Howdy. It’s Barsee again.

Happy Thursday, AI family, and welcome back to AI Valley.

Some days, it feels like AI is moving faster than I can keep up. One minute, Elon’s dropping Grok 3, then OpenAI is suddenly thinking about open source again, and now Microsoft is out here building AI that can play video games. It’s a lot. But don’t worry, I’ve got you with this.

In today’s edition:

• xAI releases its latest flagship model, Grok 3

• OpenAI going back to Open Source?

• Microsoft introduces WHAM

• Plus trending AI tools, posts, and resources

Ready, set, go…

GROK 3

Elon Musk’s xAI releases its latest flagship model, Grok 3

xAI just dropped its latest AI model Grok 3 on Monday via a live stream on X, outperforming leading models from OpenAI, Google, and Anthropic in specialized areas including maths, scientific reasoning, and coding.

Here's what you need to know:

• Musk stated that Grok 3 runs on a colossal 200,000-GPU infrastructure (that’s 10x the power of Grok 2!). This massive upgrade means way better reasoning, coding, knowledge, and instruction-following skills.

• During a live demo, xAI team showcased Grok 3’s advanced capabilities, including designing a new hybrid game blending Tetris and Bejeweled and even simulated Earth-to-Mars spacecraft trajectories like a sci-fi movie.

• The Grok 3 has family versions: "Mini" for speed, and "Think" mode for mathematical and scientific problem-solving.

• A standout feature is "DeepSearch" mode, which does deep research almost as well as Perplexity. However, it occasionally generates fake URLs and unsupported claims, making it less reliable than OpenAI’s new "Deep Research."

• Grok 3 is now available to X’s Premium users, offering advanced features like Think and Deep Research. In addition, free users can access Grok 3 with limited usage, but for a limited time.

• xAI plans to launch a voice mode and enterprise API in the coming weeks. The company also intends to release Grok 2 as an open-source model once Grok 3 achieves stability.

Why it matters:

Grok 3 is now the leading non-reasoning model and is considered the top reasoning model, surpassing o3-mini and DeepSeek R1 in several performance benchmarks. However, OpenAI's full o3 model has yet to be released, leaving room for further competition in the space.

TOGETHER WITH FLOW AI

Work at the speed you think with Wispr Flow

Meet Wispr Flow: the dictation tool that sounds just like a more polished version of yourself—minus the "ums" and "uhs." Whether it's an email, doc, or a prompt, Flow transcribes your words instantly. It's 3x faster than typing and automatically cleans up mistakes. That's time back in your day—and fewer embarrassing typos.

Download Wispr Flow for free today.

SIDE UPDATES

🧑‍🔬 Microsoft Announces Quantum Breakthrough with Majorana 1 Chip

Microsoft unveiled Majorana 1, its first quantum chip featuring eight topological qubits that use Majorana particles instead of electrons. Small enough to fit in your palm, the chip is designed to scale up to a million qubits, enabling precise simulations in fields like medicine and materials science. This breakthrough follows 17 years of research on new topoconductor materials, with DARPA selecting Microsoft to help advance scalable quantum computing efforts.

🤝 OpenAI's ex CTO Launches Human-AI Collaboration Lab

Mira Murati, former CTO of OpenAI, has launched Thinking Machines Lab, a research startup focused on developing multimodal AI systems that collaborate with humans, particularly in science and programming. The team includes 30 researchers from companies like Google DeepMind, Meta, Mistral, and some prominent figures from OpenAI. The lab is committed to open science, with plans to publish technical papers, code, datasets, and model specifications regularly.

🔓 OpenAI Going Back to Open Source?

OpenAI CEO Sam Altman is reconsidering the company’s stance on open source development, seeking user feedback on X regarding whether to release an o3-mini level model or a smartphone-optimized model. This shift comes after OpenAI distanced itself from open source following Microsoft’s investment and the release of GPT-4, citing security concerns. However, as competitors like Meta, Deepseek, Alibaba, and Mistral gain ground with open source models, Altman now acknowledges that a fully closed approach is unsustainable.

