AI News Daily
AI developments selected from public sources
- 4
- major stories
- 5
- sources
- 0
- first-party stories
Lead story
DeepSeek open-sources Super-agent architecture prototype, unveils five-module design
The DeepSeek Harness team has unveiled the underlying architecture of Super-agent, an autonomous agent exploration project, and open-sourced the prototype dsh-superagent-prototype on GitHub. The architecture consists of five modules: a two-tier Router/Worker, event-driven design, asynchronous multi-channel interaction, storage-compute separation, and an external Watchdog. Its goal is to enable agents to stay continuously online, run multiple tasks in parallel, and proactively respond to environmental events. In preliminary experiments, task completion time was 12.4 minutes at p50 and 15.7 minutes at p95, and Router batching reduced the inference time amortized per individual decision from 6.4 seconds to 1.6 seconds. The team said the experiments were based on synthetic workloads and a simplified environment, and that the plugin is only a prototype for validating the technical approach, not an official DeepSeek product.
Industry news
Apple chooses Qwen for Siri integration, Alibaba AI scores a big win
According to an announcement on Apple's official website, Qwen will be integrated into Siri, Writing Tools, and Visual Intelligence, rolling out with iOS 27.3 in December; this update does not include Siri AI or the next-generation features announced at WWDC. The author traces the 20-month process from Joseph Tsai's public confirmation of the partnership in February 2025 to the filing in July and the announcement on October 31, and argues that what Apple values is Qwen's complete range of model sizes spanning on-device to cloud, along with Alibaba Cloud's private deployment capabilities. The article also cites IDC data showing that Alibaba Cloud held a 46.6% share of China's large language model MaaS market in the first half of 2026.
Parameter Labs discloses exploit chain that can take over ChatGPT accounts
Parameter Labs has disclosed an attack chain targeting the ChatGPT HTML preview (formerly Canvas) sandbox, in which an attacker can craft a share link that steals private data or performs actions on behalf of other logged-in users when they visit it. The attack starts with the MathLive virtual keyboard's postMessage-receiving code using innerHTML, injects an iframe without sandbox restrictions, and then uses the CSP-allowed YouTube oEmbed endpoint to reflectively execute JavaScript. OpenAI has fixed the vulnerability and paid a $5,000 bounty.
Research papers
Qwenputer embeds a RISC-V CPU in Qwen3.5-4B weights and uses GDN state as memory
Qwenputer uses a compiler to write a 32-bit RISC-V CPU into roughly 11 layers of additional weights on Qwen3.5-4B, and treats the model's built-in Gated DeltaNet state as addressable RAM, running C and MicroPython in the forward pass without calling an external runtime. In tests, it reached about 97.5–100% on the DeepMind arithmetic set, raised NUPA long integers from a 48.6% baseline to 100%, and achieved zero mismatches across about 230,000 CPU output positions compared with the reference machine. The cost is a checkpoint of about 18.8 GB, several decode tokens required per CPU instruction, and the need to drive it through RvChat rather than bare generate().