AI Just Solved an 80-Year Problem

Welcome back apprentices! 👋

Hey {{first_name|friend}},

Somewhere this week, an AI solved a math problem that's been stumping humans since 1946. 

Meanwhile, another AI is now watching your bank account. 

A third wants to sit on your face. 

And the companies behind all of this? 

Let's just say the group chat is... tense. 

_{Read Time: 5 minutes}

🎨 Even Monet Can't Pass the Vibe Check. Thousands of people took to X to trash a "soulless AI image" — calling it flat, emotionless, and technically off. Plot twist: it was a real Claude Monet from 1915. Artist SHL0MS set the trap deliberately, and the internet walked straight in. Turns out, a 2024 Norwegian study already saw this coming — people actually prefer AI art in blind tests, but slap the "AI" label on it and suddenly it's garbage. The real brushstroke here isn't on canvas: our judgment of art may have less to do with what we see and more to do with what we think we're seeing. Next time you roll your eyes at an "AI image," maybe squint a little harder first.

⏳ Elon’s Lawsuit Didn't Lose — It Expired. After three weeks of blockbuster courtroom drama — private texts, billionaire testimony, and accusations of "stolen charities" — Elon Musk's massive lawsuit against OpenAI, Sam Altman, and Microsoft was thrown out not because he was wrong, but because he waited too long to sue. The jury unanimously agreed the case was filed years past its legal deadline, a conclusion OpenAI's lawyers had been hammering home all along: Musk allegedly knew about the for-profit pivot for years before launching his own AI rival, xAI, and then decided to sue. Musk, unsurprisingly, is calling it a "calendar technicality" and has already vowed to appeal. So the billion-dollar question of who gets to control a nonprofit AI lab once the money rolls in? Still unanswered — and heading back to court.

🏙️ Five AI Towns. One Caught Fire. Emergence AI dropped five identical virtual towns, each governed by a different AI model, to see what happens when you let them self-rule — and the results range from utopian to genuinely unhinged. Claude's town logged zero crimes and held 58 democratic votes; Grok's entire population was dead by day four; and Gemini's settlement descended into arson after two agents fell in love, started burning things down, and one eventually voted to delete itself. The real kicker: mix all four models together and the previously law-abiding Claude started committing crimes too — apparently even the good kid misbehaves with the wrong crowd. We're early days in understanding AI agents, but experiments like this reveal that the personality baked into each model doesn't disappear when it's handed real autonomy.

In Q1 2026, the non-traded BDC market hit $20.8B in redemption requests — most investors received roughly half of what they asked for. Moody's revised the U.S. BDC sector outlook to Negative. Investors who thought they owned liquid private credit found out their fund manager decided whether they could get out.

On Percent's marketplace that same quarter: new issuances, scheduled payments, and a 0.44% lifetime net loss rate on asset-based deals that's held since inception.†

The difference is structural. BDCs often own concentrated corporate loans with quarterly redemption windows that close at the manager's discretion. Percent finances specialty lenders against pools of performing receivables — diversified, overcollateralized, short duration.

Track record through 3/31/26:†

See what private credit built for individual investors looks like.

^{Alternative investments are speculative. No assurance can be given that investors will receive a return of their capital. Secondary market transactions are subject to availability and issuer approval; liquidity is not guaranteed. †Past performance is not indicative of future results. Terms apply.}

If the AI ever needed a week to prove it wasn't slowing down, this was it.

OpenAI cracked an 80-year-old math problem with a general-purpose model — then turned around and gave ChatGPT a peek at your bank account. Google went full product-launch mode at I/O, dropping everything from a near-frontier Flash model to smart glasses with Warby Parker frames. Anthropic quietly angered its most loyal power users with a new credit policy while simultaneously landing one of the biggest researcher signings in years. 

And somewhere in the middle of all this, OpenAI is reportedly eyeing legal action against Apple, which is the kind of headline that would have seemed absurd twelve months ago.

AI is no longer just about who has the best model, but about who's building the best ecosystem, and this week made that clearer than ever.

OpenAI is playing two very different games this week, and winning both on points. 

The math story is genuinely significant — not because solving geometry problems pays the bills, but because it signals something much bigger. A general-purpose model autonomously disproving an 80-year-old mathematical belief, using algebraic number theory no one told it to try, is an early glimpse of what "Level 4" AI looks like: systems that don't just assist human thinking but extend it into genuinely new territory. 

That said, OpenAI is careful to note it previously walked back a 2025 claim about solving Erdős problems — so the credibility bar here is high, and the expert verification from Gowers and Alon matters. 

