📝 14/4/25: Claude’s New Search, Updated Prompt Bible, Google's New Video Generator

📌 This Week’s News Highlight: Claude's Faster and More Personalised Version of Deep Research

Anthropic has launched two major updates to Claude to rival ChatGPT's Deep Research:

• Research Mode (in beta on Max plans only in the US, Japan, and Brazil for now)
• Google Workspace Integration (now live on Pro and Max plans)

These features work together, allowing Claude to both search securely across your internal tools (including Google Docs, Sheets, Drive, and Calendar) alongside external web sources to generate structured, hyper-relevant and citation-backed answers.

This means it can:

• Pull relevant info from your own files and calendars to personalise responses
• Combine that with external research to support business decisions
• Handle open-ended questions like "What are the biggest risks in our GTM plan right now?"
• Do all of this without you constantly re-uploading files, context sharing or copy-pasting content

How is this different from ChatGPT's Deep Research I hear you ask?

Anthropic claims that it operates more 'agentically', i.e. it runs multi-step searches at once, decides what to explore next, and works through questions systematically. Anthropic says this makes it both more thorough and faster than ChatGPT for research-heavy tasks. Read more here.

🕵️ Founder, Anonymous

A healthtech founder uses Claude search to stay on top of a key company objective: improving user onboarding flow.

Using Claude’s new Google Workspace integration, they:
• Pull key insights from user interviews, product notes and meeting recaps stored in Google Docs.

With Research Mode, they:
• Layer in best practices and competitor tactics from across the web
• Get a concise summary of blockers, wins and suggested next steps

Result
A clear, up-to-date view of key customer insights, progress and what to do next, all without digging through files or searching the web!

💡 Thought of the Week

Have you become ruder to ChatGPT?
This viral TikTok is calling us out for how fast we’ve dropped the pleasantries with our robot sidekick. And tbh, I’ve caught myself skipping the “please” and "thank yous" too of late.

It may just be habituation, and warranted dehumanisation bias. Whatever it is, it’s fascinating how quickly we slip into taking it for granted.

Perhaps we’ll always be the ones barking the orders.
Or maybe the moment we stop being impressed and alert is exactly when we should start paying attention?

📚 Google’s Guide to Writing Better Prompts

Google recently released a (very detailed) 68-page prompt engineering guide to help you get better results from tools like Gemini, ChatGPT and Claude.

Why it matters
If you’ve ever typed a prompt and got something vague, messy or just wrong, this guide explains how to fix that.

So that you don't have to spend all weekend reading a Google handbook, I've trawled the internet and found this cool cheat sheet, which shows the 12 prompt types, how to use them, what they're best for and some use cases. Bookmark this one.

If you insist on being a hero, here's the full PDF.

⚡ Quick Tip

Create state-of-the-art videos with a simple prompt
Google just unveiled Veo 2, its most advanced AI video model yet. It turns simple text prompts into high-quality, stylised videos (with high quality output up to 4K, realistic motion, lighting and editing).

Prompt example: "Create a 30-second cinematic video that highlights our startup’s traction, mission and product. Use upbeat music and a confident tone. Show scenes of user growth, product in action and team collaboration. The startup is a B2B healthtech platform helping clinics streamline patient intake and follow-ups.

Perfect for mocking up pitch content, inspirational careers page videos, or testing creative ads without a full production team. Veo 2 is available to try for free for a month with Gemini Advanced - definitely think this is worth a try.

🔗 Sneaky Links

• AI-generated music accounts for 18% of all tracks on Deezer 🤯
• OpenAI’s new and smartest yet o3 and o4 models beat GPT-4o at reasoning, can use all tools within ChatGPT, and can solve problems by thinking with images (not just analysing them like 4o). Imo, GPT-4o still wins for everyday, all-round support but OpenAI has said these new models are a step forward in providing better AI Agents for companies
• Google launched their most powerful model yet, Gemini 2.5 Pro, but reports say its security checks are subpar, another blow for safety & transparency in the AI race
• Claude’s getting a voice. Anthropic is reportedly launching a new voice mode this month, with three personalities to pick from: Airy, Mellow and Buttery.
• And OpenAI is building a social app to rival Meta and X...

🎯 Where are you on your AI journey?

Take our quick quiz to find out which stage you're at - from Curious Newbie to AI Evangelist - and get tailored tips to uplevel.

💬 What did you think of this post?

😍 Loved it
😐 Meh
😞 Not for me

Share this post