🧠 How I made my AI remember, reason, and redesign my data

🔥🔥🔥 Three big updates

1. Google adds Data Tables to NotebookLM

NotebookLM, Google’s AI research notebook, now supports Data Tables, letting you upload spreadsheets and query them directly. You can ask, “Which products performed best in Q4?” or “Summarize key trends by region” — and it’ll compute and explain results instantly.

It’s a small but major shift: NotebookLM is no longer just a reading companion; it’s starting to look like an AI-powered data analyst.

2. Anthropic introduces Bloom — an open-source evaluation framework

Bloom is Anthropic’s new framework for testing how AI behaves — not just how well it performs. It lets researchers run large-scale behavioral evaluations, measuring things like self-preference, bias, and ethical consistency across thousands of interactions.

Think of it as an open benchmark for AI behavior, built to help the community study alignment transparently instead of behind closed doors.

3. NVIDIA’s Nemotron 3 redefines long-context reasoning

NVIDIA released Nemotron 3, a hybrid Mamba-Transformer-MoE model that handles massive token contexts while cutting compute costs. It’s designed for agentic use cases — systems that need memory, reasoning, and continuous context flow.

For developers building persistent AI agents, Nemotron 3 feels like an early glimpse of the next generation of scalable reasoning models.

If you’re building internal agents or research systems, this one’s worth exploring — Mistral’s models continue to punch well above their size.

🔥🔥 Two tools worth trying

Manus x NanoBanana
Manus just added NanoBanana-powered AI decks that are fully editable. Generate a presentation with AI, then tweak layouts, themes, and content inside the same interface. Perfect for teams who need pitch-ready slides — fast.

StickerBox
Upload a photo or choose a theme, and StickerBox’s AI will turn it into high-quality, personalized vinyl stickers. It’s a surprisingly fun use of generative models — and yes, a solid last-minute holiday gift idea.

🔥 Things You Didn’t Know You Can Do With AI

Transfer your AI memory from one AI tool to another

If you’ve built months of memory in ChatGPT — tone, style, project context — and want to try Gemini without losing it, here’s how to transfer it seamlessly:

1. Start with ChatGPT: Ask it to “Summarize everything you know about me.” This gives you a clean snapshot of your AI memory — your tone, goals, and recurring themes.
   

2. Bring it into Gemini: Go to Gemini’s Settings → Memory and paste that summary. Gemini will instantly “learn” your preferences and context.
   

3. Set up project-specific Gems: For ongoing work, create a new Gem inside Gemini. Add related docs or notes — it’ll act as a mini version of Gemini focused just on that project.
   

4. Make it universal: If you use multiple AI tools, upload your ChatGPT summary to Prompt Genie. It’ll automatically inject your memory into any AI you use — so you never have to reintroduce yourself again.

Watch the demo here.