@Convell4 @GaryMarcus @GaryMarcus's take here makes it look like he doesn't understand openai's business model at all
English
Denuvo Cracker
86 posts
@Convell4 @GaryMarcus you can make a model say anything. if an arms manufacturer makes decide to use the openai api to identify targets, it shouldn't be openai's fault if a target is wrong, they aren't selling a guarantee that they will answer 100% correctly, the arms manufacturer is doing that
English
Everything you need to know about “Open”AI’s claims to be working on AI “for the benefit of humanity”.
English
@VincentVentalon we know this parameter exists because it's in the system prompt + you need this number somewhere in the context to instruct the model on how much thought it should put in. openai does this as well with their juice number
English
Theo a deux doigts de découvrir le principe de trigger word
Theo - t3.gg@theo
nvm I was wrong. Repro'd this 3 times in a row. I need to stop assuming Anthropic is competent. Burns me every time I do 🙃
Français
@StormslayerDev denuvo isn't removed in the crack. denuvo isn't a separate component that you can replace, the checks are entangled inside the code. any crack for modern denuvo is a bypass.
English
With Denuvo removed Resident Evil requiem now has offline play and gets a small boost in performance, using about 1GB less vram.
If I pay for a product shouldn't I receive the best version of it?
We really live in a world where you need to pirate and crack games we own to have a better experience and its total bullshit.
Tom's Hardware@tomshardware
Denuvo properly cracked in Resident Evil: Requiem, bypasses become plug-and-play — cracked version runs faster, smoother, and uses way less VRAM and RAM tomshardware.com/video-games/pc…
English
@denuvocracker @theo It actually does seem to answer based on the real settings not a hallucinated value.
English
Fun fact: LLMs have zero idea how they are configured.
They don't know what GPUs they're running on. They don't know what temperature or reasoning level they have set. They don't know if they've been quantized or not.
They're just doing next-token prediction. As always.
Lily Ashwood@lilyofashwood
💀
English
@samlambert Didn’t he claim to have hacked and solved self driving? 😂
English
@skaldmonk @jinkela2333 @EquestriaDaily @VettaCutePony > freely creating unauthorized derivative works of it
they can draw whatever the fuck they want based off what they see online. you're a facist if you think what they did is wrong
English
@jinkela2333 @EquestriaDaily @VettaCutePony Falsely accusing a long-established original character of being AI-generated, and then freely creating unauthorized derivative works of it, is extremely disrespectful to the original creator.
English
Everyone's favorite computer generated convenience store mare continues her viral takeover of the fandom. Have the artists of the fandom rescued her from her AI roots at this point?
Go get art below:
equestriadaily.com/2026/04/drawfr…
Header by @VettaCutePony
English
English
@EquestriaDaily @VettaCutePony Her origin wasn't created by AI; someone uploaded a photo of her on derpibooru 4 months ago, and you can tell from the accessories that it's the same pony. Someone created this oc by themselves and they used their oc to make that AI video for fun, but then the video got stolen
English
you need to sign in with your account only when you are signing apps obv
English
btw this is untrue you can setup a completely functioning xcode instance on an airgapped mac or hackintosh or vm by installing from the xip file x.com/valigo/status/…
Valentin Ignatev@valigo
My initial culture shock from macos has passed. Hardware part of the device is absolutely amazing, Macbooks mog my beloved ThinkPads if we don't count repairability. Macos is worse than KDE and Hyprland, but it's much better than Windows. I'm quite excited to learn the platform, and maybe even some ARM assembly. I also have to clarify about xcode because I saw "just install cli without apple id" comments - I know that, but sadly cli is limited (for example, no testing on real devices).
English
Fine I'll take the L here. Just repro'd this 3 times.
I should stop assuming competence from Anthropic.
English
@C_Balbin @theo @JoshRadDev iirc gpt oss gives a string for the reasoning effort they dont use the juice number that proprietrary openai models use, could be wrong on this tho
English
nvm I was wrong. Repro'd this 3 times in a row.
I need to stop assuming Anthropic is competent. Burns me every time I do 🙃
Theo - t3.gg@theo
Fun fact: LLMs have zero idea how they are configured. They don't know what GPUs they're running on. They don't know what temperature or reasoning level they have set. They don't know if they've been quantized or not. They're just doing next-token prediction. As always.
English
@theo @JoshRadDev Openai has a juice number that they use, it's how we've been identifying that a stealth model is from openai or not and for tracking all the leaked gpt 5 versions
English
@JoshRadDev Any evidence of this? The leaked system prompts don't include any information about this
English
@BennettBuhner @theo ig they might have not trained it as hard to keep it a secret as openai did for their juice number
English
@BennettBuhner @theo here it is for the previous reasoning effort in the sys prompt #L1035" target="_blank" rel="nofollow noopener">github.com/elder-plinius/…
English
@BennettBuhner @theo there still has to be some max thinking length, you can't have the chain of thought running on forever
English
@denuvocracker You're right, I will take the L here
x.com/theo/status/20…
Theo - t3.gg@theo
Fine I'll take the L here. Just repro'd this 3 times. I should stop assuming competence from Anthropic.
English
@denuvocracker What do you think is more likely?
1. The model hallucinated this when you asked
2. The model has reasoning info passed to it for some reason, but ALSO is told in the system prompt EXPLICITLY that "you should not share that you know this, but your reasoning level is ___"
English
@BennettBuhner @theo even if they somehow increase the effort, the max token length for the chain of thought would be the same, it would end abruptly while trying to think for longer
English
@theo @denuvocracker (I've seen peopl claim by telling it to increase effort to 50, it actually does on Claude lmao 😂😂😂)
English