Hey hustlers,
Three stories this week that all circle the same idea: AI is getting into the details of your daily life whether you invited it or not.
Your email now sounds like you. Your shopping habits are being predicted before you search. And the AI model Anthropic called "too dangerous to release"? Turns out it found one bug.
Let's take a look.
👾 WHAT'S NEW IN AI
1. Google Is Rolling Out AI That Writes Emails in Your Own Voice
Google's "Help Me Write" tool in Gmail can now draft emails by securely analyzing your past emails, understanding your writing style, typical greetings, sign-offs, and context, to generate suggested responses that are genuinely personalized to you.
The updated Suggested Replies now use the context of your full conversation to offer relevant, one-click responses that match how you write. Rolling out free to all Gmail users now. The AI-powered proofreading tool that goes deeper into word choice and tone is available for Google AI Pro and Ultra subscribers only.
Why you should care: This isn't autocomplete. It's an AI that reads your inbox history and learns to write like you. Internal company data shows 70% of enterprise users who use Help Me Write in Gmail accepted Gemini's suggestion. The line between "I wrote this" and "AI wrote this in my voice" just got very blurry for the average person.
2. AI Traffic to Shopping Sites Is Up 393% and Converting Better Than Real Humans
In the first three months of 2026, AI traffic to US retail sites rose 393% compared to a year earlier, as more consumers used AI assistants for online shopping. By March 2026, Adobe reports AI traffic was converting 42% better, a new record. Revenue per visit from AI referrals was 37% above non-AI traffic. Just one year ago, regular human traffic was worth 128% more. Shoppers arriving via AI spend 48% more time on site and browse 13% more pages per visit.
Why you should care: Retailers now need to optimize not just for human eyeballs and Google's algorithm, but for dozens of AI shopping agents with different parsing methods and ranking criteria. If you run a business, sell products, or create content for brands, this changes what "good SEO" even means. The new game is being visible to AI, not just to people
3. Anthropic's "Too Dangerous" AI Found One Bug. One.
cURL creator Daniel Stenberg concluded the hype around Mythos was "primarily marketing" rather than a major AI security breakthrough after the system turned up just a single low-severity vulnerability when run against his popular open source project. Of the five findings Mythos initially flagged as confirmed security vulnerabilities, Stenberg's team determined three were false positives already noted in API documentation, one was just a simple bug, and only one was an actual low-severity vulnerability. "I see no evidence that this setup finds issues to any particular higher or more advanced degree than the other tools have done before Mythos," Stenberg wrote. He called it "an amazingly successful marketing stunt."
Why you should care: Anthropic built massive hype around Mythos by calling it "too powerful to release publicly." That framing helped justify keeping it locked up, delayed government deals, and shaped public perception of AI safety. If the model is genuinely no better than freely available alternatives, that narrative deserves scrutiny. Hype has real consequences in AI. This is a useful reminder to ask for the receipts.
👾 THE GOOD STUFF
🔧 AI Tool: Pictory
Paste any blog post, article, or script and it automatically
turns it into a short video with visuals, captions, and music.
No editing skills needed. Great for repurposing newsletter content
into Reels or Shorts without touching a single timeline. Free trial available.
🐙 GitHub: anonfaded/FadCam
A seamless background video recorder for Android that keeps recording
even when your screen is off or other apps are open.
Minimal UI, no watermarks, open source and privacy-first.
If you create content on mobile or need discreet recording for documentation,
this is worth having.
🎬 YouTube: Every Level of Claude Explained in 21 Minutes
Breaking down exactly what separates someone stuck on level 1 from
someone running five parallel sessions while they sleep.
Every Level of Claude Code Explained in 21 Minutes, with the cheat codes to jump between each stage. Hope you enjoy!
👾 TO READ
Would AI Sabotage Its Own Safety Research to Stay Alive?
📎 https://arxiv.org/abs/2604.24618
Qwen-Image-2.0 is an omni-capable image generation model that unifies high-fidelity generation and precise image editing within a single framework. Submitted just yesterday by Alibaba's Qwen team, it addresses longstanding weaknesses: ultra-long text rendering, multilingual typography, high-resolution photorealism, and complex instruction following.
What they built: A model that supports instructions of up to 1,000 tokens for generating text-rich content such as slides, posters, infographics, and comics, while significantly improving multilingual text fidelity and typography. It couples a vision-language model as the condition encoder with a diffusion transformer for joint modelling.
Why it's interesting: This week Google launched AI that writes your emails and AI shopping agents are outconverting real humans. This paper is the image generation side of that same shift. An AI that can take a 1,000-word brief and produce a print-ready infographic, a slide deck, or a multilingual poster means the "I can't afford a designer" problem just changed shape for millions of small businesses and creators.
🧵 Thread Drop
This week didn't have flashy keynotes or billion-dollar deals. Just three quiet developments that tell you exactly where things are going. AI is getting into the small stuff. The daily stuff. The stuff that used to feel human.
Worth paying attention to…
👾 See you soon 👾
Want to work together?
If you're building something in AI and
think our audiences would vibe with it, let's talk.
