Hey friend,

Welcome back. Today, we’ll explore

• Code with AI: 9 tips for using Claude Code

• Create with AI: How AI can review and improve your content

• A new video model from Midjourney

• ChatGPT’s new Record mode

• Google’s new voice search feature

Let's dive in.

Code With AI: 9 Claude Code Principles For Software Engineers

I’ve worked as a software engineer for nearly 20 years and Claude Code has been my obsession lately. If you’re not aware, Claude Code is an AI coding tool that runs in the terminal.

Here are 9 principles I’ve found work well for AI coding with Claude Code:

1. The "Fast Intern" Mental Model 

Start with this idea: Claude is like a brilliant intern with a perfect memory who works at superhuman speed. It's an incredibly powerful assistant, but it still needs clear direction and supervision from you, the senior developer. Always verify its work, as you are ultimately responsible for quality. This mindset is key to using it effectively.

2. Give It a Screenshot to Copy 

You can give Claude Code a visual target, like a UI design mock-up, simply by pasting the image into the chat. Then, ask it to write the code to build what it sees. It's a powerful way to translate visual designs into functional code, especially for front-end development.

3. Ask Questions About Any Codebase 

Use Claude as your personal guide to navigate new or complex projects. You can ask it plain-English questions like, "How does the user authentication work?" or "Explain the purpose of this file," and it will read the code to provide you with a clear, concise explanation, dramatically speeding up your onboarding.

4. Use One Claude to Write Code and Another to Review It 

To ensure high-quality output, set up a "pair programming" scenario. Instruct one Claude to write the initial code for a feature. Then, in a separate session, ask a second Claude to act as a code reviewer. It will analyse the first Claude's work for errors, suggest improvements, and help catch bugs, simulating a real-world quality assurance process.

5. Use It for the Full Software Lifecycle (Beyond Code) 

Expand your use of Claude beyond just writing code. It excels at tasks across the entire development process. Ask it to write clear PR descriptions, generate meaningful commit messages, or even create technical diagrams for your design documents using tools like GraphViz, as its contextual understanding is remarkably deep.

6. Make Claude Learn From Its Mistakes

When Claude makes a mistake or does something you don't like, don't just correct it for the moment. Explicitly tell it: "Update the CLAUDE.md file so you don't make that mistake again." This creates a powerful feedback loop, helping the AI learn your specific preferences over time.

7. Adopt a Test-Driven Development (TDD) Workflow 

A highly effective method for ensuring your code works is to have Claude write tests first. Describe the feature you want, and ask Claude to create the tests that will check if it's working correctly. These tests will initially fail. Then, simply instruct Claude to write the code required to make them all pass.

8. Tell It to "Think" Before It Writes Code 

For any complex task, you can significantly improve the result by telling Claude to plan its approach. Starting your prompt with a phrase like, "First, think step-by-step about how you will solve this," encourages it to spend more time analysing the problem, which often leads to a more logical and robust solution.

9. Interrupt and Redirect It at Any Time 

You are always in control. If you see Claude going in the wrong direction, you don't have to wait for it to finish. Simply pressing the Escape key will immediately interrupt its current task. This allows you to jump in, provide a quick clarification, and get it back on the right path without losing any of your previous work.

Create With AI: Using AI As A Content Reviewer

One of the most useful ways to use AI as a writer that I’ve found is to ask for feedback and a numerical rating.

AI can spot ways to improve any content - whether it’s social media posts or emails. This system becomes stronger when you provide clear guidance about what great content looks like.

Here’s a sample prompt to start that works in any LLM (Gemini, ChatGPT, Claude):

Act as an expert content strategist.

Review this social media post and rate it out of 10. 

Provide specific actionable suggestions to improve it. 

Context 
- Platform: LinkedIn
- Target Audience: Software Engineers
- Business Goal: Build authority
- Content Pillar: Code with AI

Here is the post: 

<POST>
{INSERT_POST_HERE}
</POST>

AI News

Here’s my pick of AI news this week:

1. Midjourney Launches V1 Video Model

Midjourney enters the competitive AI video market with its V1 model, transforming still images into animated videos and making professional video generation accessible to all users. The release marks a strategic expansion beyond static imagery, positioning Midjourney alongside OpenAI, Runway, and Adobe in the AI video space.

• Video Capabilities: Converts single images into four five-second video clips with customizable subject and camera movements, extendable up to 21 seconds through iterative generation

• Pricing and Access: Available immediately at $10/month subscription including 3.3 GPU hours, accessible via Midjourney's website and Discord platform

• Future Vision: Company aims to develop "real-time open-world simulations" with planned releases featuring enhanced creative controls, signaling ambitions beyond traditional video generation

2. OpenAI Launches ChatGPT Record Mode, Automating Meeting Transcription and Note-Taking for Pro Users

OpenAI has introduced Record mode for ChatGPT desktop, enabling Pro, Enterprise, and Edu users to automatically transcribe, summarize, and extract action items from meetings and voice recordings. The feature transforms ChatGPT into a productivity assistant that eliminates manual note-taking while providing intelligent analysis of recorded conversations.

• Recording Capabilities: Users can capture audio directly in the ChatGPT macOS app, which automatically transcribes conversations, generates summaries, extracts key points, and suggests follow-up actions or code snippets

• Productivity Impact: Eliminates manual note-taking during meetings and brainstorming sessions, with transcripts saved in chat history for future reference and repurposing, potentially replacing third-party transcription tools

• Availability: Rolling out since June 18, 2025, exclusively for ChatGPT Pro, Enterprise, and Edu subscribers on macOS desktop at no additional cost, with no announced timeline for Windows or mobile expansion

3. Google Launches Search Live Voice Feature, Enabling Natural AI Conversations in Mobile Search

Google introduces Search Live, transforming mobile search into conversational AI experiences through real-time voice interactions. This feature represents a significant shift toward hands-free, multitasking-friendly search that mimics natural human conversation while maintaining Google's search accuracy.

• Voice Interface: Tap the "Live" icon in the Google app to start voice conversations powered by custom Gemini AI, with seamless switching between voice and text modes plus full transcripts

• Multitasking Design: Continue conversations while using other apps, enabling truly hands-free information access for users on the go with trusted, web-grounded responses

• Access & Future: Available now for U.S. users via Google Labs AI Mode on iOS/Android, with upcoming visual query capabilities using phone cameras for real-time object identification

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