I. Industry Development
[OpenAI Tests Next-Generation AI Agent, Leveraging GPT-5 to Handle Multi-Domain Tasks]
On September 25th, according to technology media reports, OpenAI is internally testing a new AI agent, named “GPT-Alpha.” Its core technology is based on the GPT-5 model. Its research and development goals focus on improving AI’s advanced reasoning capabilities and tool usage efficiency, potentially opening up new avenues for the application of AI technology in multiple scenarios.
From the information currently available, it seems that “GPT-Alpha” is not a traditional single-function chat model, but rather a comprehensive agent with multimodal processing and cross-domain capabilities. This feature means it can simultaneously complete diverse tasks such as text processing, code writing, and image-related operations within the same conversation, addressing the need to update its own knowledge base and addressing information gaps in niche fields, further expanding its service boundaries.
Alibaba (BABA) Launches New AI Creation Platform “Zhaodian,” Integrated with Tongyi Wanxiang Wan2.5
On September 24th, Alibaba’s flagship AI application, Quark, officially launched its new AI creation platform, “Zhaodian.” The platform integrates two core capabilities: AI image generation and AI video generation. It is the first platform in China to integrate with Tongyi Wanxiang Wan2.5, becoming the first platform in China to support synchronized audio and video generation.
The integration of Tongyi Wanxiang Wan2.5 is a major highlight of the platform. Leveraging the latest model capabilities, “Zhaodian” can generate voices, sound effects, and background music that match the visuals. Supporting up to 10 seconds of video duration, a single generation allows for a more complete storyline. It also supports 1080P HD video at 24 frames per second, delivering image quality that meets the requirements for creating cinematic scenes. The model’s robust command-following capabilities enable control of complex, continuous changes, such as camera movements, during video generation.
Meta (META) to Open Smart Glasses Offline Flagship Store in Los Angeles
Meta Lab is a new offline retail experience space offering experiences and purchases of Meta’s new headsets, glasses, and other devices. Previously, Meta Lab had launched small retail stores and pop-up shops in select US cities.
Meta recently announced Meta Lab’s latest store opening plans. The pop-up store on Melrose Avenue in Los Angeles will be upgraded to a flagship store, officially opening on October 24th.
The announcement stated that Meta Lab will showcase and sell the full range of Reality Labs products, including the Meta Ray-Ban Display, Meta Neura Band, Ray-Ban Meta 2, Oakley Meta HSTN, Oakley Meta Vanguard, and Quest 3/3S.
[Samsung (SSNGY) Executive Hints at New AI Device, Galaxy Glasses May Unveil on September 29th]
On September 27th, a tech media blog post stated that Samsung Electronics is challenging the chip foundry market with an aggressive low-price strategy. Samsung has significantly lowered its price for its cutting-edge 2nm process wafers to $20,000 per wafer, a 33% discount to market leader TSMC’s $30,000 price tag, signaling the start of a price war in the advanced process technology sector.
Through more competitive pricing, Samsung hopes to ensure its substantial investment in advanced process technology translates into stable market share and financial returns, thereby positioning itself for a favorable position in its long-term competition with TSMC.
[Bloomberg: Meta’s (META) Ray-Ban Display Sells Below Manufacturing Cost]
Recently, Bloomberg reporter Mark Gurman commented in his latest column on Meta’s recently released AR glasses, the Meta Ray-Ban Display: The glasses are priced at $800, significantly below the company’s manufacturing cost. The glasses support text messaging, phone calls, and video chat, and will support viewing Instagram Reels in the future. Meta Ray-Ban Display is monocular, but a binocular solution will be released in 2027, followed by holographic AR glasses that same year. Mark Gurman believes smart glasses will become a core consumer electronics product by 2030.
[Google (GOOG) Unveils Gemini Robotics 1.5 Series Models: Robot Brains]
On September 25th, local time, DeepMind, Google’s core AI R&D division, released the Gemini Robotics 1.5 series of models for robots. The series includes two models:
Gemini Robotics 1.5: A VLA model that converts visual information and instructions into motion commands for the robot to perform tasks. Unlike traditional VLA models that simply execute instructions, Gemini Robotics 1.5 considers and displays the process before executing an action, allowing it to evaluate and complete complex tasks. It also possesses excellent cross-modality learning capabilities, eliminating the need to adjust the model based on different body modalities. Gemini Robotics-ER 1.5: A virtual machine learning (VLM) model that can reason about the physical world, invoke digital tools, and create detailed, multi-step plans to complete tasks. It has achieved state-of-the-art performance on academic and internal benchmarks.
Regarding the release of this model, DeepMind stated that Gemini Robotics-ER 1.5 will be available to developers through the Gemini API in Google AI Studio. Gemini Robotics 1.5 is currently only available to select partners.
