AIGuys Digest | Sept 2025

🌟 Welcome to the AIGuys Digest Newsletter, where we cover State-of-the-Art AI breakthroughs and all the major AI news🚀. Don’t forget to check out my new book on AI. It covers a lot of AI optimizations and hands-on code:

Ultimate Neural Network Programming with Python

🔍 Inside this Issue:

• 🤖 Latest Breakthroughs: This month, it is all about Large Action Models for robotics, Cursor or Claude Code limitations, and Are energy based transformers the future?
• 🌐 AI Monthly News: Discover how these stories revolutionize industries and impact everyday life. SORA 2, Claude Sonnet 4.5 taking on Coding agents, and Google’s Gemini 1.5 for robotics.
• 📚 Editor’s Special: This covers the interesting talks, lectures, and articles we came across recently.

Let’s embark on this journey of discovery together! 🚀🤖🌟

Press enter or click to view image in full size

Follow me on Twitter and LinkedIn at RealAIGuys and AIGuysEditor.

Latest Breakthroughs

LLM has evolved to Large Reasoning models and now we are moving towards Large Action Models.

If I remember correctly, Jensen Huang (CEO of Nvidia) once talked about large action models. There is only so much AI can learn from text. Think about it for a moment, how much of your understanding of the world comes from reading only. Most of it is actually from the actions you do in the world, and now we are trying to do the same thing for the large language models. Large action models are the new trend in AI. If we want AI to properly operate in the world, it needs to learn the physical nature of the world as well.

What’s next for AI reasoning models? Action Models

Why don’t you see a 10x or 100x productivity boost with vibe coding?

This article shows exactly what the actual gain is with these tools, based on the research from Stanford.

📢 Table Of Contents
➡️ Software 1.0, 2.0, and 3.0
➡️ Why Cursor Is So Successful (Partial Autonomy and Human-AI Cooperation)?
➡️ AI Code Editor vs Coding Agents
➡️ Cursor Vs. Claude Code
➡️ Why Are We Not Seeing A 100x Productivity Boost?
➡️ My Personal Experience With AI Code Editors

Why Cursor or Claude Code is not enough?

It has been so long since the Transformers were invented, and still, there is no new major architecture that has been able to beat it fair and square. But what if we can incorporate some interesting idea and combine it with the power of Transformers? That’s where the Energy-based Transformers come in. Most AI models are quite analogous to System I thinking and are incapable of deep, long, nuanced thinking. But what if we can change that and make the system think deliberately about certain problems? This is the hypothesis we are going to discuss.

Energy-Based Transformers are Scalable Learners and Thinkers

AI Monthly News

Sora 2: The New Standard for AI Video Realism

OpenAI’s September release of Sora 2 significantly raised the bar for generative video, moving the technology closer to a true world simulator. The major technical breakthrough over its predecessor was the implementation of accurate physical modeling, allowing generated objects to obey real-world dynamics like gravity, momentum, and collision physics.

Furthermore, Sora 2 introduced native, synchronized audio, creating dialogue and sound effects that perfectly match the on-screen visuals, eliminating the need for post-production audio stitching. This model was also integrated into a new Sora iOS app, which includes a viral “Cameo” feature allowing users to insert a verified, consistent likeness of themselves into generated scenes, bridging the gap between a research tool and a consumer social platform.

Official release: Click here

Claude Sonnet 4.5: The Autonomous Agent for Software

Anthropic positioned its Claude Sonnet 4.5 model as the world’s most capable AI for autonomous coding and agent workflows. This release focused heavily on endurance and reliability, with enterprise trials demonstrating the model could maintain focus and coherence while performing complex software tasks for over 30 continuous hours.

Sonnet 4.5 achieved state-of-the-art performance on industry benchmarks like SWE-bench Verified and OSWorld, indicating a vastly improved ability to not only write code but also to autonomously interact with and control computer systems. To support these long-running tasks, Anthropic introduced the Claude Agent SDK and enhanced its internal context management and tool orchestration, making the model ideal for enterprise tasks in cybersecurity, finance, and large-scale code refactoring.

Official Blog: Click here

Gemini Robotics 1.5: The Thinking Robot

Google DeepMind introduced a groundbreaking two-model architecture for robotics that effectively separates a robot’s “brain” from its “hands.” The system consists of two specialized models: Gemini Robotics-ER 1.5 (the Embodied Reasoning model) handles the high-level planning, spatial understanding, and tool-use (such as looking up information on Google Search to inform a task).

This model then hands off step-by-step instructions to Gemini Robotics 1.5 (the Vision-Language-Action model), which is solely responsible for converting those plans into precise, physical motor commands. This decoupled system provides unprecedented motion transfer capability, allowing skills learned on one type of robot to be instantly applied to completely different hardware, making it a major step toward creating truly general-purpose intelligent robots capable of reasoning about the physical world.

Read More: Click here

Editor’s Special

• Will AI outsmart human intelligence? — with ‘Godfather of AI’ Geoffrey Hinton: Click here
• Richard Sutton — Father of RL thinks LLMs are a dead end: Click here
• The Catastrophic Risks of AI — and a Safer Path | Yoshua Bengio | TED: Click here

🤝 Join the Conversation: Your thoughts and insights are valuable to us. Share your perspectives, and let’s build a community where knowledge and ideas flow freely. Follow us on Twitter and LinkedIn at RealAIGuys and AIGuysEditor.

Thank you for being part of the AIGuys community. Together, we’re not just observing the AI revolution; we’re part of it. Until next time, keep pushing the boundaries of what’s possible. 🚀🌟

Your AIGuys Digest Team