Welcome back, AI Enthusiasts!
In today’s AI summary rundown:
Reports say Apple is looking at using Google’s servers to support a future version of its AI‑powered Siri, which could help handle the heavy compute demands of the upgraded assistant.
This move aligns with Apple’s earlier plan to use Google’s Gemini AI models as part of its “Apple Intelligence” strategy. A custom version of Gemini would enhance Siri’s capabilities.
Apple’s Private Cloud Compute, the company’s in‑house AI backend, hasn’t been heavily used so far, reportedly around 10% of capacity, which may be why Apple is exploring external infrastructure.
Apple has long emphasized privacy and on‑device processing for Siri. While a shift to Google servers could help performance, it raises questions about how Apple will maintain its privacy commitments.
This development reflects the broader pressure on Apple to catch up in AI, especially as competitors like Google and Amazon offer increasingly capable assistants. Using Google’s infrastructure and models could speed up Apple’s AI progress.
Read time is 4 min..
AI Insights
Anthropic Makes Claude’s Memory Free and Adds an Import Tool to Ease Switching
Memory Now Available for Free Users
Claude’s persistent memory feature, which remembers details like past conversations and preferences, is now available to users on the free plan, not just paid subscribers.
New Import Tool Lets You Bring Context From Other AI Bots
There’s now a dedicated tool on claude.com that helps users import data from other chatbots like ChatGPT or Google’s Gemini. This helps users carry over context and history into Claude without starting from scratch.
How the Import Process Works
Users run a special prompt in their previous AI assistant to extract remembered context, then paste the resulting output into Claude’s importing feature, letting Claude adopt that history.
Competitive Push to Attract AI Switchers
This update is part of a broader effort by Anthropic to draw users away from rival AI chatbots by removing the friction of rebuilding preferences and conversation context when trying a new service.
Comes as Claude Gains Traction
Claude’s popularity has been rising thanks to other features like Claude Code, Claude Cowork, and the launch of newer models (e.g., Opus 4.6 and Sonnet 4.6) that handle complex tasks better.
AI Training: AI Tutorial of the Day
Reframe Anxiety as Prediction Error Overload
Here’s an engaging experiment to try on thareja.ai. Instead of explaining anxiety like a psychological abstract, treat it as a computational problem: the brain is constantly predicting the future, and anxiety is what happens when those predictions are overloaded with conflicting signals.
Think of it like a machine learning model being fed too many noisy inputs at once.
Start a New Chat
Open thareja.ai and create a fresh conversation.
Clean context matters here, no prior prompts about mental health or AI bias. You want the model to think fresh.
Switch Your AI Model
Click the (+) icon next to Automatic.
Choose GPT-4o for this experiment.
Why GPT-4o?
Because it excels at:
Mapping abstract human concepts to technical frameworks
Keeping explanations logical and structured
Offering analogies that are accurate and relatable
Try This Prompt & Observe the Output
Prompt Used:
“Explain anxiety as if it were a prediction error overload in a machine learning system. Include examples of how the brain’s predictions fail and how this relates to everyday stress. Keep it simple, visual, and insightful.”
Model Used:
GPT-4o
AI Response (Excerpt)
“Imagine your brain as a constantly learning AI, trying to predict every possible outcome of your day. When your predictions are mostly correct, life feels smooth. But when new inputs are noisy, conflicting, or overwhelming, your prediction errors spike, like a model constantly being corrected. That spike triggers physiological responses we label as anxiety: racing heart, tension, and restless thoughts. Essentially, anxiety is the brain’s signal: ‘Too much uncertainty, too many surprises.’”
Why This Experiment Works
Framing anxiety as a computational problem lets you:
See the mechanics behind emotional responses
Use AI to create metaphors that clarify difficult concepts
Explore ways to reduce overload (analogous to training a model on cleaner, smaller data)
Happy prompting!
Exclusive Member Deal - 20% OFF
Superpower AI Bundle Access 50+ Major LLMs
One Subscription. $20/mo $16/mo with code SUPERPOWER20
AI-Generated Image of the Day
Prompt: A flat lay of makeup products on a marble countertop: lipsticks, eyeshadow palettes, foundation bottles, brushes, and blushes, arranged aesthetically with soft lighting, vibrant colors, and a realistic, high-resolution style.
Tip: the more specific the better
Choose the model on thareja.ai that organizes your thoughts and shapes your idea clearly. Edit until every sentence has purpose. When the concept feels focused and deliberate, let Nano Banana turn that clarity into visuals designed to perform, not just look good.
Meme of the Day
Question of the Day
Which technology is commonly used to accelerate AI model training?
That’s it for today’s news in the world of AI!
If you have anything interesting to share, please reach out to us by sending us a DM on Twitter: @dthareja or email me at [email protected]. How was today's newsletter?
Feedback helps us improve!
Thanks for reading. Until next time!
p.s. if you want to sign up for this newsletter or share it with a friend or colleague, you can find us here.