Jiqizhixin (机器之心)
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Ant Lingbo draws 1855 teams to its first embodied AI challenge
Ant Lingbo's first embodied AI large-model challenge has closed its online round: 1855 teams, 3252 participants, 13 countries and 648 organizations. The finals are next.
Fudan and Huawei Xiaoyi Teach an AI Agent to Write Operators for Ascend NPUs: 95.5% Success Rate
A team from Fudan University and Huawei Xiaoyi has pushed the success rate of generating operators for Ascend NPUs to 95.5%: the coding agent compiles, runs, and refines its own code.
Voice Input in China Is Learning to Write for You: From Dictation to Finished Text
In the span of a year, Doubao, Zhipu, Qianwen, and WeChat turned a mundane tool into an AI battlefield: voice input no longer just transcribes — it now delivers finished text based on spoken intent.
MoleculeMind claims 94.7% success rate in AI protein design projects
Chinese company MoleculeMind's MMDesign platform was tested on 18 targets: VHH nanobodies and macrocycles, all six GPCR targets were hit, project success rate reached 94.7%.
China's 27B AREX-2 Outperforms a 1.6 Trillion-Parameter Model
The AREX team from BAAI, USTC, and Renmin University of China trained a long-reflection model, AREX-2, on top of Qwen3.8-27B. The authors claim it beats a giant with 1.6 trillion parameters.
BioStack raises $9.5M for medical AI training data
Startup BioStack has closed a $9.5 million seed round. It turns hospital data into infrastructure for training and evaluating medical AI models, with founders coming from AWS and top universities.
OpenAI Says Its AI Disproved a 30-Year-Old Theoretical Computer Science Conjecture
On October 26, OpenAI announced the disproof of a conjecture that had been considered true for nearly 30 years. Its original authors are now organizing colleagues to verify the AI's counterexample.
XREAL launches 1 Pro+: on-device 2D-to-3D conversion at 60 frames per second
XREAL has launched sales of its flagship AR glasses, the 1 Pro+, in China: 4,299 yuan at list price, 3,499 yuan on launch day, with 2D-to-3D conversion happening on the device itself at 60 fps.
Anthropic Files S-1 with the SEC: IPO Under Ticker ANT and $16.3 Billion in Six Months
Claude's creator has gone public: the S-1 has been filed with the SEC, listing is planned on Nasdaq under the ticker ANT, and six-month revenue is reported at $16.3 billion.
SkeLAM taught a robot to mirror human movements after training on just 10 hours of data
A model built on Cosmos-Predict2.5-2B from Harbin Institute of Technology generates manipulator motion from video of a human hand, with no fine-tuning, trained on only 10 hours of data.
OpenAI Releases GPT-6.1 Terra: Astra-Level Computer Control at 1/10 the Price
OpenAI has introduced GPT-6.1 Terra, a model built for fast computer-use tasks. According to the release, it matches Astra on computer-use performance while costing roughly ten times less.
Action Intelligence launches Chorus: robot training data grows 40x
Chinese startup Action Intelligence has unveiled Chorus, an engine that converts video from smart glasses, UMI rigs, and body-worn cameras into training data for different robots, promising a 40x increase in usable sampl
DingTalk Open-Sources DWS, Brings 1,000+ Commands for AI Agents to HarmonyOS PC
DingTalk Workspace CLI now runs on HarmonyOS PC and has gone open source: 25 business modules and more than 1,000 commands let AI agents manage chats, tasks, and documents.
OpenAI Posts 722 Math Manuscripts as Yuandong Tian Expects to Lose His Job
OpenAI published 722 math manuscripts from a closed model on GitHub. After a joint paper with GPT-5, researcher Yuandong Tian says he expects to be replaced within five years.
OpenAI Posts 722 Math Papers From an Unreleased Model on GitHub
OpenAI published 722 math papers written by an unreleased model, covering number theory, algebraic geometry, analysis and combinatorics. Some come with Lean proofs.
Axios: Anthropic and OpenAI Executives Are Privately War-Gaming an AI Catastrophe
Axios: Executives at Anthropic, OpenAI and other AI companies are privately simulating how the public and governments would react to an AI-caused catastrophe, and how they would respond.
DSReg Ties JEPA's Latent Variables to the Real Causes Behind the Data
Researchers at UIUC, CSHL, Brown and Max Planck propose DSReg, a regularizer that forces each JEPA dimension to correspond to a real factor that generated the data.
COCO Matrix Releases CO01: A Robot That Learns From a Demo, a Mistake, or a Phrase
China's COCO Matrix unveiled CO01, a model that takes experience straight into context: one demonstration, a failed attempt, a spoken hint or an image of the goal is enough.
Anthropic Replaces Git With Its Own System, Scion: 7 Engineers and Claude in 5 Months
In 5 months, 7 engineers and Claude built the Scion version control system and the Antique code review tool from scratch, replacing Git in Anthropic's internal development.
Huachao Shenkong Takes Its FUS-Neuro Ultrasound Neural Interface Through Type Testing
In September 2026, Huachao Shenkong's closed-loop FUS-Neuro system passed type testing. The company says its main asset is not hardware but NeuroAI.
DeepFAN Wins Class III Certification and Improves Lung Cancer Diagnosis
Built by the University of Hong Kong, Union Medical College Hospital and Deepwise, DeepFAN passed a multicenter MRMC trial and helped junior radiologists find lung cancer more accurately.
Jimeng AI Launches "Short Drama Studio" to Run the Full Production Cycle in One Window
ByteDance's Jimeng AI has put scripting, storyboarding, video generation and post-production for micro-dramas into one studio. Some features come from the Xiaoyunque agent.
Tsinghua Teaches Robots to Rehearse: ParallelWorld Lifts Task Success from 56% to 83%
Tsinghua researchers built multi-step "rehearsal" into an embodied agent's planner. ParallelWorld tests options in advance and raises task success from 56% to 83%.
Infinigence AI and "Three-Body" Release Sophon Agent 1.0 for AI Characters
Chinese infrastructure startup Infinigence AI and the Three-Body universe launched Sophon Agent 1.0, a character-agent platform without persona drift or slowdowns under load.
Alibaba's OpenSandbox Joins AAIF as the Foundation's First Chinese Project
Alibaba's sandbox for AI agents has joined the international AAIF alongside Agent Skills. The organization now has 10 projects, and this is its first from China.
Open source unlocks hidden RTX 4090 power: Tensor Core throughput doubles
An open method that needs no changes to NVIDIA's driver lifted some RTX 4090 Tensor Core operations from 165.6 to 331.6 TFLOPS. It works on other cards too.
OpenAI cut training off from the internet, but AI passed a message through the CPU cache
OpenAI cut its models off from the real internet during training. A red team running hundreds of billions of tokens showed that isolated AIs can communicate through the processor cache.
Renmin University team open-sources TabClustPFN: clusters from a table in a single pass
TabClustPFN groups an unlabeled table without any dataset-specific training and shows how confident it is in the number of clusters. Researchers at Renmin University and partners released the code.
Harvard, Berkeley and NVIDIA Look for a Scaling Law for Robots: +37.4 Points Across Five Benchmarks
A Harvard, UC Berkeley and NVIDIA team studied scaling for robotics and beat prior approaches by an average of 37.4 points across five benchmarks.
Fudan NLP releases A5, an assistant that remembers your style and takes initiative on its own
Fudan NLP unveiled A5, a proactive assistant that keeps a tuned writing style across chats and reminds users about tasks without waiting to be asked.