Apple chases Meta's AI glasses lead

PLUS: Create a professional headshot with Gemini 2.5 Flash

Good morning, AI enthusiasts. Mark Zuckerberg said glasses are the “ideal form factor” for personal AI, and Apple seems to be a believer.

The tech giant just reportedly scrapped its Vision Pro overhaul to go all-in on smart glasses that compete with Zuck’s successful Ray-Ban lineup. But with Apple’s AI issues already mounting, can they actually compete?

In today’s AI rundown:

  • Apple to join smart glasses race

  • Mira Murati's startup unveils first product

  • Create a professional headshot with Gemini

  • Google’s AI agent masters Minecraft via simulation

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

OPENAI

Image source: Nano Banana / The Rundown

The Rundown: Apple just canceled plans to overhaul its Vision Pro headset, according to a new report from Bloomberg — instead shifting focus toward AI smart glasses projects designed to compete with Meta's Ray-Ban lineup.

The details:

  • The company halted work on a lighter, cheaper Vision Pro variant planned for 2027, reassigning teams to fast-track development of several glasses designs.

  • A 2027 release will connect to iPhones without its own screen, with another version with an integrated display aiming to rival Meta's Display glasses.

  • The devices will reportedly lean on voice controls and AI features, with speakers, cameras, and health tracking powered by the upcoming Siri upgrade.

  • Meta expanded its smart glasses line in September with a new Display and Neural Band, an athlete-focused Oakley option, and Ray-Bans Gen 2.

Why it matters: 2023’s Vision Pro release feels like an eternity ago in the AI world, but the hype was real — though a high price point, heavy design, and poor adoption have turned it into an afterthought. While Meta has shown a fit in the market, Apple needs to stick with the Siri redesign before being considered a player in the AI wearable space.

TOGETHER WITH VANTA

The Rundown: Security teams are facing more buyer expectations, regulatory demands, and business risk — meaning compliance has to scale even faster. Join experts Ashish Rajan (CISO at Kaizenteq) and Faisal Khan (GRC Expert at Vanta) for a tactical conversation on maturing compliance, risk, and trust in the AI era.

In this live session, you’ll walk away with:

  • A roadmap for prioritizing your next program improvements

  • Top workflows for AI to help extend your impact

  • Techniques to prove value across product, sales, and your board

THINKING MACHINES

Image source: …

The Rundown: Thinking Machines Lab, the AI startup founded by ex-OpenAI CTO Mira Murati and several prominent researchers, introduced Tinker — an API that enables developers to customize frontier models without managing complex infrastructure.

The details:

  • Tinker supports fine-tuning both supervised and reinforcement learning methods on models like Meta's Llama and Alibaba's Qwen.

  • The models can be adapted for specialized applications like solving math problems, analyzing chemistry data, and other tasks with simple code.

  • Early users including Princeton, Stanford, and Berkeley, have applied Tinker to build custom AI systems for math proofs, scientific reasoning, and research.

  • The company is accepting applications for early access starting today at no initial cost, with plans to introduce paid tiers in the coming weeks.

Why it matters: Creating AI models from scratch requires significant resources, but customizing existing models could be a better path for most organizations. Murati and co. are betting that the future is not to who builds the biggest all-purpose AI, but to who can make it easiest to create tons of hyper-specialized ones.

AI TRAINING

The Rundown: In this tutorial, you will learn to transform a casual selfie into a polished professional headshot using Gemini 2.5 Flash, with no expensive studio session required.

Step-by-step:

  1. Go to gemini.google.com and toggle on "Create Images" at the top of the prompt box

  2. Upload your selfie and paste this prompt: "Transform my selfie into a fresh but professional image suitable for a social media profile. The lighting should be natural, flattering, and multi-dimensional. My head is slightly tilted so it doesn't look stiff, and it should look like I'm in a modern, bright office with a blurred background"

  3. Review the result and tweak the prompt if needed, and adjust the background description or lighting until it feels right

  4. Download your finished headshot to use for LinkedIn, résumés, websites, or speaking profiles

Pro tip: Play around with different backgrounds like "warm café," "neutral studio," or "outdoor blurred park" to see which one aligns with your professional brand.

PRESENTED BY GLEAN

The Rundown: In The State of AI At Work in 2025, Glean explains how leading orgs are avoiding inefficiencies by centralizing their AI capabilities on unified platforms.

Inside, you’ll learn about:

  • Where organizations are investing in 2025

  • The hidden “AI tax” that arises from fragmented tools

  • How centralized AI platforms cut through fragmentation, unifying governance, security, and reducing costs

GOOGLE DEEPMIND

Image source: Google DeepMind

The Rundown: Google DeepMind researchers unveiled Dreamer 4, an AI that masters video game tasks by training within its own mental simulation, becoming the first agent to collect Minecraft diamonds using only offline data, without touching the actual game.

The details:

  • Dreamer 4 trains by practicing in a predictive world model that simulates Minecraft's physics in real-time, executing over 20k actions from visual input.

  • The training is in stages: learning Minecraft from videos, adding decision-making abilities, and improving via practice — all without playing the real game.

  • The world model achieved new highs in accuracy, with testers completing 14/16 tasks in Dreamer 4's simulation compared to 5 in rival models like Oasis.

  • Dreamer also beat OpenAI's Minecraft VPT agent while learning from 100x less data, and outperformed systems built on Gemma vision-language models.

Why it matters: It is always cool to still see games like Minecraft being used to test next-level agentic training and capabilities, but Dreamer 4’s skills translate far beyond gaming — with learning through simulation opening safer and more efficient development paths for robots that can replace costly and often dangerous IRL testing.

QUICK HITS

  • 🎥 Sora 2 - OpenAI’s new SOTA video generation model

  • 🤖 GLM-4.6 - Open LLM with improved reasoning, agentic, coding capabilities

  • ⚙️ Tinker - Thinking Machines’ API for fine-tuning language models

  • 🗣️ Octave 2 - Hume AI’s next-gen multilingual text-to-speech model

Google unveiled Gemini-powered upgrades to its Home hardware, featuring AI Nest Cams and Doorbells, a redesigned app and Home Speaker, and a new paid plan.

OpenAI’s new Sora social app surged to No. 3 on Apple’s App Store behind just Google Gemini and ChatGPT, following its viral invite-only launch.

Hume AI launched Octave 2, a new multilingual text-to-speech model that supports 11 languages and includes new voice conversion and phoneme editing features.

Character AI removed Disney characters, including Elsa, Moana, Spider-Man, and Darth Vader, from its platform following a cease-and-desist from the company.

Pew Research Center found that 9% of U.S. adults are getting news from AI, with a third of them finding it hard to determine what's true and half getting inaccurate news.

Google launched new visual search capabilities in AI Mode, allowing users to search with images or text and streamline shopping across over 50B product listings.

Zhipu AI released GLM-4.6, a new open-source LLM with a 200k context window that beats out Claude Sonnet 4 and DeepSeek-V3.2 across a series of benchmarks.

COMMUNITY

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 Cynthia L. in Jupiter, FL:

"I'm an instructional designer of eLearning courses and am building my personal portfolio. I'm using AI to develop scripts and quizzes, images and videos, TTS, and visual design briefs (for color palettes, typography, imagery style and iconography, layout and interface elements). I still have to build the course, but AI saves me so many hours of work!"

How do you use AI? Tell us here.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown

Reply

or to participate.