In partnership with

Welcome back apprentices! 👋

Hey {{first_name|friend}},

Something quietly shifted in tech this week — and no, it wasn't another chatbot that can write your emails slightly faster. 

We're talking bigger moves: the kind that changes how businesses sell, how creators create, and how governments try (bless them) to keep up. 

Grab a coffee. 

Let's make sense of it.

In today's email

  • AI images: finally editable

  • Meta's bot sells for you

  • Washington blinked on AI rules

  • Nvidia bets big on agents

  • More new AI news and tools

Read Time: 4 minutes

Quick News

🎬 Scorsese Says "Action" on AI. The man behind Goodfellas, The Departed, and roughly half of cinema's greatest hits — has quietly been using AI to storyboard his next film, and he just went public about it. As an adviser to image-gen startup Black Forest Labs, he's using their FLUX model to sketch out scenes instantly, calling it "creatively freeing" — while being crystal clear that no AI is generating actors, sets, or footage. For a director of his stature to openly embrace the tool (not the hype) is a meaningful signal in an industry that's been loudly allergic to anything AI-adjacent.

🤖 When the Bouncer IS the Burglar. Meta's AI support tool spent months handing hackers the keys to high-profile Instagram accounts — including a Barack Obama profile and a U.S. Space Force general's page — simply because someone asked nicely (and used a VPN). The exploit was embarrassingly low-tech: switch your location, request a password reset, give a new email, collect the code. No dark arts required. Meta has since patched the flaw and is recovering affected accounts, but the damage — and the awkward PR — is already done.

⚖️ AI Just Aced Office Hours. Stanford ran a blind taste test of legal tutoring — and the professors lost, badly. Sixteen contract-law faculty from 14 schools judged nearly 3,000 head-to-head matchups between their own answers and those of Google's Gemini 2.5 Pro and NotebookLM, picking the AI responses 75% of the time without knowing which was which. Extend the scoreboard further and Claude Opus 4.7 topped a nine-model ranking — with every single AI system still beating the human professors.

Together with WisprFlow

Your prompts are leaving out 80% of what you're thinking.

When you type a prompt, you summarize. When you speak one, you explain. Wispr Flow captures your full reasoning — constraints, edge cases, examples, tone — and turns it into clean, structured text you paste into ChatGPT, Claude, or any AI tool. The difference shows up immediately. More context in, fewer follow-ups out.

89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Try Wispr Flow free — works on Mac, Windows, and iPhone.

Week 22 of 2026 
Agents, Chips, and Executive Orders

If last week felt like drinking from a firehose, this week handed you a pressure washer. 

Image generation quietly grew up, Meta pushed AI agents into the pockets of a billion potential customers, Nvidia reorganized its entire empire around software that barely existed two years ago, and the White House decided it wanted a front-row seat to the AI race — just not close enough to actually slow it down. 

The throughline across all of it? Agents. 

Every major player, from chipmakers to social media giants to the Oval Office, is now either building for them, selling to them, or nervously watching them. 

Key Points You Shouldn’t Miss

  • 🖼️ Ideogram 4.0 goes open-source and beats all open models on Design Arena; Reve 2.0 jumps to #2 overall on the text-to-image leaderboard, behind only GPT-image-2

  • 💼 Meta Business Agent launches globally across WhatsApp, Instagram, and Messenger — already used by 1M+ businesses, free to start with paid tiers coming

  • 📜 Trump's AI executive order trims a proposed 90-day pre-release security review down to a voluntary 30-day window, and rules out any mandatory AI licensing

  • ⚙️ Nvidia at COMPUTEX unveils RTX Spark PCs, the Vera agent CPU (1.8x faster than rivals), Cosmos 3 robotics model, and Nemotron 3 Ultra — a 550B parameter open model now at the top of U.S. open-source rankings

The Slot Machine Era Is Over

Remember re-rolling the same prompt 40 times hoping for something usable? 

Both Ideogram and Reve just buried that workflow.

Ideogram 4.0 goes fully open-source, tops every open model on Design Arena, and finally cracks AI's historic weakness — text and typography. 

Reve 2.0 takes a more surgical approach, building images "like code" so you can tweak a single element without regenerating the whole thing. 

The real story isn't better generation anymore — it's control. And with Ideogram open-sourcing a near-frontier model, the gap between open and closed is closing faster than most people expected.

A Billion-Customer Storefront, Now With an AI Clerk

Meta embedding business agents across WhatsApp, Instagram, and Messenger is a direct play to become the world's default commerce layer — and with 1B+ daily users and 1M businesses already piloting it, the distribution advantage is almost unfair. 

The agent handles the full sales funnel, across languages, without a human unless you want one. Shopify and Zendesk integrations confirm this is infrastructure, not a feature. 

