Sponsored by
TL;DR
ChatGPT Images 2.0 can finally render text, generate infographics, and produce eight consistent images from one prompt
Uber blew its entire 2026 AI budget in four months because engineers wouldn't stop using Claude Code
Claude Code now shows a one-line session recap when you return to your terminal
SpaceX offered $60B to buy Cursor before it could close a $2B funding round
You can tap into the skincare market using a single API
The ChatGPT agent built on a real ghostwriter's system for growing 𝕏 accounts
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
FINE, THE IMAGE GEN IS IMPRESSIVE
For years, AI image gen was a party trick. Cool demos, cursed hands, garbled text. Fun to mess around with, useless for anything real.
That's mostly been the story until recently.
ChatGPT Images 2.0 dropped Tuesday and honestly it's worth paying attention to. It renders dense text accurately, in multiple languages. It generates full infographics, floor plans, UI mockups, manga sequences, maps. Up to eight consistent images from one prompt, with the same characters and objects across all of them. It can search the web mid-generation to get things right. Feed it your internal PowerPoint, it'll spit out a production-ready poster.
The reason it works differently is reasoning. When you use a Thinking model, it doesn't just draw. It plans the layout, reads your inputs, searches if it needs context, then renders. Slower than before. But the output is actually something you'd use.
I've been watching the early outputs people are sharing and the text rendering alone is a big deal. That's been the tell forever. Blurry, wrong, nonsensical. Not anymore.
Oh, and buried in the same release: OpenAI also launched workspace agents in ChatGPT. Shared agents that run in the cloud, handle long-running workflows across tools and teams, and keep working even when you're not. Powered by Codex. Feels a lot like OpenAI watching Claude Cowork eat their lunch and deciding to show up to the fight.
OUR TAKE
If you're doing any kind of content or design work, the text rendering and multilingual support alone are worth testing.
Eight coherent images from one prompt is the feature brand teams and creators have quietly needed for a long time.
The workspace agents thing is interesting but it's early (and only available to Business, Enterprise, Edu, and Teachers plans). The real question is whether teams actually adopt it or just keep using whatever Claude-based workflow they've already built.
Definitely worth opening ChatGPT this week and trying them out 🤷🏽♀️
QUICK HITS
Uber burned through its entire 2026 AI budget in four months: Engineers were encouraged to use Claude Code and Cursor, even ranked on internal leaderboards by usage. It worked… 11% of live backend code is now written by AI agents. It also blew Uber’s $3.4 billion budget. CTO Praveen Neppalli Naga says they're "back to the drawing board" on costs, and they're now eyeing OpenAI's Codex as a next option.
Claude Code now recaps your terminal session when you return: Step away for three minutes, come back, and you'll see a one-line summary of what happened while you were gone. It generates in the background automatically and is ready when you switch back. They also added a live PR status indicator in the footer, color-coded by review state (green = approved, red = changes requested, etc.) and updates every 60 seconds.
SpaceX made a $60B offer to acquire Cursor before it could close a $2B funding round: The deal gives SpaceX the option to buy Cursor later this year, or pay $10B to collaborate on AI development in the meantime. Actual acquisition is on hold until after SpaceX's IPO this summer, easier to finance with public stock. The timing makes sense for both sides: Cursor is under real pressure from Claude Code and Codex eating into its market, and SpaceX needs an AI story before it goes public.
HubSpot's ex-Head of Paid shares his 2026 playbook
Rex Gelb spent a decade building HubSpot's paid engine. Now he's showing founders exactly how to do it.
On April 27th, get the framework to structure, launch, and scale paid media that drives pipeline, not just traffic. 20 minutes. Live Q&A. Free.
AI PRODUCT WORKFLOW THAT MAKES MONEY
Turn a selfie into a full skin health report using one API call (and everything you can build on top of it)
⏱️ 2–3 hours | 🔧 Any app builder, Perfect Corp API
Why build this? The skincare market is at $180B and it's moving toward personalization fast. People want to know what's actually going on with their skin, not just generic advice. Perfect Corp has an AI skin analysis API that's been powering beauty tech for major brands globally, and it's now open to developers. One API call returns acne scores, dark circles, wrinkle zones, texture breakdowns, and an overall skin health score. You can build on top of that without training a model or writing backend code.
Steps:
Go to Perfect Corp's developer portal and grab your API key. They're offering 500 free units ($100 value) to get started, so hit the playground first and run a test image. See what the output actually looks like before you build anything.
Open Lovable (or any app builder) and start a new project. Describe the app you want: photo upload, skin analysis results, concern scores displayed visually. Lovable handles all the front-end scaffolding.
Wire up the API. You're making a POST request with the image data and your API key. That's the core of the integration. The API returns a JSON response with concern categories, severity scores, and overlay data.
Map the results into your UI. This is where the product work happens. Raw numbers aren't useful to users. Turn them into visual score cards, color-coded zones, readable labels.
Decide what you're building on top. Subscription gating, product recommendations, a B2B white-label tool for med spas, or a free lead-gen widget for beauty brands. The analysis is the foundation. What you build above it is the product.
What you'll end up with: A working skin analysis app with zero backend code. From there, you've got a real foundation to ship something. Consumer SaaS, e-commerce integration, or B2B tool for clinics. The infrastructure is already there. You're just deciding what to build on top of it.
Full tutorial:
TOOL OF THE DAY
The world's first AI head of content for 𝕏, built on a real ghostwriter's playbook.
Most AI Twitter tools just generate tweets. Paste a topic, get five mediocre options, pick the least bad one. Stanley is different. It's built on the actual system a ghostwriter used to grow accounts from zero to 10k: niche research, voice matching, content strategy, hooks, threads, and the part everyone skips, staying consistent.
Founders who know they should be posting but aren't. Creators stuck in the "I don't know what to write" loop. Anyone who wants ghostwriter-level output without the ghostwriter price tag. That's who this is for.
The moat on 𝕏 was never one great tweet. It's showing up every week with content that actually sounds like you.
