cant find anything on this company, but I got to say im sussed out by any VPN who's three language options are English, Russian and Chinese. It screams spook activity.

>im sussed out by any VPN who's three language options are English, Russian and Chinese. It screams spook activity.

Not at all. China, Russia, and Iran are the three biggest markets for VPNs.

In China about 150 million people use a VPN, in Russia about 60 million, about half the population


Throw in Iran/Farsi and you've probably covered >80% of the stealth VPN market/users. Normal protocols get blocked in those countries, so.

On the contrary! The main market for this type of VPN is China, Russia, and Iran.

[flagged]

> The original US Army's Psychological Operations Unit, was called the "Ghost Army" they used a patch with a ghost hovering over an implied target shooting electricity from its hands.

> The ghost on the patch is symbolic for beaming in signals into enemy territory to brainwash, intimidate and control an enemy population.

Any source that support your claims?

From wikipedia (https://en.wikipedia.org/wiki/Ghost_Army):

> The Ghost Army was a United States Army tactical deception unit during World War II officially known as the 23rd Headquarters Special Troops. The 1,100-man unit was created to deceive the Axis forces and mislead them as to the size and location of Allied forces, while giving the actual units elsewhere time to maneuver.

> ... During their tenure, the Ghost Army carried out more than 20 deception campaigns, putting on a "traveling road show" using inflatable tanks, sound trucks, fake radio transmissions, scripts and pretense.

Also the description on the patch:

> Ghost Army insignia, not used by the actual unit, but associated with the unit after the war


>It screams spook activity.

Like, it's a FBI honeypot or it's a CIA op to help dissidents?


Push western news narratives, propaganda, help dissidents/traitors. Cultural colonization.

Well the NSA is probably decent at building software. But none of this really matters if the OS isnt secure and the OS is much more complex and less verifiable. Also, why does the government use their own versions of Signal that are more secure, makes you wonder.

Im sure they allow bugs that benefit the NSA.

https://www.malwarebytes.com/blog/bugs/2026/10/signal-bug-co...

Downvote me, but yes I believe that Signal and the EFF are US military programs larping as civilian institutions.


White-listing contacts would have prevented the gaffe when a journalist was added to a war planning group

Would it? As I remember the journalist's number was added to an existing contact by some smart feature in iOS, not sure how a whitelist would help there.

    Well the NSA is probably decent at building software. 
Ultimately I'm sure they are but my secondhand understanding from friends is that it's difficult for the US gov't to attract and retain software engineering talent.

It's not just total compensation. Working on classified stuff can be a bit of a career killer, since you can't talk about your past work when looking for a new job.

Plus there's the antipathy toward the current Trump administration, and the general moral unwillingness of many to do defense-related work in general.

Also, last I heard, cannabis use was still a disqualifier for DoD work and/or various security clearances. There are periodic ongoing polygraph tests, not sure about ongoing piss tests. The government is big; I'm sure the exact policies differ from department to department. But this turns a lot of people off.

Of course, every single thing I've said above might also be viewed as a positive by some engineers. Maybe they attract enough "true believers" to more than make up for the other factors.


I would argue that US intelligence, especially on the technical side of things has put emphasis on privatization of these sectors.

US intelligence knows they can't go to policy makers and get funding for billion dollar domestic spy programs. So they stand up private companies and give them smallish grants at the beginning to get them on their feet, i.e Flock, Palantir, Signal Foundation, EFF, cloudflare, oracle. I could go on and on about orgs that their start off the back of US intelligence, whether through direct funding or through their alumni network.

The US Department of Defense has a long tradition of private public partnerships, but the privatization of intelligence has been accelerated in the last decade.

So while the NSA might not be able to attract talented SWEs for gov salaries to work on defense, their vendors that owe them their existence can.


That's a great insight!

(Damn, that looked like Claude-speak. I swear I'm human. But then again, that's what a bot would say...)

Its still prohibitively expensive, and models 5% the cost are faster and get the job done just as well (for the average person's workflow).

I really don't get the value proposition of these massive models, seems like they're just loss leaders simply to say you're the in the top 3. There's not much use cases outside of niche academic research (and even then I question the need). They mostly function as marketing tools, like a casino that offers high roller accommodations.

People will say "cyber" but companies didnt want to spend on security before, not sure why they would all of a sudden be compelled to spend even more.


Not everyone is writing CRUD business apps and personal use software. Pretty much any model can do those now. The frontier is for doing things those models struggle with. Even in vibe coding a simple video game the difference in results between an open source model, the previous frontier model, and the current model is pretty huge.

Security is mandatory now in a way that it wasn't, now that AI has eliminated the skill barrier to hacking.


The llm induced Dunning Kruger is getting really old.

