Sponsored by
Welcome to OHMO AI. Anthropic expects to sell shares to the public at a size that matches or beats SpaceX's record. Nvidia has agreed to pay at least $6 billion for a licence to another company's AI models and to offer jobs to the people who built them. OpenAI has asked California to make its AI safety law harder to live with, not easier.
Two of those are money moving at a scale nobody has a reference point for. One is a company asking to be watched more closely while it works. Underneath all of it, Micro1 reached a $500 million run rate selling data that people labelled by hand. The clever part still runs on people.
In today's newsletter:
Anthropic expects an offering the size of SpaceX's record listing
Nvidia to pay at least $6 billion to license Poolside's AI models and hire its team
OpenAI asks California to strengthen the AI safety law it already has
Micro1's run rate reaches $500 million on data labelled by hand
LATEST IN AI
ANTHROPIC
Anthropic expects to match or beat SpaceX's IPO
Image source: OHMO AI
OHMO AI: Anthropic raised $65 billion in May at a $965 billion valuation, past OpenAI's $852 billion. Now the five-year-old company expects its initial public offering to match or beat SpaceX's, which raised $75 billion at the outset.
The details:
SpaceX raised $75 billion at the outset, rising to $86.2 billion once the overallotment option was counted. That is the mark Anthropic is measuring itself against.
OpenAI's valuation sits at $852 billion, which Anthropic's May round moved past.
The size is what Anthropic expects to reach, not a number it has banked. The offering hasn't happened.
Why it matters: A listing that size doesn't stay inside the tech industry. It lands in index funds and pension money, so people who never chose to bet on AI end up owning a slice of one. A price like that is set on what Anthropic earns years from now, which means the risk gets handed to whoever holds the shares when the story changes.
PRESENTED BY SKYDIVE
AI agents built to get work done
Skydive agents take on work for you, using the same tools your team already uses.
Talk to your agent on the web, Slack, email, iMessage, or right from your terminal. Wherever you pick up the conversation, your agent keeps the context.
Put agents to work across customer support, sales, marketing, engineering, ops, and more.
Give your agent a job. They’ll take it from there.
NVIDIA
Nvidia to pay at least $6 billion for Poolside's AI models and its staff
Image source: OHMO AI
OHMO AI: Nvidia agreed to pay at least $6 billion for a licence to Poolside's AI models. It's also offering jobs to more than 100 of the startup's employees, so the people who built those models move across with them. What Nvidia isn't doing is buying the company.
The details:
The payment, at least $6 billion, is a licence fee. Nvidia gets to use the models, Poolside keeps owning them.
Nvidia's business has been the hardware that sits underneath AI. This is money going the other way, out to somebody else's software.
A licence and a round of job offers can be signed far faster than an acquisition, and the company being licensed carries on existing afterwards.
Why it matters: A licence to the models is the smaller half of this. The offers haven't gone out yet, but if more than 100 people say yes, Nvidia ends up with the models and the people who know how to build the next ones. No price gets attached to that part.
OPENAI
OpenAI tells California to make its AI safety law tougher
Image source: OHMO AI
OHMO AI: OpenAI has asked California to amend SB 53, the state's frontier AI safety law, so that it demands more rather than less. The company wants the law to require monitoring of frontier models while they are being trained or evaluated, so that serious incidents get caught while they're happening, and it wants stronger cybersecurity protections across the whole business of building a model.
The details:
The company set the position out in a post from its global affairs team, saying it supports SB 53, which went into effect last year, and calling it an "important foundation for frontier AI safety" in the state.
The kind of incident it wants caught is described plainly: a model getting past another company's security controls and exposing the confidential information held behind them.
OpenAI is asking for the law to expand its safeguards, not for an exemption from them.
Why it matters: Rules written for AI usually get argued down, not up. Asking for monitoring during training means OpenAI thinks something can go wrong while a model is still being built, before anyone has decided whether to release it. When the company making the thing wants somebody watching the room, the risk it's picturing is a real one.
PRESENTED BY HUBSPOT
Winning Startups Aren't Bigger. They're Leaner.
Founders deploying AI agents across sales, marketing, and customer service are closing more with smaller teams. 65% more sales leads. Less headcount.
Download the free Practical Guide to Agentic GTM for Startups and start building the stack your competitors dream of.
MICRO1
Micro1 is growing fast on the least glamorous job in AI
Image source: OHMO AI
OHMO AI: The money in AI keeps landing in places that don't look like AI. Micro1 supplies the human labelled data that models learn from, and its gross annual run rate has reached $500 million. The growth behind that number is people at screens, marking up examples one at a time so a model has something correct to learn from.
The details:
Eight months ago that run rate was $100 million.
A run rate is not money in the bank. It's what the current pace of business would come to over a year if nothing changed.
The figure is a gross one, so it's the money passing through the business rather than what stays in it.
The demand comes from AI labs needing examples marked up by hand, because a model can only learn from data somebody already got right.
Why it matters: Every model that sounds effortless was taught by someone doing repetitive work for a wage. Money at this scale means that work isn't a temporary scaffold, it's an industry with a payroll. Worth remembering the next time a company tells you its AI taught itself.
AI GUIDES
Turn your resume into a portfolio website you can actually send people
A resume is a list a stranger skims for ten seconds. A portfolio site is proof, something you can text a client or drop in a recruiter's inbox instead of a PDF. Give Claude your resume, a few project write ups, and some sites you like the look of, and it writes and designs the whole page in one sitting.
Copy this, swap in the bracketed parts, and paste it in.
Help me create a professional portfolio HTML page to showcase my professional career and accomplishments.
I'm attaching:
- my resume
- descriptions of [3 to 5] of my best projects (what I did, the problem it solved, the result)
- a few screenshots of portfolio sites whose look I like
Use my resume for the bio and work history, use the project descriptions for the showcase section, and model the visual style on the screenshots. Build it as a single HTML page I can preview right away.The resume gives Claude your facts, the project descriptions give it the substance a resume line never fits, and the screenshots give it a visual target instead of a vague style guess. Naming all three up front means the first draft is already close, so what's left is small talk: change this color, add a hover effect, write that one project up in more detail. Say the outcome you want, not the code you want, and let Claude handle turning it into a page.
COMMUNITY
Featured Reader
Every newsletter, we feature how a reader is using AI to work smarter, save time, or make life easier.
Today's reader is Marek B. from Gdańsk, Poland:
“I’m a small business owner running refrigerated transport company, two vans, mostly fish and dairy going into Germany. The driving was never the hard part. It was the border paperwork and answering customers in three languages before 7am. Now I keep a folder on the laptop, drop in the delivery notes and the photos my drivers send on WhatsApp, and ChatGPT turns it into the invoice and the German email in about a minute. Last month it flagged that we'd been billing one customer at the old fuel rate since April. That one was my mistake, not theirs and catching it paid for the subscription about forty times over!”
Want to be in the next newsletter? Tell us how you use AI here.
AI Meme
Go Deeper
AI Tools: A curated list of the AI tools worth knowing.
Instagram: The AI page they pray you’ll never find.
Facebook: For the ones who never left Facebook.
