MIT’s SCIGEN Steers GenAI Toward Quantum Materials 🧪

Generative Models Get Rules For Exotic Lattices

Author

Samantha Vale
September 22, 2025

In partnership with

It’s go-time for holiday campaigns

Roku Ads Manager makes it easy to extend your Q4 campaign to performance CTV.

You can:

  • Easily launch self-serve CTV ads

  • Repurpose your social content for TV

  • Drive purchases directly on-screen with shoppable ads

  • A/B test to discover your most effective offers

The holidays only come once a year. Get started now with a $500 ad credit when you spend your first $500 today with code: ROKUADS500. Terms apply.

Hey there, Tech Trailblazers! 🤖✨

MIT researchers have introduced SCIGEN, a system that steers generative models toward quantum-ready lattices such as Kagome and Archimedean rather than defaulting to conventional stable structures.

Published in Nature Materials, the study shows that SCIGEN generates over 10 million candidates, screens one million for stability, simulates 26,000, and identifies 41% with magnetic potential. Two compounds, TiPdBi and TiPbSb, were ultimately synthesized.

📰 Upcoming in this issue

  • 🔬 MIT’s SCIGEN Steers GenAI to Quantum-Ready Materials

  • 🎥 How To Make Viral Miniature Videos With Gemini AI

  • 🔋 iOS 26’s AI Boosts iPhone Battery Life

🔬 MIT’s SCIGEN Steers GenAI to Quantum-Ready Materials

The tool injects geometric design rules into diffusion models to produce candidate lattices, such as the Kagome and Lieb lattices. Researchers synthesized two predicted compounds, pointing to faster routes toward quantum breakthroughs.

Key Takeaways:

  • 🎯 Constraint-Guided Generation: SCIGEN adds geometric rules to diffusion models, steering them toward Kagome, Lieb, and Archimedean lattices with unusual quantum properties.

  • 🧪 From Millions to Materials: Out of 10 million candidates, researchers simulated 26,000 and synthesized two compounds, TiPdBi and TiPbSb, whose properties closely matched predictions.

  • 🧲 Quantum Targets: The focus is on spin liquids, flat bands, and topological effects that may underpin quantum computing and novel magnetism.

  • 🛠️ Tool-Agnostic Code: Compatible with models such as DiffCSP, SCIGEN allows user-defined constraints, both chemical and functional, to guide future generations.

🎥 How To Make Viral Miniature Videos With Gemini AI

This tutorial shows how Gemini AI transforms everyday scenes into miniature dioramas in under a minute. It covers prompts, lenses, motion, and sound.

Key Takeaways:

  • 🧠 Prompt Formula: Include scale cues, lens choice, time of day, and tilt shift, referencing miniature dioramas and shallow depth of field for realism.

  • 🎬 Shot Design: Use slow push-ins, overheads, and parallax, adding props in the foreground to reinforce scale and mask AI artifacts.

  • 💡 Lighting and Texture: Request noon shadows or warm lamps, specify bokeh, dust, and subtle imperfections to mimic handcrafted detail.

  • 📱 Export and Posting: Render in 9:16 at high bitrate, keep clips short, and add foley with a single hook line to boost retention.

🔋 iOS 26’s AI Boosts iPhone Battery Life

Apple is rolling out Adaptive Power, an AI-driven feature that learns user habits to extend battery life. The upgrade is limited to devices running Apple Intelligence.

Key Takeaways:

  • 🤖 Adaptive Power Mode: AI adjusts brightness, background activity, and power settings in real time to extend daily uptime without manual input.

  • 📱 Device Eligibility: Available only on iPhones that support Apple Intelligence, leaving older models behind.

  • ⚙️ Learns Usage Patterns: After a week of observation, the system predicts when to conserve energy and when to maintain performance.

  • 🔋 Battery Health Support: Works with existing charge limiting and optimized charging features to reduce heat, preserve longevity, and stabilize endurance.

Why It Matters

Encoding quantum design rules directly into generative models can shorten discovery timelines and reduce wasted computation. By promoting higher-value candidates, labs can focus resources on experiments with better odds, avoiding the cost and reputational risk of chasing irrelevant results.

The authors emphasize that breakthroughs still require real synthesis and testing. Teams that balance disciplined generation with rigorous validation are more likely to turn a single promising structure into practical impact.

Until our next issue,

Samantha Vale
Editor-in-Chief
Get Nerdy With AI

P.S. Interested in sponsoring a future issue? Just reply to this email and I’ll send packages!

How was today's edition?

Rate this newsletter.

Login or Subscribe to participate in polls.