AIGuys Digest | Aug 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 Why Agentic Frameworks Are Not Sufficient, Context Engineering, and How to build agents from scratch.
• 🌐 AI Monthly News: Discover how these stories revolutionize industries and impact everyday life. GPT-5 underwhelming release, Claude code taking on Cursor, and Google’s Nano Banana winning the AI image generation race.
• 📚 Editor’s Special: This covers the interesting talks, lectures, and articles we came across recently.

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

Follow me on Twitter and LinkedIn at RealAIGuys and AIGuysEditor.

Latest Breakthroughs

Most people entering the AI space have no clue about how to create production-grade pipelines; they simply copy simple Agentic AI tutorials. The agentic AI hype is mind-boggling. Through this article, I will try to show you why you should discard all the agentic frameworks. There are a few good frameworks like DSPy, but most are just pure fluff.

We will learn how to build better AI workflows without frameworks, what the different superior architecture patterns are. And most importantly, what are the specific anti-patterns to avoid?

Leave Agentic AI Frameworks And Build Agents From Scratch

Let’s be honest with ourselves, most agentic AI products and projects are simply trash. They are good for learning the Agentic AI frameworks, nothing more than that. Most AI devs are completely lost within these Agentic frameworks and somehow can’t see that they are overengineering their systems.

But here’s my question: why are people falling for the hype, even the experienced devs? The reasoning behind this is multifold. It is not just that these frameworks are quite new, so people are still figuring them out, but it has more to do with the economic viability and misappropriating the use case and the technology needed to solve those issues.

Agentic AI Workflows Are Seriously Broken

Prompt Engineering is Dead, it’s time for Context Engineering.

Do you remember a year or so ago, when companies were offering up to 300K USD for prompt engineering? We don’t see these posts or the company anymore. Prompt engineering was kind of hacky from the start. It was the fancy way to make everyone believe that they were engineers. But now things have changed and have taken the approach of traditional software engineering with the power of LLMs. So, let’s look at the ultimate guide on how to build real agentic workflows that actually scale and don’t break in production.

Context Engineering Over Prompt Engineering

AI Monthly News

GPT-5 underwhelming release

OpenAI officially launched GPT-5, their new flagship AI model. It features faster responses, significantly higher factual accuracy, and a massive 256,000-token context window, allowing it to process entire novels or extremely long documents in a single prompt. For the first time since 2019, OpenAI also released two open-weight models, GPT-OSS 120B and 20B, a strategic move to compete with the growing open-source community.

GPT‑5 is a unified system with a smart, efficient model that answers most questions, a deeper reasoning model (GPT‑5 thinking) for harder problems, and a real‑time router that quickly decides which to use based on conversation type, complexity, tool needs, and your explicit intent (for example, if you say “think hard about this” in the prompt). The router is continuously trained on real signals, including when users switch models, preference rates for responses, and measured correctness, improving over time. Once usage limits are reached, a mini version of each model handles remaining queries. In the near future, we plan to integrate these capabilities into a single model.

But in reality, it got very mixed reviews; it was not following the instructions properly after 3–4 messages. It was definitely not a behemoth; it was promised to be.

Blog: Click here

Claude Code is taking on Cursor

A powerful and intelligent coding assistant, Claude Code is designed to understand your entire codebase through agentic search, eliminating the need for manual context selection. It can make coordinated changes across multiple files, streamlining complex development tasks. Specifically optimized for code understanding and generation, it leverages the power of Claude Opus 4.1.

The tool integrates seamlessly into your existing workflow, living right inside your terminal or as a plugin for VS Code and JetBrains IDEs, which means you never have to switch contexts. It further enhances its capabilities by leveraging your existing test suites and build systems. You remain in complete control, as the tool never modifies files without your explicit approval. It adapts to your specific coding standards and patterns, and its configurable nature allows you to build on its SDK or run it on GitHub Actions, providing flexibility and control over your development process.

Announcement: Click here

Google’s Nano Banana is winning the AI image generation race

“Nano Banana” is the codename for Google’s new and highly advanced AI image generation and editing model, officially named Gemini 2.5 Flash Image. It was rolled out in the Gemini app and made available to developers in August 2025.

The core breakthrough of Nano Banan is its ability to maintain consistency and identity across multiple edits. This solves a major problem with previous AI image tools, where a character or subject’s appearance would often change with each new prompt.

Here are the key features and capabilities of Nano Banan:

• Character and Subject Consistency: It’s designed to ensure that people, pets, or objects in an image look consistent across a series of edits. This means you can change a person’s hairstyle, a pet’s costume, or the background of a product photo, all while preserving the original subject’s likeness.
• Prompt-Based Editing: It allows for highly precise and targeted edits using natural language. You can simply describe what you want to change, such as “remove the background and replace it with a forest,” “add a bookshelf to the wall,” or “make the person’s expression happier.” This makes advanced editing accessible to users without professional skills.
• Multi-Image Fusion: The model can understand and merge multiple images into a single, cohesive new image. This allows for tasks like placing a new object into a scene or blending different photos together.

Try it here: Click here

Editor’s Special

• The State of AI Research (talk given at UANL, Monterrey, Mexico): Click here
• The Only Trait for Success in the AI Era — How to Build It | Carnegie Mellon University Po-Shen Loh: Click here
• ’We have to stop it taking over’ — the past, present and future of AI with Geoffrey Hinton: 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