01-24-Daily AI News Daily
Daily Summary
OpenAI is taking a cut from the money you make with AI. Pharmaceutical companies using their technology to develop new drugs will share profits, forcing all API entrepreneurs to recalculate costs.
An Nvidia engineer open-sourced the world's first deep learning framework entirely generated by AI, with the paper written by AI too. Jia Yangqing called this the "AlphaGo moment of software development."
The Agent programming era has arrived. GitHub and Anthropic have brought AI assistants into the terminal, so developers no longer need to switch windows to write code.⚡ Quick Navigation
- 📰 Today’s AI News - Latest updates at a glance
💡 Tip: Want to experience the latest AI models mentioned in this article (Claude 4.5, GPT, Gemini 3 Pro) right away? No account? Head over to Aivora to grab one—ready in a minute, hassle-free support.
Today’s AI News
👀 One-Liner
OpenAI is now taking a cut from the money you make with AI—this time it’s not selling tools, it’s sharing profits.
🔑 3 Key Hashtags
#BusinessModelShift #OpenSourceBoom #AgentProgrammingEra
🔥 Top 10 Headlines
1. OpenAI’s New Business Model Exposed: AI Helps You Make Money, We Take a Cut
Just when everyone thought OpenAI only made money selling API access, they dropped a bombshell. OpenAI CFO Sarah Friar revealed at the Davos Forum that if pharmaceutical companies use OpenAI’s technology to develop new drugs, OpenAI will take a cut of the drug’s profits. This isn’t selling software—it’s “value sharing”—you get rich, we get a piece of the pie. This model could expand to energy, finance, and other sectors. Sounds great? Here’s the catch: if you used Photoshop to design something and Adobe wanted a cut of every project you sold, would you be cool with that? This shift from “selling tools to sharing profits” could completely rewrite the rules of the AI game. For startups building on OpenAI’s API, the cost calculation logic just got flipped upside down. The irony? OpenAI itself is being sued by The New York Times and others over training data copyright issues, yet now they want a cut of users’ intellectual property. This move is honestly baffling.
2. Alibaba Qwen3-TTS Open-Sourced: 97ms Ultra-Low Latency, Clone Voice in 3 Seconds
Text-to-speech used to be a trade-off: either high latency that drives you crazy, or voice cloning that sounds robotic. Now Alibaba’s Qwen3-TTS is open-sourced with end-to-end latency as low as 97ms and voice cloning in just 3 seconds. It even supports cross-language and cross-dialect transfer with zero loss. Here’s the kicker: you can design a completely new voice from scratch with just one sentence—like “give me a gentle female voice with a bit of magnetism,” and AI generates it. The model uses a Dual-Track architecture, perfect for live streaming, real-time translation, and more. For apps needing real-time voice interaction, this is a total game-changer. Developers are thrilled—finally no need to choose between low latency and audio quality.
3. Nvidia Engineer Open-Sources VibeTensor: World’s First Deep Learning Framework Entirely Generated by AI
This project is genuinely unsettling to think about. Nvidia distinguished engineer Bing Xu open-sourced VibeTensor, the world’s first deep learning system entirely generated by AI agents—not a single line of human-written code. From Python/Node.js bindings at the top level, to the C++ core scheduler, down to CUDA memory management at the bottom—all AI-generated. Even crazier? The paper was written by AI too. Sure, performance can’t match PyTorch yet (some scenarios are 1.7 to 6.2x slower), but it proves AI can build complex system software. Chen Tianqi and Jia Yangqing both gave it thumbs up, with Jia calling it the “AlphaGo moment of software development.” That said, Xu honestly pointed out the “Frankenstein effect”—AI can ensure each local subsystem works correctly, but when you stitch them together globally, unexpected performance bottlenecks emerge. It’s like AI can make every brick perfectly, but the house still leaks. The future programmer might not write code anymore—they’ll write Prompts.
4. OpenAI Releases ChatGPT Atlas Browser Major Update: Tab Grouping Support, Search Mode Enables “Auto-Navigation”
OpenAI’s Mac browser ChatGPT Atlas just got a major upgrade. New tab grouping features support vertical tab optimization for cleaner multi-tasking browsing. Even better is “Auto Mode”—the system automatically switches between ChatGPT and Google Search based on your query. Ask “what’s the weather today?” and it hits Google; ask “how do I write a web scraper in Python?” and it switches to ChatGPT. This smart navigation means you never have to manually pick a search engine again. Plus, Atlas now deeply integrates with macOS keyboard text replacement and iCloud password extensions, further smoothing the migration experience from mainstream browsers. For heavy AI users, this is definitely worth trying.
