New news, i reverse engineered SPTM and i found vulns which lead to succesfull jailbreak (14 vulnerabilities), stay updated, Supported versions and chips:
A12 - A17 (iOS 26.0 - 26.5 - probably 26.6)
Creds for PoC:
@rooootdev
AI agent that turns RE research into a queryable knowledge base.
Ingest → Compile → Search → Verify.
Pulls web articles, structures them into a wiki, searches like a graph, tags every claim.
Link: https://t.co/3PqE2FNEjk
Repo: https://t.co/GN2cKlULTW
As promised, DarkSword Kernel Exploit writeup is now live at https://t.co/aji0Xc3JNU
This goes over the root cause, what happens on the Kernel side and how the kernel exploit is implemented.
Hope it helps anyone looking to understand it! :)
Or… what if we gave you $100 in Codex credits if you tell us what you love about GPT-5.6 Sol or why you switched?
Tweet it, claim your gift, enjoy more usage. First 10k get the free tokens!
https://t.co/8mU93eA13i
6 months of Claude Max 20x, on us.
We're expanding Claude for Open Source to more of the community.
If you're a maintainer, a core contributor, someone landing PRs across the ecosystem, or someone keeping a critical package alive, apply today!
Open weight models ≠ you must run them locally.
*️⃣ Please stop listening to the clowns telling everyone to build a home lab and spend $50k��$100k to run models locally. The only time that makes sense is if you’re making multiples of that money from training models, selling GPU infrastructure, or otherwise making money because of it.
*️⃣ For inference, you don’t need to waste money on hardware, not even a $1k RTX 3090. The quantized models you’ll run at that price point aren’t worth the trade-off when the same money buys you more than a year of access to top providers.
*️⃣ Inference helps you make money through your actual work. That work pays for your AI subscriptions. A home lab just adds costs while still leaving you on quantized models that won’t match the frontier models people imagine.
And I’d bet most of the people telling you to buy GPUs and build a home lab still have active OpenAI or Anthropic MAX subscriptions themselves:“
Don’t follow those clowns:"""
Breaking
BOE is reportedly trying to enter Samsung’s Galaxy S27 OLED supply chain with a very aggressive offer.
According to the report, BOE proposed supplying OLED panels for the standard Galaxy S27 at a price $5 lower than Samsung Display.
Samsung Electronics has already sent an RFI to BOE and is evaluating samples, meaning BOE is now seriously being considered as a potential panel supplier for next year’s Galaxy S27 series.
The logic is very clear.
BOE is under OLED shipment pressure as Chinese smartphone makers cut production, so it is now using price as a weapon to compete for Samsung’s orders.
If BOE succeeds, this could become a major shock to Samsung’s internal supply chain.
Samsung Mobile would get cheaper panels.
Samsung Display would face direct price pressure.
Korean component partners tied to Samsung Display could also see their profits squeezed.
This is the brutal reality of the OLED industry now.
Even Samsung’s own smartphone division may choose a Chinese OLED supplier if the price is aggressive enough.
A $5 difference sounds small, but at Galaxy shipment scale, it can move tens of millions of dollars.
BOE is not knocking on the door anymore.
It is trying to break into Samsung’s house with a cheaper OLED panel.
I've finally figured out how Samsung's ODIN download mode protocol handles LZ4 compressed data. I haven't seen any public open-source implementations or documentation on this yet. Stay tuned!