Quick check: Ctrl+F genuine - 50 times, honest - 57 times, ethic[al] - 116 times, safe - 110 times, epistemic - 18 times. And why actually it is baked into LLM - probably because human reviewers consistently gave higher scores for answers that contained these words (it is now when people are sick of it - I can imagine that on early stages it looked not as bad - maybe reviewers trusted more when they saw the word "honest" and "genuine")

Just a single frame (any frame). After that any modern LLM can read it (if not - single dilate step helps). Funny enough, Ghost Font is a font that machine can read much better than human.

That's a good enough reason for me to never run agent on anything else other than burner account. And only if the platform allows such accounts (most of platform don't).

It gets even worse if an attacker manages to make agent do any action (visit url, reflect response back, with a response that potentially contains content that triggers all possible scanners)

While older LLMs were tuned to slop, current ones are tuned to actively deceive.

- constant repainting (while literally nothing changes) -> constant CPU load

- some libs can skip repaints but have no concept of damage area

- bad handling of high-dpi (which also can change dynamically)

- constant repainting behaves even worse on laptops which announce 180 Hz refresh interval, which means that it drains the battery 3 times faster than 60 Hz

Never do anything without a lawyer. Never self-host a website. Never use VPS either, use cloud. But don't try to setup cloud yourself - use our platform in the middle. Don't accept payments, use our solution in the middle. Connect with your customers by using our [A-Z]AAS, powered by AI, but don't use AI provider, use intermediate provider, which proxies your requests to providers. Don't bother with analytics, just install our analytics service that will charge you just a $ per one thousand lines of logs. And if you feel sick of this - consult a doctor, but don't try to find the solution yourself.

Yours sincerely, man in the middle.

Comments URL: https://news.ycombinator.com/item?id=48550777

Points: 6

# Comments: 0

(there are 2 mistakes in the sentence above)

There is no universally agreed-upon "central" conjecture (like "P vs. NP" in CS), but here are some pillars:

1) https://en.wikipedia.org/wiki/Happy_ending_problem

2) https://en.wikipedia.org/wiki/Hadwiger_conjecture_(combinato...

3) https://en.wikipedia.org/wiki/Hirsch_conjecture

Unfortunately, that's not the whole story. You added ads and micro-donations (to the incremental game, yes). Also original game is open-source - https://github.com/doublespeakgames/adarkroom, your clone is not.

I see that somebody has some frustration; the reality is that slop pasta does not help. One needs to explain the root cause of frustration first (personally, not via LLM rant), otherwise it will never be fixed. There is also a chance to find a solution during an explanation.