AI Tutor + Human Tutor = Gains

Hey human,

You may have noticed a change in the header this week. We’re rebranding to better support schools, and are now called ‘AI for Schools’ and to you can find us on the web at https://forschools.ai. We hope you like it!

Now, here’s what’s been happening in education / AI this week:

📚 AI+education news

• Eedi + AI > A new exploratory trial from Eedi and Google DeepMind offers early evidence that human-in-the-loop AI tutoring can match high-quality one-to-one tuition while beginning to solve the scalability problem. In UK secondary maths classrooms using Eedi’s platform and Google’s LearnLM, a human–AI tutoring team was as effective as expert human tutors at fixing errors and addressing misconceptions, and outperformed them on “knowledge transfer”.

🌍 Wider AI updates

• Claude Use Cases > The team at Anthropic have published some use cases on their website which I have not doubt are sure to inspire readers. There are plenty of inspiring ideas for both educational and personal purposes that are worth checking out.

• Notebook’s Deep Research > A favourite use case of AI outside of classic ‘chatbot’ use case. Now NotebookLM can search for additional sources using its own Deep Research function. I suspect this is the same model that powers Gemini, but it’s good to see this come directly to Notebook so we can avoid going between the two of them.

🎯Top Tip

• Mastering GPT’s Memory: A Quick Guide

  GPT’s memory features help you reduce repetition and maintain continuity across work. This summary captures what each setting does and doesn’t do so you can use them with confidence. Go to Personalization and scroll down until you see ‘Memory’.

  

  The Memory Settings

  Manage Memories

  A simple list of what GPT currently remembers about you (e.g., preferences, ongoing projects). You can review, edit, or delete items at any time.

  Important: GPT only saves something if you explicitly tell it to or if it clearly helps future conversations. You stay in control.

  Reference Saved Memories (On/Off)

  When on, GPT can use the memories stored in your list. When off, GPT ignores them.

  Turn on when you want:

  • Consistent tone and preferred formats

  • Faster starts to recurring tasks

  • Support that builds on your ongoing work

  Turn off when you want:

  • Every conversation treated as new

  • No long-term personalisation

  Reference Chat History (On/Off)

  This affects only your current conversation. With this on, GPT uses earlier messages in the same thread to stay coherent. With it off, GPT treats each message independently.

  Turn on for multi-step tasks or discussions.

  Turn off if you’re switching topics often and want clean boundaries.

  Choosing the Right Setup

  • Both on: Best for ongoing projects, planning, and writing workflows.

  • Both off: Best for high-privacy use or completely isolated tasks.

  • Chat History on + Saved Memories off: Keeps coherence within a task but avoids long-term personalisation.

Till next week.

Mr A 🦾