Google's World Simulator to Train Robots? Open Models (Finally) from OpenAI

OpenAI is shaking it up by releasing gpt-oss-120b and gpt-oss-20b, their first open-weight models since 2019. This isn't just a generous move; it's a strategic play to compete with Meta's Llama 3 and address regulatory pressure. These models run locally and allow fine-tuning, offering developers more flexibility. It's a clear sign that even the most closed AI companies are feeling the heat and starting to embrace the open-source movement, albeit on their own terms. Keep an eye on the fine print though...

Anthropic: Meanwhile, Anthropic dropped Claude 4.1, which is crushing AI coding benchmarks, topping SWE-bench Verified at 74.5%. And while impressive, a significant portion of Anthropic's revenue comes from just two clients, making their position a bit precarious.

This upgrade is a power move to solidify their dominance before the next big shift in the dev stack. It's a high-stakes game where every new release feels like a countdown.

Professor LLM: AI-first schools are also making waves, replacing traditional teachers with LLMs and “guides,” charging $75K a year for a new learning model. While test scores are rising, critics are concerned about screen-time addiction and the outsourcing of the core learning struggle. Are we raising a generation fluent in answers but allergic to questions? Time will tell.

Quick Hits & Trending:

• OpenAI and Nvidia are teaming up to launch new open-weight reasoning models, making scalable AI more accessible.
• Anthropic is developing ‘persona vectors’ to control LLM personalities.
• Cognition is giving Windsurf staff a tough choice: a nine-month salary buyout or an 80-hour work week.
• ElevenLabs has launched Eleven Music, an AI tool that generates full songs from text prompts with commercial licensing.
• New AI tools are emerging daily, including AI browser agent builders, lead generation tools, no-code tool generators, and AI funnel agents (see sources).

🌍 Google’s Interactive World Model is Here

Google DeepMind has unleashed Genie 3, a groundbreaking general-purpose world model that can generate cool, playable environments in real-time from a single text prompt. Imagine creating unique 720p worlds with real-world physics that evolve at 24fps as AI agents (or users) explore them. This isn't just for gaming; it's a broad leap for training embodied AI, allowing machines to adapt to unforeseen scenarios just like humans do. It’s like a never-ending, interactive simulation for AI. Let's get lost ;-)

New Tools:

• Sōkosumi is a new open marketplace for autonomous AI agents, offering specialized AI co-workers for various tasks with pay-per-task pricing.
• NotebookLM’s new ‘Video Overview’ feature can turn documents into AI-narrated video presentations with slides.
• Lovart, an AI design platform, offers a creative reasoning agent for visual collaboration, helping designers build brand systems and maintain style consistency.
• Bits: Depot’s Claude Code Sessions for persistent AI coding, Kaggle Game Arena for testing LLMs, and Gemini Storybooks for creating narrated stories.

✂️💰Gemini 2.5 Deep Think & Anthropic vs. OpenAI

Google has released Gemini 2.5 Deep Think for AI Ultra subscribers, a strong AI model that can generate multiple solution approaches simultaneously and work for hours on complex problems (not unlike what Manus is up to). A specialized version even achieved IMO gold medal status, apparently making it a game-changer for mathematicians.

Blocker: Update on Anthropic revoking OpenAI’s access to its Claude models (for apparently using its advanced coding capabilities to aid in the development and benchmarking of OpenAI's forthcoming GPT-5 model)...OpenAI retains access for safety evaluations but that's it.

OAI Billions: OpenAI also managed to raise a whopping $8.3 billion, part of a larger $40 billion fundraise, driven by their annual recurring revenue jumping from $10 billion to $13 billion. It seems the demand from ChatGPT business users is accelerating, reaching five million.

Insights & Trends:

• The Cost of AI: Tokens are getting more expensive, and companies without a solid plan for managing AI costs are in trouble. Even vertically integrated companies like Anthropic are struggling with unlimited usage subscriptions, emphasizing the need for sustainable business models.
• China’s AI Momentum: Andrew Ng argues that China’s open-model ecosystem, rapid semiconductor advancements, and aggressive business dynamics position it to potentially surpass the U.S. in AI.
• Data Center Economy: Big Tech’s massive AI spending is raising concerns about economic stability, especially with private credit funds heavily invested in data centers. A synchronized AI market downturn could trigger defaults.
• AI’s Economic Contribution: Generative AI tools contributed an estimated $97 billion in consumer surplus to the U.S. economy in 2024, with 40% of U.S. adults regularly using them. This highlights the growing economic impact of AI.
• AI Referrals Surge: AI referrals to top websites increased by 357% year-over-year in June, reaching 1.13 billion, with ChatGPT accounting for over 80% of these referrals.
• New Tooling
  • Cohere launched Command A Vision, an open-weight vision-language model for business.
  • ByteDance’s Seed-Prover proved five of six problems at IMO 2025.
  • Manifest AI open-sourced Power Attention,
  • v0 Platform API now provides a text-to-app interface for web app creation.
  • Spinach for meeting summaries.
  • Harvey for legal contract review.
  • FastWan for 5-second video generation.
    

🍞 Microsoft Says These Jobs Are Toast!

Microsoft’s analysis of 200K Copilot chats confirms OpenAI’s 2023 predictions about AI’s job impact were scarily accurate! Jobs requiring communication and information-sharing are most affected, while those needing physical touch or manual labor remain safest. This means specialists who master skills AI can’t copy, work with AI, build strong reputations and focus on real problem-solving will thrive.

Key Takeaways:

• AI’s Impact on Jobs: AI is hitting communication-heavy roles the hardest, but physical and manual jobs are relatively safe. The key is to adapt and collaborate with AI, natch.
• Freelancing & Entrepreneurship: AI tools empower individuals to do more, potentially increasing opportunities for skilled freelancers and entrepreneurs, while also intensifying competition.
• Prompt Engineering: The reporting highlights the importance of effective prompt engineering and verification, emphasizing that AI still needs human oversight to produce quality work.

Sources

• The Internet
• AI Secret
• The Rundown AI
• The Neuron