You dont need Fable to do most things. Your video game is still a "CRUD" app and you certainly wont make a great video game because you have access to fable. But marketing works well, people actually believe they can buy competency for themselves.

It hasnt eliminated the barrier to hacking. Please go find me a valid crit on any program on h1 with your fable 5.2 within the next 3 days, Ill bet my entire life savings it can't, and if you did it would cost 3x the bounty. You've been reading too much engagement farming articles/posts

Security has always been highly important, not any more than it is now. Companies are not going to suddenly start spending six figures on fable to protect themselves from imaginary AI powered hackers.

Its there for people who fall for cheap marketing who thnk they can shortcut meritocracy and make video games, or do security research with no skill. Good luck! Dont max out your CCs on llms.


Reminds me of how I always see people on X announce to no one in particular that they use Fable on xhigh for everything. Like it's a status symbol to run a more expensive, bigger, slower model for Git CLI commands and to summarize their emails or whatever.

I think the general notion that a lot of people are using models that are overpowered for their use case is right, but I think the big question is where the line actually is and how many people fit on either side. My suspicion based on my experience is that there are significantly more people on the "actually need the frontier model" side than the bear case suggests, but I'm not speaking with any authoritative data here and could be totally wrong.

I used to work on evals/observability at Opik: https://github.com/comet-ml/opik And I can say from working with lots of users running agents at scale, there is definitely a pretty commonly reached frontier of difficulty/complexity/abstraction where the capability difference between something like Fable/Astra and smaller models is a big deal. I find this to be especially true the more autonomous your workflow gets. You can still often tackle the hardest tasks with something like Opus, it just typically involves more direct intervention and design work from the user.

But again, this is just anecdata on my part. All of this is still pretty ambiguous and lots of companies are trying to solve the problem of matching models to task difficulty (I.e. routers), with varying degrees of success.


Fable 5.2 on low is both cheaper and more intelligent than Opus 5.5 on high. Prohibitively expensive is a gross oversimplification.

People looking at frontier models release are not the average person, and let's say that you pay $150 with Fable to save $1000*, it's much more interesting than using $10 with a cheap model to save even $500.

*Scale numbers to whatever suits you.


If you think llm usage provide that much value I got a bridge in Brooklyn to sell you.. Their main value is as a tool for wage theft (which im cool with, jusy recognize it), so lazy people can cheat at their jobs and get by doing less. Basically every RFH engineer admits this to me. Work two hours, company pays for 8 plus a tool that does mediocre work, that everyone else has access too and thus provides no edge.

Fable class models are for people who fall for cheap marketing and are lazy and think they can now bypass a meritocracy and buy themselves competency. Really they'll just continue to be lazy, and the people with more competency will run circles around them with 1/10th the amount or llm usage, or none at all.


Well for tasks that can be done with another model you just use another model.

I thinks more likely ya'll are experiencing a phenomenon that has been discussed on here before.

At the model release people become high/manic/addicted to the newest model think they now have some new superpower and then that "high" wears off and you think the model isnt as impressive.

The truth is llms have a slot machine-esque UX to them, and the labs designed it this way.

Dont worry you'll get your new model with some mystical name any day now and a month or two afterwards you'll be posting about how it no longer feels as good to use.


Why its been disclosed and patched....

Finally companies are realizing that subsidizing unlimited AI for your employees is basically just paying for the tools for them to work 2hrs a day while you pay them for 8.

Listen, I dont care if employees scam their bosses by offshoring their work to AI, but its absolutely the case that 90% of people irresponsibly use AI to be lazy, get by doing less work.

Unfortunately, as Ive personally witnessed when you do this and take away the cheating tools engineers just start copying and pasting IP into their own AI subs where you have less garuntees about ZDR. Employees will gladly spend an extra $200 a month so they can work 4hrs less a day and just get by.

Only 10%, maybe less, are actual super contributors with AI. The rest are literally getting a form of diabetes of the brain and the exercise required to reverse that disease is quite uncomfortable, especially when you've been hands off using a cheating tool for a year straight.


Companies deserve it and then some.

I wont argue with you there, I dont really care about the companies at all

Im just saying employers will notice that its basically being used as a cheating tool by a large subset of users and they 'll stop subsidizing that behavior.

What Im really trying to say is its not a producitivity enhancer for an enterprise, maybe for a small percentage of employees, so I dont really see the value in it as an enterprise wide investment.


Most detection cant be mitigated on the fly by clicking a turnstile or typing captchas anymore, so no idea what AI is doing for you.

Web scraping at scale requires solvers. Even if it wasn't cost prohibitive, an llm controlling a browser would be quite a heavy/inefficient way to solve captchas.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: