Matt Wolfe

AI News: These Google Updates Are Dividing People

⏱ 45 min video · 4 min read22 May 2026Worth watching

TL;DR

Matt Wolfe recaps Google I/O 2026, covering 100+ announcements including new Gemini models, a video-generation model called Gemini Omni, an autonomous agent called Gemini Spark, AI-powered search changes, and AI glasses. He also covers Andrej Karpathy joining Anthropic, OpenAI's personal finance feature in ChatGPT, Spotify's AI podcast tools, and Cursor's new Composer 2.5 model.

Key points

Gemini 3.5 Flash launched: faster and cheaper than competitors (input $0.15/1M tokens, output $9/1M tokens) while matching Claude Opus 4.7 and GPT-5.5 on most benchmarks, with Gemini 3.5 Pro still to come.

Gemini Omni is a new video-generation and editing model grounded in world knowledge, supporting avatar insertion, style transfer, and element addition to existing videos -- any modality in, video out now, more outputs planned.

Gemini Spark is Google's autonomous agent running on Google Cloud servers 24/7, connecting to Gmail, Calendar, Drive, and third-party services via MCP, with permission controls before external actions; currently in trusted tester rollout for Ultra subscribers.

Google Search is shifting toward AI-first responses by default, reducing organic link visibility, threatening traffic to content creators and publishers -- a major controversy because Google relies on that same content to train its AI overviews.

Andrej Karpathy, OpenAI co-founder and former Tesla AI lead, joined Anthropic -- a significant talent move that signals growing competition between frontier AI labs.

Actionable insights

→

Developers building agents should evaluate Gemini 3.5 Flash as a default model -- it leads on agentic benchmarks and is 3x faster and dramatically cheaper than GPT-5.5 and Claude Opus 4.7.

→

Content creators and SEO-dependent businesses should monitor Google Search AI mode rollout closely, as AI overviews increasingly replace organic clicks and a structural traffic decline appears likely.

→

Cursor users can switch most tasks to the new Composer 2.5 model for near-flagship performance at under $1 per task, reserving GPT-5.5 or Opus 4.7 only for the hardest problems.

→

Think carefully before connecting bank data to ChatGPT via Plaid -- OpenAI is building an ad-supported business model and combining financial data with ad targeting raises legitimate privacy concerns.

→

Google Glasses (audio version) are launching in fall 2026 via Warby Parker and Gentle Monster partnerships -- now is a good time to decide whether Meta Ray-Bans or Google Glasses better fit your ecosystem.

Notable quotes

“If Google essentially scrapes the responses from all of these blog posts, puts them in an AI overview so that nobody ever ends up clicking into the blog post -- that's going to mean a drastic reduction in traffic from Google to these websites, and the responses that Google is giving is based on the content that they got from these websites.”

“I just, this is so wild to me because I remember three years ago or so when I was playing with ModelScope and ZeroScope, I would generate a monkey on roller skates and you can barely tell what it even was.”

“There are six different Google tools right now that have AI baked into them that I can go to to build my website -- I think where to go to do certain things is sort of getting convoluted and confusing.”

Worth watching?

✅

Worth watching the full video?

Watch if you want the live demo footage of Gemini Omni and Gemini Spark in action -- the written summary captures all key facts, but seeing the avatar and video editing features in real time adds useful context for evaluating these tools.

Topics

AI & Tech Google I/O

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →

More like this