A Front-Row Seat, Not a Guardrail

What started as a 90-day mandatory security review of frontier AI models — enough to genuinely rattle release timelines — got signed as a voluntary 30-day window after Trump said the longer version would "get in the way of" the race with China. 

Former AI czar David Sacks, who reportedly fought the original draft, came around once the clock shrank. The order rules out mandatory licensing entirely and points the DOJ at AI-powered cyberattacks, which is arguably its most useful clause. 

The honest read: the government wants visibility into what's coming off the frontier, just not enough to actually slow it down.

Building the Infrastructure for a World Run by Agents

Nvidia's COMPUTEX thesis was refreshingly blunt: AI agents are the next major consumer of compute, and Nvidia is building everything around them. 

RTX Spark chips run agents directly on PCs, the Vera CPU (1.8x faster than rivals) is already deployed by Anthropic, OpenAI, and the NYSE, and Cosmos 3 gives robots the ability to plan ahead rather than just react. 

Oh, and Nemotron 3 Ultra — a 550B parameter open-source model — just landed at the top of U.S. open-source rankings, taking shots at Chinese rivals Qwen3.5 and Kimi K2.6. For a $5 trillion company, Nvidia's message was surprisingly simple: we're not just making chips anymore — we're building the roads for a world where the cars drive themselves.

What's the Deal for You?

Whether you're a freelancer, a business owner, a developer, or just someone who uses the internet — this week's news affects your daily life more than most AI headlines do. 

Image tools now work the way creative professionals actually think, making high-quality design accessible without a steep learning curve. 

Meta's business agents mean the next time you message a brand on Instagram, there's a very good chance an AI is closing the deal. 

The Nvidia stack means the hardware race is being quietly reoriented around autonomous software — which will shape what your next laptop, phone, and workplace tool actually do. 

And the White House order tells you something about the pace of all this: even the government, which usually moves at geological speed, is trying to keep up.

If you run any kind of business with a social media presence, go look at Meta Business Agent today — it's free to start, it's already live, and your competitors are probably setting it up right now. The early-mover advantage on AI customer tools won't last as long as it did on, say, having a website.

Help Your Friends Level Up! 🔥

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

Last call:

The World's Biggest Dev Event Hits Silicon Valley

From AI and cloud to DevOps and security — WeAreDevelopers World Congress brings the entire modern stack to San Jose. 500+ speakers. 10,000+ developers. One epic September. Use code GITPUSH26 for 10% off.

🧪 Test the Prompt

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

👑 The winner is…

Last week, we challenged you to test GPT-4o’s visual generation skills with this prompt.

Here’s the WINNER:

Congrats to Nadia for this creation!🥳

Want to be featured next? Keep those generations coming!

🎨 Prompt: The Museum of Alternate Versions

Inside a bright, modern design museum, a circular gallery displays dozens of radically different versions of [your object] arranged in chronological order around the room. Each version represents a different future: one ultra-minimal, one luxury-grade, one built for space travel, one made entirely of transparent materials, one powered by advanced AI. Visitors walk through the exhibit comparing the possibilities, while the most advanced version stands on a central pedestal under dramatic lighting. Ultra-detailed, photorealistic realism, vibrant colors, premium materials, architectural photography, sharp focus, cinematic depth of field.

We’ll be featuring the best generations in our next edition!

The Framework Behind our Prompts

If AI outputs feel inconsistent, it’s usually not the model, it’s missing structure.
We documented the exact 6- Part System we use to get reliable results across ChatGPT, Claude, and Gemini.

It’s a short guide you can finish in under an hour, with plug-and-play prompts + exercises so you actually build the skill and fix the frustrating AI inconsistencies.

Subscriber Price: $10 (normally $19).

The Last Time Stocks Were This Expensive Was December 1999.

"Right now, it's good. But it was in '72, '86, 2000, and 2007." - Jamie Dimon, May 2026.

The Shiller CAPE ratio just hit 42.3. The only time in 140 years it's been higher? December 1999.

Stocks can stay expensive for a long time...

It’s one metric to consider, but when your portfolio is built around the most expensive equities in modern history, what else you diversify with could really matter.

Blue-chip contemporary and post war art has shown near-zero correlation with the S&P since 1995.* Prices are largely driven by private collectors competing for a fixed supply of artwork by artists like Banksy, Basquiat, and Picasso.

Masterworks lets you invest in shares of that market.

  • $1.3B deployed across 500+ artworks

  • 29 exits to date

  • Net annualized returns like 16.5%, 17.6%, and 17.8%, not including those unsold

*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.

DISCLAIMER: None of this is financial advice. This newsletter is strictly educational and is not investment advice or a solicitation to buy or sell any assets or to make any financial decisions. Please be careful and do your own research.

Keep Reading