@TomZarebczan@1MILLIONMRR@robustus We can do a network upgrade which trustlessly proves the effective circulating supply without waiting for coin holders to move their coins!
@shitcoin_maxxi@trading_axe >****** Foundation finds undetectable infinite mint in ZEC
> Drum up ****** influencers to relentlessly shill ZEC to increase EL
> ???
> Profit
@Apeguru I fully agree, I was just illustrating a point, and this isn't how you'd go about auditing an entire codebase.
If you or anyone else is willing to sponsor the compute needed to replicate this with the whole repo, I'm more than willing to do it. I think it would be quite trivial.
I was able to replicate finding the Zcash Orchard vulnerability using GPT 5.5 without a harness and little to no steering (except for passing the files).
GPT 5.5 is consistently able to find this vulnerability.
Opus 4.8 released on May 25th, and the vulnerability was found one day later, on May 26th. GPT 5.5 released on April 23rd, meaning someone could’ve found the vulnerability using GPT 5.5 for an entire month.
(Link to chat in next post)
@Apeguru@AriDavidPaul@CraigSalm@zooko I’m assuming this is about my post.
The prompt of course contains the code that is known to be vulnerable. But the prompt doesn’t have to be biased. Any prompt works. Like “find the scary vulnerability pls”.
https://t.co/QAXE5xHRnX
Addressing some criticism from the comments:
The model uses search:
Search is turned off in settings. The model is prompted to not use search. Thinking traces can be checked to see if search was used.
Brute-force prompts until one sticks:
The prompt doesn't matter. GPT 5.5 can consistently find it using any prompt (see attached picture). The only thing that matters is the code supplied. Which could be found by using a harness and enough compute.
I was able to replicate finding the Zcash Orchard vulnerability using GPT 5.5 without a harness and little to no steering (except for passing the files).
GPT 5.5 is consistently able to find this vulnerability.
Opus 4.8 released on May 25th, and the vulnerability was found one day later, on May 26th. GPT 5.5 released on April 23rd, meaning someone could’ve found the vulnerability using GPT 5.5 for an entire month.
(Link to chat in next post)
Addressing some criticism from the comments:
The model uses search:
Search is turned off in settings. The model is prompted to not use search. Thinking traces can be checked to see if search was used.
Brute-force prompts until one sticks:
The prompt doesn't matter. GPT 5.5 can consistently find it using any prompt (see attached picture). The only thing that matters is the code supplied. Which could be found by using a harness and enough compute.
Addressing some criticism from the comments:
The model uses search:
Search is turned off in settings. The model is prompted to not use search. Thinking traces can be checked to see if search was used.
Brute-force prompts until one sticks:
The prompt doesn't matter. GPT 5.5 can consistently find it using any prompt (see attached picture). The only thing that matters is the code supplied. Which could be found by using a harness and enough compute.
@btcplanet Supplying the code location is a fair criticism, which I highlighted in my original post, but can be trivially solved using enough compute and a harness
As for brute-forcing pompts, see:
@zeroSp3c The hard part is finding the relevant pieces of code, but this can be solved by using a simple harness.
Any medium technical person could’ve found this bug imo
Thats not how it works. You can check the thinking traces to see if search was used. Thats why I provided the chat link. This is the same way the erdos problem was solved by a teenager without a mathematics background.
The knowledge cutoff of the model also doesn’t change.
https://t.co/gxtARohh3D
Thats not how it works. You can check the thinking traces to see if search was used. Thats why I provided the chat link. This is the same way the erdos problem was solved by a teenager without a mathematics background.
The knowledge cutoff of the model also doesn’t change.
https://t.co/gxtARohh3D
I was able to replicate finding the Zcash Orchard vulnerability using GPT 5.5 without a harness and little to no steering (except for passing the files).
GPT 5.5 is consistently able to find this vulnerability.
Opus 4.8 released on May 25th, and the vulnerability was found one day later, on May 26th. GPT 5.5 released on April 23rd, meaning someone could’ve found the vulnerability using GPT 5.5 for an entire month.
(Link to chat in next post)
I was able to replicate finding the Zcash Orchard vulnerability using GPT 5.5 without a harness and little to no steering (except for passing the files).
GPT 5.5 is consistently able to find this vulnerability.
Opus 4.8 released on May 25th, and the vulnerability was found one day later, on May 26th. GPT 5.5 released on April 23rd, meaning someone could’ve found the vulnerability using GPT 5.5 for an entire month.
(Link to chat in next post)
I was able to replicate finding the Zcash Orchard vulnerability using GPT 5.5 without a harness and little to no steering (except for passing the files).
GPT 5.5 is consistently able to find this vulnerability.
Opus 4.8 released on May 25th, and the vulnerability was found one day later, on May 26th. GPT 5.5 released on April 23rd, meaning someone could’ve found the vulnerability using GPT 5.5 for an entire month.
(Link to chat in next post)