Weekly AI Digest: Meta’s Smart Glasses, DeepMind’s “Historic” Win & Google’s £5B UK Bet
Week 38, 2025
This week’s big story is Meta’s full push into AR wearables. Their new Ray-Ban “Display” smart glasses bring a built-in heads-up display, translation, and gesture controls straight to consumers. It’s the clearest signal yet that Big Tech sees AI-enabled hardware as the next frontier for interfaces.
Other headlines: DeepMind claims a “historic” programming breakthrough, YouTube is rolling out generative AI editing tools for Shorts, and Google just announced a £5B AI investment in the UK.
Let’s dive in!
Meta Launches Ray-Ban “Display” Smart Glasses
Meta unveiled new Ray-Ban smart glasses with a lens-embedded display, real-time translation, navigation, and an EMG wristband for gesture input. US launch is set for Sept 30 at $799, with European rollout in 2026. They also revealed Oakley Meta Vanguard sports glasses.
This marks a bigger bet on consumer-facing AI hardware after years of Meta Reality Labs’ mixed results.
DeepMind Claims “Historic” AI Programming Breakthrough
DeepMind’s Gemini 2.5 variant solved a gold-medal ICPC task and ranked second overall, which the company calls a “historic” milestone for AI problem-solving.
A version of the company’s Gemini 2.5 AI model solved a complex real-world problem that stumped human computer programmers to become the first AI model to win a gold medal in this programming competition.
YouTube Introduces Generative AI Tools for Shorts
YouTube announced a new set of creator tools powered by generative AI. Leveraging Google’s Veo 3, Youtube announced text-to-video for Shorts, automated editing features, and new remixing options.
The rollout aims to speed up Shorts creation, giving YouTube another competitive edge against TikTok and Instagram.
How People Are Using ChatGPT
OpenAI and NBER released the largest study yet on ChatGPT usage, analyzing over 1.5M conversations. Findings show ~30% of use cases are work-related, while ~70% are personal. Most interactions fall under “asking and decision support” rather than pure automation.
This data clarifies how generative AI is embedding into daily life at scale. It will be interesting to see how these patterns evolve as AI tangles itself with the real world.
Google Commits £5B to UK AI
Ahead of Trump’s visit, Google pledged £5B over two years into UK AI and infrastructure, including a new datacenter at Waltham Cross.
UK leaders framed the deal as a strategic vote of confidence in the country’s AI ecosystem.
OpenAI Upgrades Codex
OpenAI introduced GPT-5-Codex and a set of enhancements, including faster coding agents, integrated code review, and stronger IDE/CLI integrations.
This is part of a broader push to make AI agents more directly useful in software development workflows.
Podcast of the Week: Chris Dixon on AI-Native Products
In a16z’s latest podcast, Chris Dixon discusses building networks, movements, and products that are truly AI-native.
His framework highlights how distribution and retention will shift as AI is built into consumer products from day one.
This week showed AI breaking into three fronts: hardware you can wear (Meta), reasoning advances at the model level (DeepMind, OpenAI), and capital deployment shaping the geography of AI (Google UK).
If you have the time, I also really recommend that you read the paper on how people are using ChatGPT — very valuable insights on that one!
As always, stay curious,
– Ivo