[Qualcomm (QCOM) Launches Snapdragon X2 Elite and 5th-Generation Snapdragon 8 Extreme, Fully Betting on the Edge-to-Edge AI Ecosystem]
On September 24th, at the Snapdragon Summit 2025, which kicked off simultaneously in Hawaii and Beijing, Qualcomm unveiled two versions of the processor: the Snapdragon X2 Elite for high-end PCs and the Snapdragon X2 Elite Extreme, designed for ultra-high-end Windows 11 PCs. Qualcomm unequivocally called them “the fastest and most efficient Windows PC processors available today.” Supporting this confidence is Qualcomm’s in-house developed third-generation Oryon CPU core. Built on a 3nm process, the Snapdragon X2 Elite Extreme integrates up to 18 cores, two of which boast a maximum clock speed of 5GHz. According to Qualcomm, this is the first time an Arm-compatible CPU has reached this speed.
[Samsung (SSNGY) XR Headset Debuts at Qualcomm Snapdragon Summit 2025]
Recently, Qualcomm held its 10th annual Snapdragon Summit in Hawaii. During the event, Samsung’s yet-to-be-released headset was on display, though no hands-on experience was available. The product’s hardware specifications are now clear: the glasses will be equipped with a Qualcomm Snapdragon XR2+ Gen 2 chip, dual 4K Micro-OLED displays, support for controllers, eye tracking, and gesture interaction.
Sources indicate that the Samsung headset will be launched online on October 21st, initially targeting the Korean market, with a price tag likely between $2,000 and $3,000.
II. Market Dynamics
[Apple (AAPL) Vision Air Screen May Use the Same CoE Display Package as the 20th Anniversary iPhone]
On September 23rd, it was reported that Apple may introduce the same advanced display technology to two devices in 2027. Specifically, Apple is planning to adopt a new approach to using CoE display packaging technology in the upcoming Vision Air, and they may also use this technology in at least one iPhone in the same year.
An industry insider stated, “Apple plans to use CoE in a bar-shaped smartphone in 2027 to celebrate the 20th anniversary of the iPhone.” “We are trying to apply the new technology to the top-tier model to highlight that Samsung has implemented it before Apple.”
[Wimi Hologram Cloud Inc. (WIMI) Deep Links Delegated Proof-of-Stake Consensus Mechanism to Achieve Node Interaction and Data Consistency]
Amid the global wave of digitalization, blockchain technology, with its unique encryption, distribution, and chain-like structure, is profoundly changing the way information is exchanged and value is transferred. It ensures data consistency and tamper-proofing, laying a solid technical foundation for the transformation from the information internet to the value internet. As a distributed ledger system, blockchain possesses powerful capabilities for processing data from diverse sources. Its decentralization, traceability, and immutability enable deep integration with diverse sectors, including finance, the Internet of Things, and smart manufacturing, fostering a range of innovative business models and application scenarios.
WiMi is reportedly actively exploring a new Deep Linked Delegated Proof of Stake (DL-DPoS) consensus mechanism, aiming to overcome the bottlenecks of the existing DPoS consensus mechanism and promote the further development of blockchain technology. The Deep Linked Delegated Proof of Stake consensus mechanism incorporates a linking mechanism and comprehensive reputation assessment, which, through synergy, optimizes and upgrades traditional consensus mechanisms.
WiMi’s DL-DPoS consensus mechanism incentivizes more node participation, enhances inter-node collaboration and oversight, and, combined with a scientific and rational reputation assessment system, effectively resists various malicious attacks and misconduct, ensuring the stable and efficient operation of blockchain networks in complex network environments. It not only provides practical solutions to the pain points of existing consensus mechanisms, but also opens new avenues for the application of blockchain technology in a wider range of fields, including public and consortium chains.
[Microsoft (MSFT) Announces Integration with the Claude Model! On September 28th, Microsoft officially announced that Microsoft 365 Copilot will integrate with Anthropic’s Claude model. Additionally, Microsoft 365 Copilot will continue to be powered by OpenAI’s latest model, GPT-5, and users can now also flexibly utilize Anthropic’s Claude Sonnet 4 and Claude Opus 4.1 models.
[OpenAI Rumored to be Considering Developing AI Glasses]
Recently, it was reported that OpenAI is poaching design, manufacturing, and supply chain personnel from Apple to build a “full production line” of devices. OpenAI’s hardware is reportedly targeting “pocket-sized devices” and smart speakers, with the first device expected to ship as early as late 2026 or 2027.
In July of this year, OpenAI made a high-profile announcement about its entry into the hardware field and the acquisition of io, a hardware company founded by former Apple chief designer Jony Ive. At the time, Ive revealed to employees that its first product would be a screenless, pocket-sized device, specifically not glasses or any other wearable device. It’s still unclear whether OpenAI will develop smart glasses, but with the investment and rapid development of companies like Meta, it’s not ruled out that OpenAI will change its mind.
