84% Deploy Gen AI Use Cases in Under Six Months – Real-Time Web Access Makes the Difference

Your product is only as good as the data it’s built on. Outdated, blocked, or missing web sources force your team to fix infrastructure instead of delivering new features.

Bright Data connects your AI agents to public web data in real time with reliable APIs. That means you spend less time on maintenance and more time building. No more chasing after unexpected failures or mismatches your agents get the data they need, when they need it.

Teams using Bright Data consistently deliver stable and predictable products, accelerate feature development, and unlock new opportunities with continuous, unblocked web access.

Power Your AI

Hey there,

I hope your week is off to a great start. I’m sitting here with my coffee, staring at my screen, and honestly, I’m having trouble deciding which tab to keep open, there is just too much happening right now.

Today, we woke up to a new SOTA open-source model, a massive leap in AI video, and a very strange (but smart) business move from OpenAI.

Today:

• Huawei, DeepSeek, and Chips: How China Closes the AI Gap

• DeepSeek-V3.2: The Open Source King Returns 👑

• New #1 Video AI: Runway Reclaims the Throne 🎬

• From APIs to ownership: OpenAI’s Thrive Holdings play

• Ex-Microsoft & Google exec to lead Apple’s AI push

• Kling claims O1 tops Google Veo & Runway

China Just Popped America's AI Bubble: Cyrus Janssen Reveals What Happens Next!

China is rapidly closing the gap with the U.S. in AI and chip development, driven by a massive STEM workforce, strong government support, and long-term planning. Unlike U.S. politics, China’s one-party system allows for fast execution on national tech goals. Companies like Huawei and DeepSeek are competing globally, and the nation leads in AI robotics and renewable energy. Open-source AI and chip self-sufficiency are strategic moves. 

Despite perceptions, China does not aim for conflict but seeks peaceful rise and global integration. Its transformation and innovation pace—especially in infrastructure, energy, and AI—make it a superpower that cannot be ignored.

WATCH THE VIDEO ON YOUTUBE

The Open Source King Returns: DeepSeek-V3.2

If you only pay attention to one thing today, make it this. DeepSeek just dropped DeepSeek-V3.2, and the claims are frankly wild.

For a while, we’ve been waiting for an open-source model that truly rivals the absolute best closed models (like GPT-4o or Claude 3.5 Sonnet). DeepSeek says they’ve done it—and then some.

Why it matters:

• The Power: They are claiming performance levels comparable to "GPT-5" territory (specifically with their new Speciale reasoning model). Even their standard V3.2 is trading blows with the heavyweights.

• The Tech: It uses a massive 671 billion parameter architecture (Mixture-of-Experts), but here's the kicker: they introduced DeepSeek Sparse Attention (DSA). In plain English? The model doesn’t waste energy looking at every single word in a document if it doesn't have to. It’s smart about what it ignores.

• The Result: This makes it incredibly cheap and fast to run, especially for long documents (up to 128k tokens).

We are talking about a model that can supposedly handle complex reasoning and coding tasks at a fraction of the compute cost. If the benchmarks hold up in the real world, the gap between "Open AI" (the company) and "open AI" (the community) just disappeared.

Runway Gen-4.5: Physics Finally Makes Sense

While DeepSeek is handling the text/code side of things, Runway just decided to own the video space again. They released Gen-4.5, and it is stunning.

We’ve seen good AI video before, but it usually falls apart when things move. Objects morph, people walk through walls, and water looks like blue jelly. Gen-4.5 seems to have fixed a lot of that.

The cool part:

• Physics: The model understands weight and momentum. If a character drops a cup, it falls like a cup, not a feather.

• Prompt Adherence: It actually listens to you. If you ask for a specific camera movement or a complex scene setup, it follows instructions much better than previous versions.

• The Score: It’s currently sitting at #1 on the Artificial Analysis leaderboard, beating out everyone else.

It feels like we are inching closer to the "one-person film crew" reality. If you’re a creator, your toolkit just got a massive upgrade.

OpenAI’s "Boots on the Ground" Strategy

Finally, some big business news coming out of San Francisco. OpenAI has taken an equity stake in Thrive Holdings.

This is a fascinating pivot. Instead of just selling API access, OpenAI is now embedding its own engineering teams directly into Thrive’s portfolio companies—mostly "unsexy" industries like accounting and IT services. The goal is to force-multiply AI adoption from the inside out.

It’s a bit of a controversial move—some are calling it a "circular deal" since Thrive Capital is already a major investor in OpenAI—but it shows that even with a $500B valuation, OpenAI is looking for deeper ways to integrate into the economy than just a chat interface.

🧠RESEARCH

Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer

Z-Image is a new image generator that matches leading tools while being smaller, cheaper, and faster. It runs in under a second on normal computers, supports strong photo-realistic pictures and English-Chinese text, includes an image-editing version, and is fully released with code and models for public use worldwide by developers.

REASONEDIT: Towards Reasoning-Enhanced Image Editing Models

ReasonEdit is a new image editing method that lets an AI think through edits step by step. It first explains your instruction in simple terms, then edits, then checks and fixes mistakes. This think edit reflect loop greatly improves how well existing image editors follow directions and avoid unwanted changes.

AnyTalker: Scaling Multi-Person Talking Video Generation with Interactivity Refinement

AnyTalker is a new video model that turns audio into realistic videos of several people talking together. It trains mostly on single-person clips, then learns group interaction from a small multi-person dataset, keeping costs low while improving lip-sync, image quality, natural back-and-forth reactions, and smooth, believable coordination between different faces.

🛠️TOP TOOLS

Each listing includes a hands-on tutorial so you can get started right away, whether you’re a beginner or a pro.

Alpha3D – Turn Text & Images into 3D Models - converts text prompts or 2D images into game‑ready 3D models

AltIndex – AI Stock Picks & Alternative Data Analytics for Investors - an investing analytics platform that blends alternative data with traditional market signals to surface AI‑driven stock and crypto ideas

Amadeus Code – AI Music Generator & Royalty‑Free Library - AI‑powered music platform that combines three pillars in one place

📲SOCIAL MEDIA

🗞️MORE NEWS

• Apple AI chief John Giannandrea will retire and Amar Subramanya, from Microsoft and Google, will lead its AI push as Apple struggles with weak Apple Intelligence reviews while rivals pour money into AI computer systems.

• Kling AI’s Video O1 creates short videos and edits them from simple text instructions, keeps characters and style consistent across shots, and reportedly beats Google and Runway video tools in the company’s own internal tests.

• Nvidia released Alpamayo-R1, an open AI model that helps self-driving cars see and reason about roads, plus a Cosmos Cookbook of guides and tools so developers can train smarter robots and vehicles for real-world use.

• Chinese tech giants Alibaba and ByteDance are training AI models in Southeast Asian data centers to keep using Nvidia chips despite U.S. export limits, while DeepSeek partners with Huawei on AI chips, the FT reports.

• OpenAI and Accenture are partnering to bring ChatGPT Enterprise and AI assistants into large companies, training thousands of staff and launching a program to build custom AI agents that automate work across core business functions.

• AI-made songs are now topping charts and flooding streaming platforms, but most listeners can’t tell they’re artificial, leaving many fans and musicians uneasy about ethics, lost jobs, creativity, and the importance of human backstories in music.

• Flock, a US surveillance camera company, secretly used low-paid workers in the Philippines to watch and label Americans’ footage, raising worries about privacy, who sees this data, and how its artificial intelligence systems are trained.