📵 Humane’s AI Pin is Dead, as HP Buys Startup’s Assets for $116M

HP has acquired most of Humane’s assets, including its employees, intellectual property, and software platform, after the startup decided to shut down. However, the deal excludes Humane’s AI Pin business, which failed to gain traction and will be discontinued. Despite raising $230 million since 2018, the AI Pin was widely criticized for its poor functionality and high price, ultimately leading to the startup’s collapse.

📈 True AGI Means 10% Economic Growth, Says Satya Nadella

According to Microsoft CEO Satya Nadella, true AGI shouldn’t be measured by AI progress alone, its real success lies in driving global economic growth to 10%. Nadella believes the biggest winners will be industries that leverage AI for productivity, fueling accelerated economic expansion.

🧪 Google Develops AI Co-Scientist to Aid Researchers

Google has introduced an AI co-scientist built on its Gemini 2.0 platform to help researchers develop new hypotheses and research strategies. By simply describing a research goal in plain language, the tool provides testable hypotheses, a summary of relevant literature, and suggestions for experimental approaches. Designed to work alongside experts, this tool aims to enhance research without replacing the scientific process.

WHAM

Microsoft introduces WHAM

Microsoft has introduced Muse or WHAM (World and Human Action Model), an AI world model that can understand and simulates interactive game worlds by analyzing video footage alone.

Here's what you need to know:

• WHAM was developed by Microsoft Research Game Intelligence and Xbox Game Studios using actual human gameplay from Bleeding Edge, a 4v4 online brawler. It learned from visuals and controller inputs.

• This AI model has watched the equivalent of seven years of gameplay, crunching through a massive dataset with the help of V100 and H100 GPU clusters.

• Unlike older AI world models that struggle with continuity, WHAM can maintain coherent gameplay for up to two minutes, a big leap over Google’s Genie 2. It also keeps track of new objects in scenes with 85-98% accuracy.

• Despite its advances, WHAM faces significant constraints that currently restrict its practical applications. The model currently outputs at just 300×180 resolution and 10 frames per second, with occasional inconsistencies in character models.

• Microsoft is open-sourcing WHAM’s weights and sample data, allowing developers to experiment and refine the models.

Why it matters: 

WHAM could revolutionize how we design and experience games. Developers might use it to create more dynamic, responsive worlds, while players could see AI-powered environments that react and evolve in real-time.

TRENDING TOOLS

• Surfshark > VPN that hides your IP, blocks trackers, and encrypts your connection, keeping snoops and advertisers out while staying private, stay secure. *

• Reflect > Personal data tracking, insights, and experiments.

• Lovable Visual Edits > Faster and more precise edits.

• Google Career Dreamer > Use AI to find patterns between your experiences, skills, and interests to connect you with careers that might be a good fit.

• OmniParser V2 > Turn any LLM into a Computer Use Agent.

• Breyta.ai > Extract key insights from multiple files instantly.

_{* this indicates sponsored}

THINK PIECES / RESOURCES

• Tesla's robotaxi strategy looks a lot like Waymo's.

• AI killed the tech interview. Now what?

• AI boyfriends, a billion-dollar industry in China.

• New York Times goes all-in on internal AI tools.

• Here's why he switched 50% of my AI work to Claude.

• Quick video tutorials and screenshots to help you master Cursor AI.

• AI can interpret animal emotions better than humans.

• Your most important customer may be AI.

CONTENT CORNER

1/ You can use this prompt for when the AI keeps giving you the same error output.

2/ World’s first USB stick with local LLM. AI in your pocket.

3/ Great progress by Clone Robotics.

4/ AI is just getting started. This decade, we will see many many breakthroughs that will change so much of our lives.

5/ In terms of coding, claude is the best.

THAT’S ALL FOR TODAY

That’s all for today’s issue, folks.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ Like what you see? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.