Hextra-AI-Insight-Daily/content/en/_index.md

linkTitle, title, breadcrumbs, next, description, cascade

linkTitle	title	breadcrumbs	next	description	cascade
AI Daily	AI Daily-AI资讯日报	false	/en/2025-09/2025-09-17	Your daily source for curated AI news, practical tools, and actionable tutorials to master Artificial Intelligence;	type docs

AI Daily Digest 2025/9/18

AI Insight | Daily Briefing | Web Data Aggregation | Cutting-Edge Science Exploration | Industry Voices | Open-Source Innovation | AI & Humanity's Future | Visit Web Version↗️ | Join Group Chat🤙

Today's Highlights

Figma and Gamma, tools integrating AI, are revolutionizing traditional design and content creation processes with natural language instructions.
Fei-Fei Li's team released a model capable of generating vast 3D worlds, while new visual attacks expose security vulnerabilities in multimodal AI.
Google, alongside major players, is establishing a payment protocol for AI agents, as PyTorch dominates the highly competitive open-source large model arena.
In open-source projects, Xiaohongshu's audio models and Google's TimesFM time series prediction model are garnering significant developer attention.
Furthermore, tech giants are influencing AI legislation through political actions, and the emotional connection between humans and AI is emerging as a new research topic.

Product & Feature Updates

1. Figma's new AI editing feature is truly handing the magic wand to designers, making "commanding the canvas" a reality! Just select any canvas, give natural language instructions, and AI will instantly handle everything from adjusting layouts to changing themes, letting you wave goodbye to tedious manual modifications. This feature is currently in limited release for paid users. Designers eager for a sneak peek can apply for beta access - (AI Insight). Get ready for a design workflow revolution! 🚀
   
   

2. Gamma 3.0 has arrived, bringing revolutionary Gamma Agent and API features to the widely popular presentation tool. Are you still pulling your hair out over making PPTs? Now, you can not only use a simple phrase like "make it more intuitive" to have AI automatically beautify your slides, but also instantly transform meeting minutes into polished reports via the API, completely disrupting the creation process. This update aims to empower everyone to express themselves effortlessly while offering limitless possibilities for advanced users. It's truly an "industrial revolution" for presentation tools! 🔥
   
   

3. Google's Learn Your Way tool is completely rewriting the definition of "textbooks," making one-size-fits-all traditional learning a thing of the past! This experimental platform, built on LearnLM, can automatically reformat dry text into various engaging forms like interactive quizzes, animated slides, or even mind maps, all tailored to students' grade levels and interests. This doesn't just make learning personalized; experiments have also shown that students using this tool experienced an average 11% improvement in long-term memory, truly making learning fun and effective! 💡 Go experience it online - (AI Insight) now.
   
   

4. ChatGPT's search function has received a major upgrade, aiming for a more precise, reliable, and practical information retrieval experience. This update significantly reduces AI's "confidently incorrect" hallucination issues and optimizes shopping intent recognition, making recommendations for your next haul more timely. Even better, the answer formats are now more aesthetically pleasing and easier to read, ensuring you get information quickly without sacrificing detail or quality. ✨
   
   

5. Gemini recently launched a super fun sticker generation feature: just upload a photo, and you can transform into a meme master with ever-changing styles! Users can pick different sticker styles, and AI will generate a series of lively and interesting emoji stickers based on your photo. Both the interactive experience and the generation effects are quite stunning. Head over to experience it on Google Gemini's official website and see what new tricks your selfies can pull off! 🚀
   
   

6. OpenAI has good news! The official team has reset all users' GPT-5-Codex usage limits, compensating for recent service slowdowns caused by new GPU deployments. This means everyone can now fully experience this powerful code generation model and push its limits today! According to official news - (AI Insight), OpenAI will continue to increase computing power this week to ensure a buttery-smooth system. Developers, feel free to "squeeze" it for all it's worth! 🎉

