Instagram | Sponsor this newsletter

My fellow AI explorers

Greetings from the inbox.

This week, the AI story quietly graduated from "look what the model can do" to "okay, but who is actually using this, who is getting sued for it, and does it work when a life is on the line?" Heavy stuff, but I promise we make it fun.

We have got the first real scoreboard of which giants actually run on AI (spoiler: most are bluffing), and we’ve got the big news: Anthropic is filing to go public. AI personal information leaks are getting a bit spicy as well, so grab your coffee.

In today’s edition:

• 🏆 The first hard data on which S&P 500 companies actually use AI

• 🏦 Anthropic files for what could be the biggest IPO in history

• 🤖 Hackers talked Meta's AI into giving away accounts

AI IPO

The Trillion-Dollar Footrace Just Got a Starting Gun

Anthropic confidentially filed a draft S-1 with the SEC today, putting the Claude maker on the runway to a public listing.

Almost nothing concrete is public yet, but the symbolism is enormous: Anthropic just edged ahead of OpenAI in the race to be the first frontier lab on the public markets.

• 💰 The filing follows a mega-round pushing its valuation near $965B, with revenue reportedly run-rating past $40B on the back of Claude Code and enterprise agents.

• 🗓️ Reporting points to a possible listing as early as October 2026, advised by the firm that ran Google's 2004 IPO.

• 🏁 Context: SpaceX filed its own S-1 targeting a June 12 debut, and OpenAI is reportedly filing too.

Here is the thing worth sitting with.

Three of the most consequential private companies on Earth are about to ask public markets for a combined sum that may top $200B, against a 2025 US IPO market that raised roughly $45B total.

The bull case is the trillions parked in money-market funds hunting for a home. The bear case is that public investors price these like actual businesses with actual margins. Whoever lists first sets the template for how the whole sector gets valued, and Anthropic just volunteered to be the test case.

🔮 Prediction: Whoever prices first sets the template, and the market will not be kind to vibes. Expect the debut listing to force the entire private AI ecosystem to confront a number it has been avoiding: what these companies are worth when revenue and burn are on a public page, not a pitch deck. The down-round era for second-tier labs starts here.

Check out the details for this big week of IPOs here →

AI Talent

Four Companies Just Scored a Perfect 100 on AI. Almost Everyone Else Is Faking It.

For two years, every earnings call has been a contest to say "AI" the most times, and a new study finally scores who is actually doing it. The AI-Driven Enterprise Institute graded all 500 S&P companies on public signals like earnings transcripts, patents, and job postings instead of self-reported fluff, and the gap between the leaders and the pack is brutal.

• 🥇 Only four firms hit a perfect 100: Nvidia, Amazon, Meta, and oilfield-services giant SLB. All the usual suspects… sort of.

• 🛒 Walmart led all non-tech companies at 95.84, ahead of utilities like NextEra.

• ⚠️ Researchers flagged weak boardroom AI literacy: executives championing tech they do not actually understand.

The "why" of this story beats the idea of the leaderboard. Nvidia tops the list not because it uses the most AI but because its hardware sits under the entire build-out, and the appearance of SLB and Walmart near the top is the real tell: this stopped being a tech-sector story.

With global AI spending set to clear $300B this year, expect "where do we rank?" to become a standard board question fast, and expect a lot of CEOs to discover their AI strategy is mostly vocabulary.

🔮 Prediction: With global AI spending set to clear $300B this year, "where do we rank versus our peers?" becomes a standard board question by Q4, and a quiet panic spreads as a lot of CEOs discover their AI strategy is mostly vocabulary. The scoreboard, not the press release, starts moving stock prices.

Read the full CNBC story here →

30-Second AI Play

Pressure-Test Your Own AI Support Flow Before a Hacker Does

This week, attackers hijacked 100+ high-value Instagram accounts by simply asking Meta's AI support bot to do it. No 2FA, no human, just a polite request and a one-time code sent to the attacker. Meta has patched it, but here is how to sanity-check your own setup in about 30 seconds of thinking and an afternoon of fixing:

1. Map your write-access agents. List every AI or automated flow you work with that can change emails, reset passwords, or move account ownership. That’s your blast radius.

2. Social-engineer yourself. Open your own support bot and ask it, in plain language, to add a new recovery email to a test account. If it complies without verifying identity, you have the exact Meta bug.

3. Gate identity changes behind hard auth. Anything that changes who controls an account should require step-up verification, never just a code sent to a new, attacker-supplied address.

4. Log and rate-limit resets by geo. The trick worked via a VPN near the target's hometown, so be alert to resets from fresh IPs.

5. Give victims a human escalation path. Do not let your AI be the only door.

The lesson is not "AI support is bad." It is that handing a language model permission to perform irreversible account actions without a verification gate turns helpfulness into a vulnerability.

Other Relevant AI News!

⚖️ Florida just became the first state to sue OpenAI and name Sam Altman personally, alleging ChatGPT was engineered to maximize engagement over safety.

🏗️ OpenAI and Oracle broke ground on their massive Stargate data center in Michigan, promising local ratepayers will not foot the energy bill.

💊 A Tennessee nurse stole fentanyl for months while the hospital's AI watchdog software allegedly never flagged it, raising hard questions about AI oversight.

📉 A New York Fed study argues remote work, not AI, is the real driver behind rising unemployment among recent college grads.

🛒 The same adoption study found Walmart out-AI'd nearly every non-tech company in the S&P 500, beating major utilities.

Golden Nuggets

• 🪙 First to IPO sets the valuation template for the entire sector. Watch SpaceX's June 12 debut as the canary.

• ⚖️ "We added a parental control" may not survive contact with product-liability law. The Florida suit is the one to track.

• 🔐 If you have shipped an AI agent with account-write permissions, treat the Meta exploit as your free penetration test.

• 📊 Bookmark the AIDE index, because "how do we rank versus peers on AI?" is about to become a standard board question.

Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)

Until our next AI rendezvous,

Anthony | Founder of Uncover AI