Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers to help you understand technology trends and innovative AI product applications.
New AI products click to learn more:https://app.aibase.com/zh
1. Douyin announced a campaign against the abuse of AI technology, taking measures such as removal and expulsion for violators.
Douyin E-commerce Security and Trust Center issued a notice stating that it will take strict measures against merchants and influencers who abuse AI technology, including removing and expelling violative videos and accounts. The notice pointed out that some merchants use AI to generate false product displays, exaggerate product functions, and imitate celebrity images for misleading promotions, seriously damaging consumer rights and platform order. The platform requires creators to make a self-declaration when publishing AI-generated content; works without declaration will be subject to governance.
【AiBase Summary:】
🤖 Using AI to generate fake product displays and exaggerated promotional content
🎭 Using AI to synthesize celebrity images for misleading promotions
📢 AI-generated content must be actively declared, otherwise it will be governed
2. ElevenLabs Audio Model Updated to Version 2: Supports Long Clips, Seamless Loops, and High-Fidelity Audio
ElevenLabs' AI audio model has been upgraded to version 2, bringing multiple improvements, including support for longer audio clips, seamless loop functions, and higher sampling rates, significantly enhancing the quality and flexibility of audio generation.
【AiBase Summary:】
🎧 New version model supports generating up to 30 seconds of audio clips
🔄 New seamless loop function, convenient for background audio production
🔊 Supports 48kHz sampling rate, improves audio fidelity
3. OpenAI Acquires Data Analysis Platform Statsig
OpenAI acquired the product development data analysis platform Statsig, which will further enhance its capabilities in product iteration and data analysis. The founder of Statsig, Vijaye Raji, will join OpenAI as CTO and continue to operate independently. The platform processes 1 trillion events per day, providing efficient experimentation and data analysis support for product development.
【AiBase Summary:】
📊 The Statsig platform processes 1 trillion events per day, providing efficient product experiments and data analysis.
🚀 Vijaye Raji will help OpenAI achieve the scaling of next-generation products.
🌟 OpenAI acquires Statsig, the founder will serve as CTO and continue independent operations.
4. Amazon Launches Lens Live AI Feature: Real-Time Scanning for Shopping Experience
Amazon launched a new AI feature called Lens Live, allowing users to scan items in real-time through their phone cameras and purchase them directly on Amazon. The feature integrates the AI assistant Rufus, which can summarize product information and answer questions, aiming to improve user shopping efficiency and experience.
【AiBase Summary:】
🛒 New feature: Amazon launches Lens Live, allowing users to scan goods with their camera and purchase them directly.
📱 Convenient operation: Users can aim at specific items to scan, get matching items in real-time, and add them to the shopping cart.
🤖 AI Assistant: Lens Live integrates the AI assistant Rufus, which can summarize product information and answer related questions.
5. Google AI Launches Stax: Help Developers Evaluate Large Language Models According to Custom Criteria
Google AI launched the Stax tool, aimed at helping developers evaluate large language models according to custom criteria, providing a structured evaluation method to improve model testing efficiency and accuracy.
【AiBase Summary:】
🧪 Stax provides a fast comparison function, enabling developers to test different models side by side with various prompts.
📊 Supports custom and pre-built evaluators, helping developers obtain evaluation results relevant to actual needs.
📦 Project and dataset features support large-scale testing, ensuring consistency in evaluation standards.
Details link: https://stax.withgoogle.com/landing/index.html
6. WordPress Launches AI Tool Telex to Simplify Website Building Process
WordPress founder Matt Mullenweg introduced the AI tool Telex at WordCamp US2025, aiming to simplify the website building process through an AI interface, allowing users to create website content without coding. Although still in the experimental stage, its potential has attracted widespread attention. In addition, Mullenweg also mentioned a legal dispute with hosting provider WP Engine, as well as his optimistic view on the future of AI technology.
【AiBase Summary:】
🤖 Telex is an AI-based WordPress development tool designed to simplify the website building process.