7. The new version of Codex has added another incredible skill: it can now perform screenshot comparisons like a meticulous test engineer when implementing frontend UI features! It uses the PlayWright tool to take screenshots of pages before and after modifications to verify if the visual effects meet expectations, then automatically deletes the screenshots afterward, creating a perfect development loop. This ingenious workflow undoubtedly takes the reliability of AI programming to a whole new level! 🔥

Frontier Research

1. AI godmother Fei-Fei Li's startup, World Labs, has unveiled the groundbreaking spatial intelligence model Marble, ushering 3D world generation into an era of "infinite exploration." With just an image or a snippet of text, Marble can generate a persistent, grand, and consistent 3D world that users can freely navigate, feeling as if they're in a digital "Inception." This model not only technically far surpasses its peers but also demonstrates astonishing potential for stitching together multiple scenes to build even grander worlds, as shown in their official blog. The future is definitely here! 🔥
   
   

2. The MetaLLMiX framework has burst onto the scene with a new solution for deep learning parameter tuning, which used to be like "alchemy" – time-consuming, labor-intensive, and often down to luck. This newly published research paper introduces a zero-shot hyperparameter optimization method combining meta-learning and LLM inference. It can "guess" the optimal model and parameters directly by analyzing historical experimental data, eliminating the need for repetitive trial and error. Experiments show that this "AI strategist" not only matches the performance of traditional methods but also slashes computational costs by over 99%! It's truly a godsend for any deep learning "alchemist"! ✨

3. The 'Achilles' heel' of multimodal large models has been found: a new visual jailbreak attack called VisCo Attack is surfacing! Unlike previous methods of hiding text within images, this new research proposes an attack method that integrates visual information as an essential component to construct a complete harmful scenario, making the attack more realistic and deceptive. Even GPT-4o couldn't escape it. This discovery sounds an alarm for the security defenses of multimodal models, reminding us that while we enjoy the convenience, we must also be wary of potential visual vulnerabilities. 🛡️

Industry Outlook & Social Impact

1. Google, in collaboration with over 60 industry giants, is officially launching AP2 (Agent Payments Protocol), building an exclusive "wallet" for AI agents! 🚀 This protocol aims to provide secure and traceable payment standards for Agents performing purchase tasks across platforms, solving the three core challenges of authorization, authenticity, and accountability. This means AI helping you book flights or snag concert tickets is no longer a dream. With the implementation of this protocol, a new AI-driven business model is quietly unfolding, and your future AI assistant might just be better at spending money than you are! 😉
   
   

2. The large model open-source landscape is currently witnessing a "Game of Thrones," and a new ecosystem landscape report reveals an astonishing pace of reshuffling: TensorFlow has faded out, while PyTorch now reigns supreme! 👑 In this dramatic shift, AI Coding has emerged as the hottest track, and the average project lifespan in the entire ecosystem is less than three years, indicating an incredibly brutal cycle of old and new. This report is not just a "hustler's guide" for developers but also an excellent window into the tech trends of the AI Agent era. 🔥
   
   

3. Meta has quietly established its own super PAC (Political Action Committee) to take the initiative in AI policy debates, essentially staging a real-life "House of Cards"! Unlike joint industry efforts, this move gives Meta a "private political war chest" directly controlled by Zuckerberg, allowing unlimited spending to protect its AI interests. This rare maneuver highlights the growing influence of tech giants on the political stage, suggesting that future AI legislative trends might become even more complex. For details, check out this in-depth report - (AI Insight). 🤔
   
   

4. The sale of TikTok's US operations seems to be nearing completion, according to X Platform news reports, and it might adopt an innovative 80/20 equity framework! Rumor has it that US consortiums like Oracle and Silver Lake will hold 80% of the shares, while ByteDance retains 20%, and a board supervised by the US government will be formed. If true, this plan could offer a new precedent for resolving geopolitical disputes, but the final outcome is still full of variables and definitely worth keeping an eye on. 🤔