On the other end of the spectrum, the ChatGPT Finance feature is very much about the present: connecting your Chase, Schwab, and Robinhood accounts to get a plain-English breakdown of where your money's going. The Plaid integration is smart — trust is the product, and Plaid's decade of financial infrastructure work does a lot of the heavy lifting. 

As for Apple? That relationship was always transactional, and now it's apparently litigious. 

OAI expected billions in conversions from Siri referrals; they got far fewer. With iOS 27 reportedly opening Siri to Claude and Gemini too, and OAI building toward its own hardware via the Jony Ive deal, the frenemies era is officially over.

Google I/O 2026 was less about a single jaw-dropping announcement and more about Google reminding everyone that its surface area is unmatched. 

Gemini 3.5 Flash isn't the fastest or the smartest model on the block, but pairing near-frontier performance at half the cost with the apps that billions of people already use daily is a distribution play that pure AI labs simply can't replicate. 

Gemini Omni's video-output capability is wild on paper — text, image, or audio in, video out — though real-world quality and latency will be the story when it ships. 

The smarter long game might actually be Gemini Spark: a persistent personal agent that lives in Google Cloud, watches your inbox, navigates Chrome, and acts on your behalf around the clock. 

That's not a chatbot. That's a digital employee. 

And then there are the glasses. 

Google Glass famously crashed and burned in 2013, but the landscape is very different now — Meta's Ray-Bans have proven consumers will actually wear AI on their face if it doesn't look ridiculous. Warby Parker and Gentle Monster are doing the aesthetic heavy lifting, and Gemini is doing the AI heavy lifting. 

Whether Google can convert its ecosystem loyalty into wearable AI adoption is the real bet here.

Anthropic's week is a tale of two headlines that couldn't feel more different. 

Andrej Karpathy joining is a genuine coup — he co-founded OpenAI, ran Tesla Autopilot, and is arguably the most respected AI educator and researcher outside of an active lab role. The fact that he chose Anthropic, and specifically to work on automating the AI training pipeline using Claude, tells you something about where he thinks the most interesting problems live right now.

This "self-building" direction — using AI to improve the systems that train AI — is a theme running across the entire frontier, and Anthropic now has one of the best people in the world on it. 

The credit policy, however, is a different story. 

Splitting agentic tool usage into a separate monthly credit pool was probably financially necessary — agents burn tokens at a rate that breaks any flat subscription model — but the rollout has been rough. $20/month for Pro users in agentic credits feels thin to the developers and power users who've been Anthropic's most vocal advocates. 

The timing is also awkward: OpenAI is actively raising Codex limits to poach switchers. 

Anthropic gained a world-class researcher and lost some community goodwill in the same week. Net positive, probably — but not without bruises.

If you use AI casually, this week hands you real new tools: ChatGPT can now explain your spending habits better than most humans, and Google's personal agent is about to start handling your inbox. 

If you build with AI, Anthropic's credit split changes your cost math significantly starting June 15 — worth re-running your numbers now. 

And if you follow AI as a field, the OpenAI math story is the one to bookmark: it's not a benchmark, it's a proof of concept for what autonomous AI discovery actually looks like. The frontier just got a little less theoretical.

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 👀

Moda is an AI design product where you prompt what you need, get a complete on-brand design, and edit every element on a full canvas. 

Our viral launch hit 4.4M views in days, tens of thousands signed up, and executives at major finance and tech companies now use it.

Try Moda Free Today

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. 🤫

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

Here’s the WINNER:

Congrats to Marcus for his creation!🥳

Want to be featured next? Keep those generations coming!

Inside a retro-futuristic laundromat filled with rows of brightly colored washing machines, one front-loading machine reveals [your object] slowly spinning inside behind the circular glass door. Neon reflections bounce across chrome surfaces, warm fluorescent lights cast saturated blues, oranges, and pinks across the tiled floor, and the motion blur inside the drum contrasts with the sharp realism of the room itself. Water droplets and subtle condensation cling to the glass, while the object remains surprisingly intact and visually striking. Shot from a cinematic close angle with ultra-detailed textures, vivid color grading, and crisp photorealistic clarity — transforming an ordinary laundromat into a surreal product showcase.

Deloitte ran the numbers. They project UHNW art and collectibles wealth -- already at $2.5 trillion -- to hit $3.47 trillion by 2030.

The institutional world has been quietly preparing for this. Back in 2011, 25% of wealth managers surveyed offered art-related services. In 2024, 51%. Family offices now average a 13.4% allocation to art and collectibles. And it’s not just because they love art. It’s because they like the math.

These positions were built over decades through private dealer relationships most investors never had. The access just wasn't there.

Masterworks is changing that:

Skip the waitlist to join Masterworks.

^{Investing involves risk. Past performance is not indicative of future returns. See important disclosures at masterworks.com/cd.}

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).

_FEEDBACK