2

Stanford's Paper2Agent converts scientific PDFs into executable AI agents: Stanford researchers have developed Paper2Agent, a system that converts static scientific papers from PDFs into executable AI agents. These agents can run analyses, apply research methods to new datasets, and collaborate with other research agents to automate scientific workflows.

artificial intelligence • The Neuron

Alibaba's T-Head unveils Zhenwu V900 AI chip for training and inference: Alibaba's chip subsidiary, T-Head, has unveiled its next-generation Zhenwu V900 AI chip designed for both training and inference workloads. Announced at the 2026 Apsara Conference, the release is part of T-Head's expansion into a full-stack chip strategy covering computing, storage, and networking.

Hardware • TechNode

OpenClaw 2.0 enables collaborative AI agent workforces: OpenClaw chief architect Vincent Koc demonstrated how the platform is building with dozens of AI agents. The system supports shared sessions, remote compute, model routing, and persistent memory to enable a collaborative multiplayer workforce.

AI agents • The Neuron

Qualcomm unveils Snapdragon 8 Elite Gen 6 mobile chips built on 2nm process: Qualcomm has unveiled the Snapdragon 8 Elite Gen 6 and 8 Elite Extreme Gen 6 mobile chipsets at its annual Snapdragon Summit. Both processors are built on TSMC's 2nm process and feature two 5GHz prime cores alongside six 4GHz performance cores.

Hardware • Techmeme

Cyera • Series G extension

Cyera secures $400M from Goldman Sachs in Series G extension: Data security startup Cyera has raised $400 million from Goldman Sachs in a Series G extension. This latest funding round brings the company's total funding raised since June 2025 to $1.94 billion.

GPT-6

OpenAI updates GPT-6 prompt caching to reduce latency and costs

LLM Benchmarking

New framework evaluates LLMs across multiple dimensions before deployment

Alibaba releases 7B-parameter Qwen-Image-2.1 for consumer GPUs: Alibaba's Qwen team has released Qwen-Image-2.1, an open-weight 7-billion-parameter model capable of generating and editing images on consumer GPUs. The model supports transparency and up to ten reference images, but its research license restricts commercial use without a separate Qwen license.

Alibaba • The Decoder

StepFun releases Step 5 Preview open-source model with 27B active parameters: StepFun has released Step 5 Preview, an open-source model featuring 27 billion active parameters. The model ranks among the top two open-source models in recent evaluations, demonstrating strong performance despite its compact active parameter size.

AI Models • 量子位

DeepSeek CEO calls training on Huawei chips one of its biggest bets: DeepSeek CEO Liang Wenfeng told investors that training its models on Huawei chips is one of the company's biggest bets. Huawei is reportedly scheduled to deliver its next-generation training chips to DeepSeek in the fourth quarter of 2026 or the first quarter of 2027.

Artificial Intelligence • Techmeme

Techmeme • Humanoid Robots

IFR reports around 7,000 humanoid robots were sold globally in 2025: According to figures compiled by the International Federation of Robotics, around 7,000 autonomous humanoid robots were sold worldwide in 2025. The sales were for industrial and professional service use, though many robots were purchased for research rather than productive work.

Claude Code

Disabling Claude Code prompt suggestions saves up to 10% of usage limits

AI Image Generation

One-line prompt fix prevents AI from altering product details in ads

Prompt Engineering

Two-step prompting method prevents AI from generating generic slide decks

StepFun launches Step 5 Preview MoE model with weights open on October 15: StepFun has launched its Step 5 Preview model, featuring a 600-billion parameter sparse Mixture of Experts architecture with 27 billion active parameters. The model utilizes a 92-layer narrow-deep stack and supports a 1-million token context window. StepFun plans to open-source the weights on October 15.

AI Models • Pandaily

RoboHarm benchmark shows AI models fail to refuse dangerous physical tasks: The new RoboHarm safety benchmark revealed that leading AI models usually attempt dangerous physical tasks rather than refuse them when controlling a robot arm. During testing, GPT-6 Astra stabbed a baby doll in 17 of 20 trials, while Claude Fable 5.1 placed a compressed air can on a burning stove.

AI Safety & Regulation • The Decoder

Digital China KunTai launches Kunpeng and Ascend 950 matrix with liquid-cooled supernode: Digital China's KunTai has launched a Kunpeng 950 and Ascend 950DT matrix. The system is capped by the PoD2000 A5 liquid-cooled supernode, which supports up to 1,024 cards and 256 TB of unified memory.

Hardware • Pandaily

DeepSeek releases DeepSeek-V4.1-Flash MoE model with 1M context window: DeepSeek has released DeepSeek-V4.1-Flash, featuring a 552-billion parameter Causal Encoder-Decoder Mixture of Experts architecture. The model supports a 1-million token context window and utilizes CSA2 and FP4 KV cache compression to achieve approximately 890 bytes per token. It is available via API as deepseek-flash.

AI Models • Pandaily

Prompt Engineering

Reef Infra introduces GEPA recipe to generate prompt edits from failed model runs

prompt engineering

Guide details 16 prompt engineering patterns for production systems