You stop rebuilding context every time you open a new tab.

📡 Your browser tabs became AI context →
📡 The internet's AI teacher picked Anthropic →
🛠️ Run one messy question end-to-end →
🔍 Your Monday inbox gets triaged overnight →

Jump to the one that matters to you ↓

What's new: Google's rebuilt AI Mode search box now accepts open Chrome tabs, screenshots, and files directly inside the same thread. Instead of manually comparing tabs yourself, AI Mode synthesizes across all of them in one conversation using Gemini 3.5 Flash. (Google’s I/O Roundup)

Dan's take: The real shift isn't smarter search. It's that tab management itself is starting to disappear as a workflow. I tested this during Google's May 2026 rollout week and realized how much of my "research process" was just me manually carrying context between windows.

The screenshot upload mattered more than I expected. I dropped in a campaign dashboard screenshot and skipped the usual copy-paste cleanup entirely.

Here's the honest part: Test 3 didn't hold. Google is still caught between what it was and what it's trying to become — and that gap is real.

The test: Tab synthesis on hiking-trail comparisons, screenshot upload on a campaign report — both held. The multi-step planning thread collapsed to link results on the third follow-up.

Lindy sorts your inbox, drafts replies in your voice, and preps you before every meeting over iMessage. 400,000+ professionals use it. Try it free.

Try Lindy Free

What's new: Andrej Karpathy joined Anthropic on May 19 after pausing Eureka Labs, his AI education startup. He's building a new research group focused on using Claude to speed up Claude's own pre-training. (Karpathy’s X Post)

Dan's take: The signal isn't corporate chess. The person who taught millions of people how AI works just decided the real work is inside the lab. Anthropic crossing a reported $1 trillion secondary-market valuation is interesting. Karpathy pausing his own school to join them is the part I can't stop thinking about.

What this means for Claude over the next 12 months: I don't know yet. Nobody does.

But when the guy who coined "vibe coding" decides he'd rather build than teach, I pay attention.

Karpathy is the talent signal. The money signal is stranger.

ChatGPT is still the app most people know. But in April, Anthropic said Claude had passed $30B in run-rate revenue, up from about $9B at the end of 2025. Anthropic also said customers spending $1M+ per year doubled from 500+ to 1,000+ in less than two months.

For comparison, OpenAI was reported around $25B in annualized revenue. Different companies, different accounting, but the direction is hard to ignore.

ChatGPT attention: 🟩🟩🟩🟩🟩🟩 bigger consumer mindshare
Claude money: 🟨🟨🟨🟨🟨🟨 business use caught up fast

This workflow keeps context alive across follow-ups instead of forcing you to restart every search from zero.

Step 1: Go to google.com and look for the AI Mode tab under the search bar. Don't see it? Enable it at search.google.com/search-labs.

Step 2: Type one real work problem the way you'd explain it to a smart colleague:

Step 3: Follow up in the same thread. Don't start a new search:

Step 4: One more follow-up:

Step 5: Notice what you didn't have to re-explain. That's the shift.

One common mistake: Typing a keyword instead of a real question, then wondering why the response feels like a search result.

Expected output: A synthesized answer to your actual situation plus two follow-ups that build on each other without restarting context. Under 8 minutes.

Or reply GOOGLE and I'll send it directly.

Going Deeper: If you're already running AI Mode with Chrome tab context, run the same multi-step thread with 3–4 competitor pages attached. The synthesis layer handles competitive research faster than any manual stack.

🧠 AI sounds confident when it's wrong — 51% 😵 I lose important chats and files — 22% ⏱️ I keep repeating the same prompts — 19% 🔄 My workflows break between tools — 9%

Half the room picked hallucination. Not memory loss, not broken workflows. The specific failure mode where AI says something wrong with complete confidence. That's 102 people who've been burned in a real situation — and why the Trust Check in this issue exists.

Three replies that stayed with me:

A reader in the UK: "The Lord Justice was not impressed when the quoted case law didn't exist."

Kathy was away from home. AI confidently misread her security camera. She called her daughter to physically check the house. Trick of the camera angle.

One reader worked a thread for 4+ hours. When the final output wouldn't come, the AI started defending its own failures instead of fixing them.

I went from Google to ChatGPT for search because I wanted a response, not a list. Then to Perplexity because I needed citations I could actually trust. Now AI Mode is pulling me back to Google more than I expected — which I didn't see coming.

But I still miss old search. Not for the results. For the experience of landing on a blog post you didn't know you needed and reading the whole thing.

What did you actually lose when search stopped feeling like that?