{{first_name | Leader}}, welcome back.
The AI market had a rough week, and the cracks are starting to show in unexpected places. These are today’s updates.
This tool lets you launch production ready sites in hours*
ChatGPT is about to look very different
Chipmaker earnings just broke the market's winning run
Why Google is renting compute from a rocket company
Tools, resources, and a prompt to design a lightweight internal AI assistant.
. Together with Framer.
Framer helps teams design, build, and launch their marketing sites lightning fast. With the ability to publish hundreds of CMS pages in a single click, operate at a global scale with seamless localization, and even host unified content across multiple domains, teams have never been able to ship faster.
Trusted by companies like Miro, Bilt, and Perplexity
- Speed without chaos: ship pages and updates faster without turning the site into a fragile set of one-off hacks
- Reduce dependency: shift routine brand and marketing work out of product engineering queues.
- Production-grade foundation: Run real marketing systems (CMS, SEO, performance optimization) with governance and collaboration
. Top News.
A regulatory filing revealed that Google will pay SpaceX roughly $920 million per month to access 110,000 Nvidia GPUs. The deal surfaced days before SpaceX's planned market debut on June 12.
The demand for GPU capacity has outrun what even the largest hyperscalers can build fast enough internally. Google is one of the best-resourced infrastructure builders in the world, and they're still renting external compute at nearly $11 billion a year to keep up.
For SpaceX, a guaranteed revenue stream of this size landing just before an IPO targeting a $2 trillion valuation is exactly what institutional investors want to see. For the rest of the market, it confirms that large-scale GPU capacity is now actively traded at the highest levels, and competition for Nvidia's supply isn't slowing down.
The S&P 500 had been on a nine-week winning streak, carried largely by megacap AI names. That streak ended this week.
Falling odds of a near-term Fed rate cut reduced speculative appetite for high-multiple tech names at exactly the wrong moment, turning a contained sector reaction into something much broader.
When AI infrastructure names are priced for perfection, any gap between expectations and reality gets punished fast. The valuations running through this sector leave very little room for disappointment.
ChatGPT started as a chat interface. OpenAI wants it to end up as the app you never have to leave.
OpenAI is planning to convert ChatGPT into a superapp bundling Codex, AI agents, image generation, and third-party tool integrations into one unified interface. Rollout expected in the coming weeks.
Recent product moves signal the direction. Lockdown Mode reduces prompt-injection risks. A revamped memory system improves context retention and privacy controls. In-chat email sending lets users compose and send without switching tools.
Each of these individually looks like a product update. Together they look like a platform strategy. OpenAI is building for stickiness, expanding monetization through subscriptions and third-party integrations, and positioning ChatGPT as the workflow hub rather than one tool among many.
. Signals.
Tools
Links
. Poll.
To hit the best ROI with social-media ads, which AI-powered approach helps most?
. Market.
Funding
Pinterest signed a $4B AWS agreement (through 2031) to accelerate its AI-powered discovery platform.
PointFive raised $60M (Series B) to help enterprises monitor and reduce AI and cloud spend.
Roles In AI
Senior Machine Learning Operations Engineer II at Life360
Lead Software Engineer, 3D Computer Vision at DroneDeploy
Socials
. Prompt of the Day.
Design a Lightweight Internal AI Assistant
When to use this?
When teams keep asking for the same data, updates, or explanations, and you want an AI assistant to handle it.
I want to design a lightweight internal AI assistant for a specific task.
Your job:
Define what the assistant should handle (questions, summaries, lookups)
List inputs it needs (docs, links, dashboards, FAQs)
Define clear boundaries (what it should not answer or do)
Propose a simple setup using tools we already have
Give 3 example questions the assistant should answer well
Keep it practical. Assume no custom engineering.
Task to automate:
[describe the recurring questions or requests]P.S. Get more such prompts in the Prompting Playbook (free for you)
Stay curious, {{first_name | leaders}}
PS. If you missed yesterday’s issue, you can find it here.