The Rundown AI / Articles / AI / Meta's smart glasses get neural upgrade

Meta's smart glasses get neural upgrade

PLUS: Patients control AI and robotics with thought

Zach Mink

September 18, 2025

Good morning, AI enthusiasts. Meta just took smart glasses to the next level at Meta Connect, revealing three new products — including a Neural Band that shifts control from your voice to your thoughts.

With the ability to control your glasses through subtle muscle signals (and no awkward voice commands), Meta is building a future where your phone never leaves your pocket.

ICYMI — Rowan (CEO at The Rundown) sat down for an exclusive Q&A with CEO Mark Zuckerberg ahead of Meta Connect. Watch the full interview on YouTube.

In today’s AI rundown:

Meta’s smart glasses get a neural upgrade
OpenAI, Google models take gold at ICPC contest
Build agents, automations using Replit Agent 3
AI forecasts patient risk for 1,000+ diseases
4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

🕶️ Meta's smart glasses get a neural upgrade

Image source: Meta

The Rundown: Meta just revealed three smart glasses at Meta Connect, including Ray-Ban Display glasses that pair with a Neural Band to detect muscle signals before you visibly move, an athlete-focused Oakley line, and the Ray-Ban Meta Gen 2.

The details:

Display glasses use the Neural Band to detect electrical muscle signals, enabling messaging, navigation, and more via subtle finger movements.
The Neural Band will also learn and personalize to the user’s unique patterns, making controlling the interface nearly imperceptible.
Gen 2 doubles battery life to 8 hours while adding 3K Ultra HD video recording and conversation focus to amplify voices in noisy environments.
The Oakley Meta Vanguard targets athletes with 9-hour battery life, water resistance, and Garmin connectivity for real-time performance metrics.

Why it matters: Meta is solving two problems that have plagued other smart glasses: making wearables people actually want to wear, with controls that work without awkward voice commands. With the Neural Band detecting intention before movement even occurs, this is potentially the start of tech that responds to thought over touch.

TOGETHER WITH SANA

💡AI agents for instant insights

The Rundown: Imagine having a personal data analyst at your fingertips. Sana Agents lets anyone upload spreadsheets, analyze trends, and generate charts in plain English — helping companies like PE-firm Norvestor make decisions 2x faster with secure, enterprise-ready analytics for finance, sales, and marketing.

Use Sana Agents to:

Instantly analyze spreadsheets and files
Empower every team with self-serve reporting
Ask questions in plain English — get SQL, charts, and insights from your databases

Request a demo today.

OPENAI & GOOGLE

🥇OpenAI, Google models take gold at ICPC contest

Image source: OpenAI / Google

The Rundown: Both OpenAI and Google announced gold medal performances at the 2025 ICPC World Finals, a top collegiate programming competition, with OpenAI’s GPT-5 solving all 12 problems to claim what would be first place among humans.

The details:

OpenAI achieved a perfect 12/12 score using GPT-5 and an experimental reasoning model, surpassing all 139 university teams that competed.
Google’s Gemini 2.5 Deep Think finished with 10/12 correct to earn gold-medal status, notably solving one problem that stumped every human competitor.
Both companies competed under official contest conditions with identical constraints as human participants, with the top human team scoring an 11/12.
The achievement comes after OpenAI and Google’s gold-level performances at the IMO, OAI’s IOI gold, and a silver finish at the AtCoder Finals.

Why it matters: Competitive programming has been solved internally, and there’s no going back for humans to regain control of the leaderboards. While AGI hasn’t been declared (and is largely subjective), it’s clear that models have already reached superhuman abilities in more than a few domains — and are only getting better.

AI TRAINING

🎬 Create professional marketing videos using AI agents

The Rundown: In this tutorial, you will learn how to use Claude Code to automatically build complete video projects using frameworks like Remotion, turning simple prompts into marketing videos with animations and professional effects.

Step-by-step:

Type in your terminal: npm install -g @anthropic-ai/claude-code
Navigate to your project folder and type claude to activate the AI agent
Prompt: “Create a premium marketing video for [your company] using Remotion”
Refine with follow-ups and run npm run start to preview your video locally

Pro Tip: Add your own images, logos, and UI screenshots to the project folder before prompting, and the AI will incorporate them into professional video sequences.

PRESENTED BY IBM

🔎 Investigate IT incidents fast

The Rundown: IBM Instana’s latest feature, Intelligent Incident Investigation, helps teams resolve incidents fast using agentic AI.

IBM Instana’s AI-powered observability:

Uses agentic AI for fast, autonomous investigations
Helps reduce manual troubleshooting with smart automation
Delivers remediation steps and documentation automatically
Helps teams act quickly—even in high-stress, 2 am moments

Learn more in the DevOps Guide to Observability.

AI RESEARCH

🔮 AI forecasts patient risk for 1,000+ diseases

Image source: Reve / The Rundown

The Rundown: European researchers just developed Delphi-2M, an AI system that analyzes medical records to calculate individual disease risks across more than 1,000 conditions up to 20 years into the future.

The details:

The model studied health data from 400K U.K. patients, learning patterns from doctor visits, hospital stays, and lifestyle habits to spot early warning signs.
Delphi-2M matched or exceeded single-disease models while simultaneously reporting probabilities for 1,258 conditions, including cancer and diabetes.
Researchers verified accuracy by having the AI predict diseases for patients with already known health outcomes, tested on 1.9M Danish records.

Why it matters: While these are probabilities for medical outcomes, a predictor that works across over 1,000 conditions helps show how different conditions connect and influence each other, and provides a more proactive approach to health than the current reactive treatments often found across the medical world.

QUICK HITS

🛠️ Trending AI Tools

💵 AP2 - Google’s new open protocol for agentic transactions
🧊 Hunyuan3D 3.0 - Tencent’s upgraded 3D model
⚙️ Orchids - AI full-stack engineer for building prototypes, apps, and websites
📚 Tongyi DeepResearch - Alibaba’s open-source SOTA research agent

📰 Everything else in AI today

World Labs launched Marble, a beta platform that generates explorable, persistent 3D worlds from text or image prompts.

OpenAI and Apollo Research published new data on scheming behaviors across AI models, developing new training methods to reduce the deceptive actions by 30x.

China’s internet regulator banned major tech firms, including ByteDance and Alibaba, from purchasing Nvidia’s AI chips, pushing the use of domestic alternatives.

Elon Musk posted on X that he believes Grok 5 has “a chance of reaching AGI”, saying the next-gen model will begin training in a few weeks.

Zoom introduced AI Companion 3.0, featuring the ability to streamline meetings, create custom AI agents, use photorealistic avatars, and more.

AI models are becoming too smart for humans to train, with experts reportedly struggling to create tasks difficult enough for OAI’s advanced models.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Anonymous in New Zealand:

"AI helps me all day long. From responding to emails in a more polite, professional manner, creating ad copies for the properties I'm listing, to coming up with creative ways of writing my weekly newsletter — even reading architectural plans and researching the market for the purpose of writing appraisals."

How do you use AI? Tell us here.