🌐 Users can generate content blocks with simple prompts and download them in .zip format for use.
⚖️ Mullenweg mentioned a legal dispute with WP Engine, emphasizing the importance of fair use of WordPress achievements.
7. Liquid AI Launches LFM2-VL Model, Bringing "Compact and Sensitive" AI Visual Language Capabilities to Mobile Phones
Liquid AI's newly released LFM2-VL series model has achieved significant breakthroughs in the field of visual language processing. The series includes two models, LFM2-VL-450M and LFM2-VL-1.6B, suitable for resource-constrained devices and single GPU deployment, respectively. The model demonstrates efficient performance in multi-modal tasks, supports high-resolution image input, and optimizes inference speed and flexibility.
【AiBase Summary:】
🧠 **Lightweight Design**: LFM2-VL-450M is designed for resource-constrained environments, while LFM2-VL-1.6B, although with more parameters, remains lightweight and suitable for device-side deployment.
⚡ **Efficient Inference**: GPU inference speed is improved by up to twice, and it performs excellently in multiple multi-modal tasks.
🖼️ **Flexible Compatibility**: Supports high-resolution image input, processes large-sized images using non-overlapping slicing, and is compatible with mainstream development tools.
Details link: https://huggingface.co/LiquidAI/LFM2-VL-1.6B
8. Apple Open Sources FastVLM and MobileCLIP2: 85x Speed Increase, iPhone Turns into AI Powerhouse!
Apple recently open-sourced two vision-language models—FastVLM and MobileCLIP2—on Hugging Face. These models offer excellent performance optimization and efficient local execution capabilities, bringing new possibilities for edge device AI applications. FastVLM achieves a qualitative leap in speed and efficiency, while MobileCLIP2 focuses on efficient alignment between images and text, suitable for resource-constrained edge devices.
【AiBase Summary:】
🍎 FastVLM is optimized for high-resolution image processing, with the first token response time increased by 85 times, suitable for mobile devices.
📱 MobileCLIP2 is a lightweight CLIP model, supporting real-time multimodal interaction, suitable for image search and content generation.
🚀 Apple's open sourcing of these models promotes the popularization of vision-language models and empowers developers to build smarter AI applications.
Details link: https://huggingface.co/collections/apple/fastvlm-68ac97b9cd5cacefdd04872e
9. MetaGPT Launches End-to-End Automated Testing Tool RealDevWorld, Achieving 92% Accuracy
The article introduces MetaGPT's RealDevWorld automated testing tool, which achieves full-process automation testing through a multi-agent collaboration framework, featuring natural language-driven, self-healing test scripts, and full-stack test coverage, significantly improving testing efficiency and reliability.
【AiBase Summary:】
🧪 Natural language-driven testing: Users can describe requirements in natural language to generate test cases, lowering the technical barrier.
🔄 Self-healing test scripts: AI automatically repairs test scripts that become invalid due to UI updates, reducing maintenance costs.
🌐 Full-stack test coverage: Supports testing across Web, mobile, API, and other platforms, covering the entire workflow.
Details link: https://github.com/tanghaom/AppEvalPilot
10. Switzerland Launches Autonomous Open Source AI Large Model Apertus
Switzerland launched its first fully open-source large language model, Apertus, developed jointly by EPFL, ETH Zurich, and CSCS, with multilingual processing capabilities and an open and transparent development principle, aiming to provide users with an alternative to OpenAI and promote the sustainable development of artificial intelligence.
【AiBase Summary:】
🧠 Apertus is Switzerland's first fully open-source large language model, with multilingual processing capabilities, covering over 1000 languages.
🌐 The development process reflects transparency and openness, with model architecture, training data, and processes all made public.
⚡ Apertus offers two versions, suitable for individual users and complex scenarios, and is trained on supercomputers powered by renewable energy.
Details link: https://huggingface.co/swiss-ai/Apertus-8B-Instruct-2509
