Hey friend,

This was another crazy week of AI updates. So many things to cover.

Today, we’ll explore:

• How to build a YouTube content researcher that finds viral video ideas using N8N

• Claude 4!!! : )

• Some of the many things released at Google I/O

• Google’s Agent Development Kit (ADK) support for Java

Let's dive in.

Tutorial: Automate YouTube Research With N8N And Airtable

Check out the video here.

After an unfortunate break (due to being highly disorganised), I published a tutorial this week showing you how to build your own YouTube content research assistant using Airtable and N8N.

The goal of this system is to research content and give you insight into high performing content ideas. The automation builds a database of all YouTube videos for any number of channels including likes, view count, and content multiplier.

AI News

Here’s my pick of AI news this week:

1. Google I/O 2025

Google released so many amazing things at I/O 2025, it’s hard to know where to start. Gemini 2.5 models with Deep Think reasoning, Imagen 4 for improved image generation, and Veo 3 for video capabilities, alongside a new "Gemini Ultra" premium subscription service priced at £249.99 monthly.

If you’re a developer interested in building AI Agents, it’s worth paying attention to the Agent Development Kit (ADK). They released support for Java here.

• Gemini 2.5 includes Flash and Pro variants with a new Deep Think reasoning mode, complemented by Imagen 4's enhanced text rendering capabilities and Veo 3's advanced video generation features.

• Google launched Flow, an AI filmmaking tool powered by Veo 3 with audio capabilities, expanding its creative suite that includes image remixing tool Whisk.

• AI Mode in Google Search has been released in the U.S., offering conversational search experiences with virtual try-on and agentic checkout features, powered by a custom version of Gemini 2.5.

2. Anthropic's Claude 4 (Massive Release)

The big news this week: Anthropic released Claude Opus 4 and Claude Sonnet 4, its next-generation AI models. These flagship models establish new benchmarks for coding capabilities and complex reasoning while maintaining Anthropic's commitment to safety.

What this shows to me is that Anthropic are doubling down on having the best AI models for coding and AI agents.

• Claude 4 models excel at analysing large datasets, executing long-horizon tasks, and multi-step reasoning, with specific optimisation for programming that has reportedly surprised even Anthropic's senior engineers with productivity improvements.

• Two variants have been released: Opus 4 (the most capable model) priced at $15/$75 per million input/output tokens, and Sonnet 4 (balancing capability and efficiency) at $3/$15 per million tokens.

• Both models are available through Anthropic's API via Amazon Bedrock and Google Vertex AI platforms, with Sonnet 4 accessible to all users while Opus 4 is limited to paying customers.

3. Github Copilot Agent

Github launched it’s Copilot agent. What I’m excited about is having AI integrated into the Github flow. Check out the docs here.

4. Mistral AI Launches Devstral

Mistral AI has launched Devstral, a new open-source LLM specifically designed for software engineering and coding. Released under the Apache 2.0 license, the model outperforms competitors like Google's Gemma 3 27B on coding benchmarks while running efficiently on consumer hardware.

• Devstral excels in "agentic" use cases, allowing developers to navigate codebases, edit multiple files simultaneously, and power intelligent software agents that interact with code and testing environments.

• The model can run locally on a single Nvidia RTX 4090 GPU or a Mac with 32GB RAM, eliminating the need for expensive cloud infrastructure for deployment.

Your Opinion Matters

What did you think of today’s email? Your feedback helps me create better emails for you!

Got more feedback or want me to cover a specific topic? Reply to this email and let me know.

Owain