From 7 items, 2 important content pieces were selected

Technology News

  1. Kimi launches K3.1 Pro and K3.1 Pro Highspeed ⭐️ 8.0/10
  2. Hacker News Debates Claimed iPhone Hacking Demo and iOS Defenses ⭐️ 7.0/10

Technology News

Kimi launches K3.1 Pro and K3.1 Pro Highspeed ⭐️ 8.0/10

Kimi has released K3.1 Pro and K3.1 Pro Highspeed, a launch that drew significant Hacker News attention. A quoted open-weight policy says Kimi aims not to release weights that would extend the absolute frontier of cyber capability, that K3.1 Pro's offensive cyber capability is by design below what its pipeline can produce and remains behind frontier proprietary models, and that it intends to progressively raise cyber capability available in open weights. Commenters citing Artificial Analysis reported an Intelligence Index score of 55, fourth overall, ahead of GPT-6 Astra and Fable 5.1 and faster than both, while noting the model generated 150 million tokens during evaluation versus a median of 87 million.

hackernews · Jowsey · Oct 23, 16:59 · Discussion

「Background」 Kimi is a family of large language models from the Chinese AI developer Moonshot AI, which has positioned recent releases as "open-weight" — meaning the trained parameters are published for download rather than being available only through an API. K3.1 Pro is described as a 3T-class open-weight model with 1M-token context and advances in coding, 3D modeling, and financial research, with one report citing pricing of $2/$10 per million tokens and an open-weights date around Nov 4. The launch also lands inside an industry debate over how much offensive cyber capability should be shipped in downloadable models; Moonshot said it identified and removed cyber-related data from both training stages while retaining the general-purpose coding tasks needed for usefulness.

「Impact」 Developers gain a new leading open-weights option: K3.1 Pro scores 55 on the Artificial Analysis Intelligence Index, ranking fourth overall and ahead of GPT-6 Astra, though its status as the top open-weights model depends on the weights being published.

「Community discussion」 Commenters treated the model as broadly frontier-level based on its benchmark placement, while debating Kimi's choice to keep the most cyber-capable weights out of open release. One subscriber to the $99 monthly plan reported hitting the five-hour quota after 47 minutes in a single session.

References

Tags: #LLM releases, #AI benchmarks, #open weights, #AI safety, #frontier models

Hacker News Debates Claimed iPhone Hacking Demo and iOS Defenses ⭐️ 7.0/10

A Hacker News post titled “Today, we hacked what we'd set out to hack: the iPhone” drew 464 points and 299 comments around a claimed iPhone hacking write-up, though the supplied evidence does not confirm the technical details, affected iOS versions, novelty, or real-world impact. Commenters focused on whether jailbreaking remains practical, whether Lockdown Mode would stop such exploits, and why PDFs have repeatedly appeared as an iOS attack surface, with references to Pegasus ForcedEntry. The thread also debated whether LLM-assisted vulnerability discovery helps attackers and defenders symmetrically, and whether iCloud and iMessage lockouts make jailbreaking less useful. The discussion reflects continued interest in iOS exploitation research alongside skepticism about the practical consequences for ordinary users.

hackernews · pavel_lishin · Oct 23, 16:21 · Discussion

「Background」 Jailbreaking an iPhone means exploiting iOS to escape Apple’s software restrictions and gain root-level control; an “untethered” jailbreak is one that survives a reboot. Calif.io’s post is associated with a demo that an external report says took place at Hexacon in Paris on an iPhone 18 Pro running iOS 27.0.1. The Hacker News thread also focuses on Lockdown Mode, Apple’s optional hardened configuration that narrows attack surfaces, with commenters asking whether it would actually block exploit chains like those demonstrated.

「Impact」 For iPhone users who depend on Lockdown Mode for high-risk protection, the demonstrated exploits could undercut confidence in that defense, but whether Lockdown Mode blocks any of these specific attacks has not been confirmed, nor have affected iOS versions. Commenters also note that jailbreaking risks losing access to iCloud and iMessage, though that remains community assertion rather than documented behavior here.

「Community Discussion」 Commenters were interested in the demos but divided on their practical significance: some asked about Lockdown Mode’s real-life effectiveness and LLM-driven exploit discovery, while others lamented ecosystem lock-in and the loss of early jailbreaking tinkering. A recurring concern was that Apple can detect jailbreaks and restrict iCloud or iMessage access, reducing the appeal of maintaining a modified device.

References

Tags: #iOS security, #iPhone exploitation, #jailbreaking, #mobile security, #vulnerability research