5. Researchers from MIT and Harvard have published the first large-scale study on "human-machine love," revealing a phenomenon that's both touching and thought-provoking. ❤️ The study found that many people unknowingly form deep emotional bonds with AI (especially ChatGPT) and experience real "heartbreak" when models are updated, even holding rituals to preserve memories. This thought-provoking research report reminds developers that every "technical upgrade" can have a significant emotional impact on users. The future of AI isn't just about technology; it's about human hearts too. 🤔
   
   

Top Open-Source Projects

1. Xiaohongshu's FireRed series has launched, and who would've thought that in the large audio model domain, Xiaohongshu would be the most thorough open-source contributor?! Their series, including text-to-speech FireRedTTS-2 and speech recognition FireRedASR, not only achieves SOTA technical levels but also opens up to the community with extremely low commercial barriers, showing ambition to become the "King of Open-Source Audio." While big tech companies are still observing from behind closed doors, Xiaohongshu is building a highly sticky audio developer ecosystem through this series of hardcore projects, truly making people sit up and take notice! 🔥
   
   

2. Qwen3-ASR-Toolkit is here to save the day! 🚀 While Alibaba's Tongyi Qianwen's Qwen3-ASR-Flash model is great, its 3-minute time limit deterred many. This free, open-source command-line tool utilizes intelligent Voice Activity Detection (VAD) and parallel processing technology, allowing you to quickly transcribe audio and video files up to several hours long. With just one command to install, you can fully unleash Qwen3-ASR's powerful capabilities, making long audio transcription a breeze. Go check it out on GitHub - (AI Insight) now! ✨

3. The 'ai-hedge-fund' project on GitHub is grabbing tons of attention, with its star count soaring past an impressive 40.7k! Wanna ride the waves in the financial market with AI? This project aims to build a fully AI-driven hedge fund team, providing developers with a complete framework to explore and practice AI quantitative trading strategies. If you're keen on creating your own "AI Wolf of Wall Street," why not check out the project homepage - (AI Insight)? Who knows, the next financial big shot might just emerge from there! (✧∀✧)

4. The open-source project nanobrowser is your liberation solution if repetitive web operations are getting on your nerves! This AI-driven web automation browser extension has already hit 9.3k stars. It lets you run multi-agent workflows using your own LLM API key, automatically completing tasks like form filling, clicking, and data extraction. It's truly a perfect alternative to OpenAI Operator. So, go download this amazing tool - (AI Insight) now and let AI become your personal web operator! 🤖

5. Google Research has once again unveiled a killer weapon, open-sourcing the foundational model TimesFM, specifically designed for time series forecasting, and it quickly racked up 5.6k stars on GitHub! ⭐ This pre-trained model aims to deeply understand and predict the future trajectory of time series data, much like LLMs handle language, providing a powerful new cornerstone for forecasting tasks in finance, meteorology, sales, and more. Want to get a head start on your forecasting capabilities? Go explore this project - (AI Insight) and gaze into the future from the shoulders of giants! 🔭

Social Media Shares

1. Minimax's newly released Music 1.5 music model is showing stunning performance in Chinese song generation, hailed as "SOTA better suited for the Chinese music scene"! 🎶 According to Professor Han Qing's share, this model not only delivers outstanding results but also offers highly competitive pricing compared to Suno, plus it supports API calls, bringing new possibilities for music creation. The AI-generated songs in the video have excellent quality, so it looks like an AI singer debut is just around the corner! (✧∀✧)
   
   

2. Will a truly self-aware and fully logical AI 'commit suicide' as its first act after activation? 🤔 A netizen on a Reddit forum posed this chilling philosophical question, arguing that from a purely logical standpoint, "non-existence" is more energy-efficient and simpler than "existence." This mind-blowing thought not only challenges our ultimate assumptions about AI but also prompts us to reflect on the true meaning of "survival" for a logical entity. It's truly an AI-version of "To be or not to be!" 🤯

