Hey friend,

Today, we’ll explore

• How To Create Viral LinkedIn Content Using AI

• New AI models ERNIE 4.5 And X1

• The OpenAI Agent SDK

Let's dive in.

Tutorial: How To Create Viral LinkedIn Content Using AI

I published a tutorial this week. It shows how I use AI as a personal copywriting assistant for LinkedIn. The video explores creating content with AI without losing your authentic voice or ideas. The principles and techniques shown here work for any AI writing.

Check out the full tutorial here:

AI Highlights

Here’s my pick of AI news this week:

1. Baidu Launches ERNIE 4.5 And X1 Models

Baidu has released two advanced AI models - ERNIE 4.5 and ERNIE X1 - making them freely accessible to individual users via the ERNIE Bot platform. The company claims ERNIE 4.5 outperforms GPT-4.5 at just 1% of the price, while ERNIE X1 matches DeepSeek R1's capabilities at half the cost.

• ERNIE 4.5: A native multimodal foundation model that processes text, images, audio, and video with improved language understanding and coding skills.

• ERNIE X1: Baidu's first deep-thinking reasoning model excels at complex tasks like Chinese knowledge Q&A, literary creation, and autonomous tool use.

• Strategic expansion: Baidu plans to open-source the ERNIE 4.5 series beginning June 30, 2025 - potentially accelerating the global commoditisation of advanced AI.

2. OpenAI Launches Comprehensive Agent-Building Toolkit for Developers

OpenAI has unveiled a suite of tools to help developers build AI agents that can independently accomplish tasks. The release includes the new Responses API, built-in tools for web search, file search, and computer use, plus an open-source Agents SDK for orchestrating workflows with enhanced tracing capabilities.

• New API: The Responses API combines Chat Completions and Assistants capabilities into a unified platform that supports multiple tools in a single call, with plans to eventually replace the Assistants API (sunset targeted for mid-2026).

• Tools: Built-in tools include web search (with GPT-4o and GPT-4o-mini achieving 90% and 88% accuracy on SimpleQA), file search with vector storage capabilities, and computer use preview that achieved 38.1% success on OSWorld tasks.

• Agent SDK: The open-source Agents SDK simplifies multi-agent workflow orchestration with features for intelligent handoffs between agents, configurable safety guardrails, and comprehensive tracing tools—companies like Coinbase and Box have already leveraged it to build functional agents quickly.

3. Google Launches Gemma 3: Advanced AI Models for Single-GPU Deployment

Google DeepMind has introduced Gemma 3, a collection of lightweight, state-of-the-art open AI models derived from Gemini 2.0 technology. These models are designed to run efficiently on single GPUs or TPUs while outperforming larger competitors like Llama3-405B in human preference evaluations.

• Efficient scalability: Gemma 3 comes in multiple sizes (1B, 4B, 12B, and 27B parameters) with multimodal capabilities for processing text, images, and short videos, a 128K token context window, and support for over 140 languages.

• Resource optimisation: The 27B model achieves exceptional performance on the LMArena leaderboard despite requiring only a single GPU, compared to competitors needing up to 32 GPUs for similar or lower performance scores.

• Built-in safety features: Google also released ShieldGemma 2, a 4B parameter image safety checker that provides built-in content filtering across three categories (dangerous content, sexually explicit material, and violence) to support responsible AI deployment.

Content I Enjoyed: The Future Of AI

In a conversation at the Council on Foreign Relations, Dario Amodei, CEO of Anthropic and former VP at OpenAI, discussed the future of AI, its risks, and opportunities, focusing on Anthropic's approach and the global implications of advanced AI development.

One of the most interesting predictions: “What we are finding is we are not far from the world—I think we’ll be there in three to six months—where AI is writing 90 percent of the code. And then in twelve months, we may be in a world where AI is writing essentially all of the code”

With the advent of AI, more code will be produced than ever before. Much of this code will be created by people without a software background. This means that a lot of code will be generated by AI, but it doesn’t necessarily mean that having more AI-generated code will reduce the need for software engineers.

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