5. Musk’s Davos Debut: Optimus Robot Going on Sale in 2027, Ushering in the “Household Standard” Era
Musk made his Davos debut and spilled the timeline for Optimus commercialization. He says the robot will be available to the public in 2027 with extremely high performance. By end of 2026, Optimus will be capable of executing complex industrial tasks, though initial production will face “S-curve” challenges. Musk predicts that as AI and robotics proliferate, the global economy will enter an unprecedented era of “abundance.” Still, Optimus’s success depends on scalable manufacturing proof and clear unit economics. Imagine every household having an Optimus to do chores, watch the kids, even chat with you. Sounds like sci-fi, but Musk says it’s 2027 reality.
6. Tencent Releases 2025 Anti-Fraud Report: AI Modeling Enables Precision Crackdowns, 70+ Cases Exposed
Tencent used AI to catch internal fraudsters—results are stellar. Throughout 2025, Tencent’s AI analysis tools built dynamic risk models and uncovered 70+ fraud cases, resulting in 90+ terminations and 20+ criminal referrals to police. Tencent also published a “Never-Cooperate Entity List,” blacklisting 15 companies for fraud. This AI anti-fraud system proactively detects and precisely targets fraud, way more efficient than traditional manual review. For enterprises, this is a case study worth learning from—AI doesn’t just help you make money, it helps you save it.
7. GitHub Copilot CLI Released: Bringing Copilot’s Powerful Coding Agent Capabilities Directly to Your Terminal
GitHub brought Copilot straight into your terminal. Now you can use Copilot directly from the command line without opening an IDE. For developers who spend a lot of time in the terminal, this is a godsend. You can ask Copilot in natural language “how do I delete all .log files” and it generates the command right there. This “what-you-see-is-what-you-get” experience makes coding way smoother. For terminal-loving developers, this is a must-try tool.
8. Anthropic Open-Sources Claude Code: An Agent Coding Tool Residing in Your Terminal
Claude Code is an agent coding tool that lives in your terminal. It understands your codebase and helps you code faster by executing everyday tasks, explaining complex code, and handling git workflows—all through natural language commands. For example, you can say “refactor this function for me” and Claude Code automatically analyzes the code and suggests optimizations. This “AI assistant” mode makes coding way more efficient. For developers frequently dealing with complex code, this is worth trying.
9. DeepSeek Open-Sources FlashMLA: Efficient Multi-Head Latent Attention Kernel
DeepSeek open-sourced FlashMLA, an efficient multi-head latent attention kernel. It significantly boosts large model inference speed, especially when handling long contexts. For applications needing to process massive data, this is a super practical tool. DeepSeek has always excelled at inference efficiency, and open-sourcing FlashMLA further solidifies their leadership in this space.
10. Microsoft Open-Sources VibeVoice: Cutting-Edge Open-Source Speech AI
Microsoft open-sourced VibeVoice, a cutting-edge speech AI tool. It supports multiple speech tasks including speech recognition, text-to-speech synthesis, and voice cloning. For developers building speech applications, this is an incredibly practical tool. Microsoft has deep expertise in speech AI, and open-sourcing VibeVoice further lowers the barrier to entry for speech AI development.
📌 Worth Watching
[Product] Agentation Officially Launched - Point-and-shoot AI programming by combining Agent with Annotation, giving AI coding assistants precise webpage structure visibility.
[Business] Amazon Reports Another Major Layoff - This round could cut 16,000 jobs, with AI automation as the main driver, concentrated in high-paying headquarters positions.
[Research] Tencent Collaborates with Tsinghua and Peking University on Render-of-Thought (RoT) - Renders text reasoning steps as image visual embeddings, achieving 3-4x Token compression rates.
[Open Source] OpenBMB Open-Sources UltraRAG v3 - A low-code MCP framework for building complex and innovative RAG pipelines.
[Open Source] Remotion Open-Sourced - Programmatically create videos using React, perfect for scenarios requiring batch video generation.
😄 AI Fun Fact
Tech Enthusiast Weekly (Issue 382): The Twilight of Independent Software
Today’s most interesting read comes from Ruan Yifeng’s Tech Enthusiast Weekly. He mentions that the era of independent software might be heading toward twilight. Back in the day, many programmers made a living selling independent software, but now AI’s arrival makes everything uncertain. AI can write code for you, generate content, even automate tons of tasks. So where’s the future for indie software developers? Ruan thinks future programmers might not write code anymore—they’ll write Prompts. This article is worth deep reflection for every developer.
❓ Related Questions
How do I experience OpenAI’s ChatGPT Atlas browser?
ChatGPT Atlas is currently Mac-only and requires download and installation. Domestic users might face download difficulties or account registration restrictions.
Solution: Visit Aivora to get ready-made accounts with instant delivery and hassle-free support.