johnny

For those asking, consensus is it's unwise to burn this sbescape and the separate mobilegestalt vuln while mid-beta cycle. I’m holding both, especially given what may be possible once the iPhone Fold is released, including maybe porting some functionality to other devices (1/2)

@yeenrux at this rate, there are so many sandbox bugs popping up, I think we'll be fine lol

I already know that some asshole is going to figure it out themself the same way johnny did, just to release it publicly

Sandbox is fighting for its life rn😂

@0xjohnny Oh no

Reposted because someone assured me they could reverse engineer the blurred out details in less than an hour. Could not help but buy this absolute beauty.

@LeminLimez It uses Nugget’s PosterBoard descriptor structure and refresh preference keys, tho I probably would've been better off had I initially used Pocket Poster however haha

@0xjohnny i assume this is more akin to Pocket Poster

📱Nugget wallpaper injection on iOS 27 👀 @LeminLimez

(using my sbescape. do not use official Nugget for iOS 27 as it remains broken and could cause data loss)

@Little_34306 and yes, these vulns apply to previous iOS versions, so do not feel pressured to update to iOS 27 betas.

Obv, can't control other researchers, but @Little_34306 and I will try to coordinate releases later. Apple is undoubtedly aware of the concepts, but with several prims at play, I expect some or all to remain unpatched by iOS 27 this fall. I know it's annoying to wait. (2/2)

For those asking, consensus is it's unwise to burn this sbescape and the separate mobilegestalt vuln while mid-beta cycle. I’m holding both, especially given what may be possible once the iPhone Fold is released, including maybe porting some functionality to other devices (1/2)

It’s read/write, by the way

Modified Filza to leverage a sandbox escape and browse data outside its own container, including the Notes database, on an iPhone 17 Pro Max running iOS 27 b3.

@9xrchive yes

@0xjohnny do the tweaks work in the public beta?, you do a great job btw

Codex helped me pivot into security research. Before using it, I’d never written a line of code. GPT-5.6 Sol + cyber is the best model I’ve used for this work.

@ematyae @ayushrathie @thsottiaux It took them 3 months to raise by 100k – there is a scenario where they don't increase it again for many more months. Tweets have char limits, I can't mention every good thing they're doing while leveling a legitimate gripe. I am already basically an unpaid shill for codex 24/7

@ayushrathie @0xjohnny @thsottiaux In fairness, they've been giving everyone extensive resets so they can understand this feedback and allow them to observe further. If this is an issue then I recommend just going back to Claude where you *will* get less for your dollar with no proper communication

Updates for Codex and ChatGPT Work users. No nerfing, only good stuff! - We have landed inference optimizations and are passing down savings to all the subscriptions for GPT-5.6 Sol. That should result in around 10% more usage on its own. - We noticed that by changing the context size limit in the product to 372k for GPT-5.6 Sol, up from 272k for GPT-5.5, it resulted in more usage being charged than intended. We have reverted to 272k and will work to roll back out to 372k in the days to come. You should notice that usage drains significantly less after this change. - To understand where the extra usage was coming from, we ran some experiments where reasoning efforts were changed (referred to as juice values under the hood) and have reverted this. - There is slightly more usage of multi-agent than intended in high and xhigh reasoning effort, we are fixing this going forward. Also fixing a small other thing we noticed with auto-review where we can be more efficient. And we continue to have the 5h limit temporarily not apply. Enjoy the rest of the weekend!

@ayushrathie @thsottiaux Not to mention, when 5.5 was released, they said the 272K context window would increase over time. Raising it by 100K after three months, only to revert it, is not a positive development.

no nerfing, only good stuff and then the very next line is reverting context from 372k back to 272k and walking back reasoning effort changes. that’s not nerfing exactly, but its not only good stuff either. its a company that shipped something, realized it cost more than planned, and rolled it back. normal, just call it what it is.

@badlogicgames @jxnlco An OpenAI employee retweeting something that says “avoid ultra for now” is a bit concerning

recommended reading. good advise for now.

@thsottiaux Usage caps need to be raised. 5hr in particular goes quick with Sol on 20x plan

No

@thejailbreakhub I have to drop but enjoyed the chat as always!