A reverse engineer discovered a hidden Extensions Framework in iOS 27 and macOS Golden Gate that lets Siri hand off voice and text requests to third-party AI models like Claude, with Siri acting as a middleman for system tasks requiring privacy access. A separate, deeper integration point called Model Manager Services could let providers such as a GPT-5.6 model replace Apple's server-side Siri AI entirely, though Apple hasn't confirmed any such partnership.
A leaker known as pdfu shared iOS 27 and macOS Golden Gate framework code showing Siri AI can hand off tasks to Anthropic's Claude or OpenAI's GPT-5.6 models, with Siri acting as an intermediary for system-level access. One demo even shows GPT-5.6 fully replacing Siri as the on-device agent, searching emails and sending iMessages while logging activity in OpenAI's interface.
Andon Labs has released Pion, a platform that lets AI agents operate companies with full autonomy, following two years of research that began with the simulated Vending-Bench test and progressed to real deployments running vending machines, a store, and a cafe. The company is now opening Pion to outside users through a waitlist so more people can experiment with autonomous businesses.
Anthropic reportedly told select investors that its adjusted operating income will be positive for a second straight quarter, according to the Financial Times. That metric excludes major recurring costs like R&D and compute, so the company still posts a substantial net loss on its actual bottom line.
A developer documented practical issues encountered while porting large system prompts from Anthropic's Claude Opus to a self-hosted Ollama setup, citing the move as a response to growing distrust of frontier AI providers. The post argues that inference session data itself—not just user-submitted content—may hold competitive value, and cites the recent controversy over OpenAI's handling of a Navier-Stokes proof as evidence that providers may train on user activity without adequate transparency.
A developer analyzing iOS 27 and macOS 27 code found hidden APIs that would let third-party AI models like Claude act as Siri Extensions, effectively replacing Siri's backend. The feature isn't active yet, but a demo shows Claude appearing within Siri's interface using an App Intents 'Model Delegation' mechanism.
A hobbyist asked the Fable 5 AI tool to design a printed circuit board pairing a Raspberry Pi Pico 2350 with a GDEY0154D67-FL04 e-ink display, four buttons, and exposed I2C and GPIO pins, using only a single plain-English prompt. Unlike an earlier attempt with Claude Opus 4.8 that botched component orientation and routing, Fable 5 worked unsupervised for a few hours and produced a completed 31.8 x 37.32mm four-layer schematic and layout via KiCad's MCP integration, with no manual edits or checks from the designer before manufacturing.
Oak Park High School students Aayush Bathija and Prince Rohatgi, working with UCLA postdoctoral researcher Daniel Soskin, published a 75-page arXiv paper resolving an open question about coefficient ratio bounds in Lorentzian polynomials, a theory associated with Fields Medalist June Huh. The work generalizes earlier results on quadratic polynomials to arbitrary degree, pinning down which coefficient ratios have universal upper bounds and what those optimal bounds are. The students used AI tools, including Claude Opus 5 and GPT-5.6 Sol, for exploration and drafting, while independently verifying every calculation and proof step.
A developer who examined private iOS 27 and macOS Golden Gate frameworks found evidence that Apple has built mechanisms letting Siri delegate tasks to third-party AI models like Claude, or even swap out its own server-side model entirely for one such as GPT-5.6. In demonstrations, Claude handled tasks such as file creation while Siri still executed system actions like setting reminders, and a separate protocol showed an external model receiving Siri's planner prompts and tool definitions to perform system actions directly.
Anthropic's September 2026 threat report reveals a small group of Russia-based freelancers used Claude Code to develop swarm coordination, computer vision, and terminal guidance software for a combat drone system called DronDoc or Serafim. The system could select targets, including people, and issue detonation commands without human oversight, and developers trained it on Ukrainian combat footage while evading Anthropic's geographic blocks via VPNs. Anthropic banned the associated accounts and says the group was not tied to the Russian state, though it hasn't named the organization publicly.
Anthropic tasked its Claude Fable 5.1 model with solving Sir Thomas Urquhart's unsolved 17th-century cryptogram, the Cyphral Distich, and the AI produced a solution within about a day after roughly 44 minutes of reasoning. The cipher, two lines of 32 numbers each, had resisted human cryptographers for centuries and was even featured on Klaus Schmeh's list of top unsolved historical ciphers. Fable 5.1 succeeded by noticing that the number 32 and surrounding textual clues in Urquhart's book pointed to the decoding method that human researchers had overlooked.
Anthropic's September 2026 threat report describes Chinese-based actors using its Claude models across at least five programs, including two tied to military applications. One actor drafted fire-control specifications for an anti-torpedo system and benchmarked it against U.S. Navy systems, while another built roughly 16 software modules for suppressing enemy air defenses, with a default scenario targeting 12 sites in Taiwan such as Patriot batteries and air bases. Metadata reviewed by Anthropic suggested links to Chinese research institutions, including the PLA Academy of Military Sciences.
GoKawiil