ChatGPT Now Personalizes Search Using Your Memory

✨TodayOnAI’s Daily Drop

• ChatGPT Now Personalizes Search Using Your Memory

• Ben Goertzel’s Startup Wants to Replace Human Labor with AGI Robots

• OpenAI’s o3 Under Fire for Benchmark Discrepancy

• 💬 Let’s Fix This Prompt

• 🧰 Today’s AI Toolbox Pick

OPEN AI

ChatGPT Now Personalizes Search Using Your Memory

🚀 TodayOnAI Insight: OpenAI is quietly rolling out Memory with Search, a new ChatGPT feature that personalizes web search queries using remembered user preferences—marking a significant step toward more contextual and customized AI interactions.

🔍 Key Takeaways:

• New feature: Memory with Search rewrites web search prompts by incorporating past user details like dietary preferences or location.

• How it works: If memory is enabled, ChatGPT can adjust queries—e.g., turning “restaurants near me I’d like” into “good vegan restaurants, San Francisco.”

• Privacy control: The feature is only active when memory is enabled and can be disabled in ChatGPT’s settings.

• Gradual rollout: OpenAI hasn’t disclosed availability details, but users on X report early access.

• Strategic positioning: This move strengthens ChatGPT’s competitive edge against Anthropic’s Claude and Google’s Gemini, which also leverage memory.

💡 Why This Stands Out: OpenAI is pushing beyond static chatbot experiences by fusing search with persistent personalization. This not only boosts relevance but hints at a future where AI assistants proactively tailor outputs to individual users. Will memory-driven interactions become the new baseline for intelligent search?

AI

Ben Goertzel’s Startup Wants to Replace Human Labor with AGI Robots

🚀 TodayOnAI Insight: Apple is launching AI-generated App Store review summaries in iOS 18.4, now in public beta. Powered by Apple Intelligence, the summaries condense user reviews into short, AI-written blurbs, offering a quick overview of app sentiment—starting in the U.S. in English.

🔍 Key Takeaways:

• AI-Powered Review Summaries: Apple Intelligence uses LLMs to generate weekly-updated review summaries for apps with sufficient user feedback.

• U.S.-Only at Launch: The feature debuts in English in the U.S., with global expansion and language support planned later this year.

• Moderation Tools Included: Users can report issues by long-pressing a summary; developers can flag problems via App Store Connect.

• Broader AI Rollout: iOS 18.4 and iPadOS 18.4 also bring Apple Intelligence to more users, including in the EU, along with new notification, Siri, and Visual Intelligence features.

💡 Why This Stands Out: This update signals Apple’s deeper integration of generative AI across its ecosystem. Review summaries aim to streamline decision-making for users—but may also amplify the impact of fake or manipulated reviews. As with Amazon and Google, Apple’s move reflects a growing trend: turning raw customer feedback into digestible AI-driven insights.

OPENAI

OpenAI’s o3 Under Fire for Benchmark Discrepancy

🚀 TodayOnAI Insight: OpenAI’s new o3 model is under scrutiny after third-party benchmarks revealed a performance gap between its public release and earlier internal claims. While OpenAI’s December figures suggested a 25% success rate on a challenging math benchmark, independent testing from Epoch AI shows o3 scores closer to 10%.

🔍 Key Takeaways:

• OpenAI initially claimed o3 could solve over 25% of questions on FrontierMath; independent tests found ~10%.

• The high internal score was achieved using a more compute-intensive version than what’s now publicly available.

• Epoch AI’s evaluation used an updated FrontierMath set and acknowledged different test conditions.

• ARC Prize confirmed the public o3 is smaller and optimized for product use, not benchmark performance.

• OpenAI has since released o4-mini and plans to launch a more powerful o3-pro variant soon.

💡 Why This Stands Out: This episode highlights growing tensions around AI benchmarking transparency. As vendors push to impress with bold claims, differences between internal and public model versions are becoming harder to ignore. The question isn’t whether benchmarks are useful—but whether they’re being used responsibly.

✨ See how a simple prompt upgrade can unlock better AI output.

🔹 The Original Prompt

At first glance, this prompt might seem okay. But it's too broad — and that limits the quality of AI-generated results. Let’s improve it using prompt engineering best practices.

✅ The Improved Prompt

💡 Why It's Better

• Specifies the brand type (sportswear and fitness)

• Targets a clear audience (active lifestyle consumers)

• Includes relevant content pillars (training, gear, health)

• Aligns with marketing goals (SEO, engagement, loyalty)

🛠️ Learn how to adapt this prompt for SaaS, AI tools, dev teams & more →
Read the full PromptPilot breakdown

💡 Bonus Tool: Want to generate and master prompts instantly?
👉 Try PromptPilot by TodayOnAI (Free to use)

📰 More from the AI World

• ChatGPT is referring to users by their names unprompted, and some find it ‘creepy

• Former Y Combinator president Geoff Ralston launches new AI ‘safety’ fund

• Google’s latest AI model report lacks key safety details, experts sa

• OpenAI partner says it had relatively little time to test the company’s o3 AI model

🧰 Today’s AI Toolbox Pick

• 🥁Suno (Music Tool): Transforms your ideas into music.

• 😴AI Employe (Email Tool): Automates email-to-CRM/ERP data transfers like understanding emails, receipts, and invoices to gain hours back.

• 🐙OctoML (Cloud Tool): Delivers GenAI solutions running on the most efficient compute, empowering builders to launch AI applications.