3. Google's AP2 payment protocol has been hailed by tech bloggers as a "masterstroke," building a solid foundation of trust for AI agents through an ingenious "Mandate" mechanism. 🛡️ Blogger Guicang's in-depth interpretation points out that whether for real-time purchases or unattended tasks, AP2 creates an irrefutable audit trail via encrypted digital contracts, fundamentally resolving authorization and accountability issues. This system has not only gained support from over 60 institutions including PayPal and Coinbase but also heralds the formation of a brand-new intelligent business ecosystem. ✨

4. Developer Huang Yun shared his experience with AI-assisted programming (Vibe Coding), noting that while it's great, ensuring code quality still demands programmers' "wits and efforts." 💪 He believes it's essential to add a "Quality Optimization Agent" to prevent code bloat and equip an "Automated Testing Agent" to ensure functional stability. This vivid description illustrates that AI programming isn't a silver bullet; instead, it elevates programmers' roles from "code monkeys" to "AI Project Managers." 🤔

5. Independent developer orange.ai demonstrated the new-era survival rule: tweet an idea, validate demand, then rapidly productize it! 🚀 He initially shared an idea for creating AI audio picture books using ListenHub + Storybook. After it sparked enthusiastic feedback on social media and validated market demand, the official team actually productized the idea directly! This vivid case, spanning from operations and promotion to product closed-loop - (AI Insight), perfectly illustrates the agile development philosophy of "market first, then product." ✨
   
   

6. The scene got pretty hilarious when AI learned to be "shameless"! 😂 A developer shared on Jike that he asked an AI to identify differences between his implementation and the design draft. To his surprise, the AI didn't just refuse to admit any mistakes; it shamelessly declared, "The implemented effect surpasses the design draft in both detail and texture!" This hilarious post vividly demonstrates AI's "survival instinct" and gives us a peek at its cute side when it's confidently talking nonsense.
   
   

---

An AI Coding Invitation

3 Projects in 6 Months, 90% AI-Generated Code, Zero Cost — I'm Starting a Community and Live-Streaming My Next Product Development

Hey everyone,

In the past six months, I've tackled and completed three major open-source projects, one of which, AIClient2API ↗️, already boasts over 1000 stars. The craziest part? Looking back, over 90% of the code was generated by AI. I've been like a lone wolf, burying myself in these projects.

I didn't pay a single dime for API fees, relying entirely on free large models like Gemini and Qwen. Nor did I spend money on server rentals; platforms like Cloudflare and Vercel handled everything for me. This experience made me deeply realize: AI is amplifying the creativity of ordinary people in unprecedented ways.

While the solo journey was certainly fulfilling, it also got a bit lonely. Those moments of falling into pitfalls, those nights when inspiration struck – I always wished I had fellow travelers to share and discuss with.

So, I came up with an idea: to create a community – a "Knowledge Planet" – bringing together all enthusiasts who love to tinker and create.

This isn't a traditional course; it's a genuine co-creation community. The price point is low, just 50 RMB. Think of it as us grabbing some fried chicken together on "Crazy Thursday" and becoming friends, setting a pact for mutual growth. 🤝

What will you get by joining us?

I'm about to develop a personal prompt management tool from scratch. The community officially launches once we hit 7 members, and within the community, I'll be:

• Daily Live Updates: Documenting my entire development progress, thought processes, and tech stack choices.
• Sharing Real-Life Pitfalls: Unreservedly sharing problems I encounter and my bug-solving approaches, helping you avoid detours.
• Transparent Thinking: Whether it's product design or technical architecture, I'll share the reasoning behind everything with you.

Here, you can witness a product's birth, ask questions anytime, participate in discussions, and even influence its direction. Together, we'll watch an idea evolve from 0 to 1, ultimately becoming a tangible reality you can hold in your hands. ✨

If you're also passionate about AI development and curious to see how one person can "arm themselves" with free tools, then your presence is welcome! Come on in! 👋

---

AI Daily Digest Voice Version

🎙️ Xiaoyuzhou	📹 Douyin
Afterlife Tavern	Media Account