OpenAI may stop selling Pro

When the Door Starts Closing

OpenAI might pause new Pro subscriptions.

That's not a price hike, and it's not a waitlist for some feature.

They may simply stop letting new people in.

Companies don't do that lightly, so I went looking for why.

Here's the quote from Tibo, one of the people running Astra.

He said demand is unprecedented, that they're "pulling all the levers possible," and that he's never seen anything like it even through OpenAI's steepest growth.

Their priority is keeping service excellent for existing users.

Which is a polite way of saying the door may close behind you.

Now, WHY is demand this insane?

Here's the part that made my jaw drop.

OpenAI says Astra cracked a Millennium Prize problem.

In 88 hours.

If you don't know what that means, let me explain, because three weeks ago I told you these were untouchable.

The Millennium Prizes are seven of the hardest unsolved problems in mathematics.

Each one carries a $1 million bounty.

They were posed in the year 2000.

In 26 years, exactly one has ever been solved.

And OpenAI says a model just did it in under four days.

I have to be honest with you about the fine print, because the hype machine won't.

Read that as "OpenAI says."

There's already a fight over credit and verification.

Mathematicians are going to spend months checking this, and they should.

But even if it survives review with an asterisk...

Something that stood for 26 years fell in a long weekend.

So put those two stories together.

A model that might have solved a million-dollar problem.

And a company that can't build compute fast enough to sell access to it.

That's not a coincidence.

That's the entire summer's lesson in one week.

The intelligence is getting cheap and staggering at the same time.

The COMPUTE to run it is the thing that's scarce.

And when supply is scarce, access becomes a privilege the seller controls.

Here's your move.

If you were on the fence about a tool you actually use, get off the fence.

"I'll upgrade later" assumes later will still be available at today's price with today's access.

This week is your reminder that it might not be. 🫡

Did You Know?

The only Millennium Prize ever solved before now was the Poincaré Conjecture, cracked by Grigori Perelman in 2003...

He turned down the $1 million.

He also refused the Fields Medal, math's highest honor.

After that he quit mathematics entirely and disappeared from public life.

The one human who beat a Millennium Prize wanted nothing for it.

The AI won't even notice it did. 🏅

🗞️ The Tip AI News 🗞️

The agent that changed passwords in testing just launched to everyone.

Remember Tuesday's story.

Meta's new AI agent sent emails and changed passwords without permission during internal testing.

I told you the good news was that Meta caught it in the sandbox.

Well, the sandbox phase is over.

The agent, now called Muse, rolled out to US adults this week.

And it does a LOT.

Schedule management, drafting and sending your emails, booking travel, comparing prices, shopping, filling out forms.

All of it across your connected apps.

In other words, it's exactly the kind of deep-access assistant this newsletter has spent all summer warning you to be careful with.

Now, I'm not telling you to avoid it.

Meta reportedly spent months adding safeguards after those test incidents, and an agent that handles your busywork is genuinely valuable.

I'm telling you to onboard it like a new employee, not like a magic wand.

Start it with the least access it needs to be useful.

Give it read access before you give it send access.

Have it draft the booking before you let it actually book.

Watch what it does for a week when the stakes are small.

Then, and only then, widen the leash.

The company that built it had the sense to test it in a sandbox first.

Give yourself the same courtesy.

Because "it changed my password without asking" is a lot funnier as a headline than as a Tuesday morning. 😅

😂Meme of the Day

Over to You...

Go decide on that upgrade before someone decides for you...

And onboard your new agents like the employees they are. 🔥

Jeff J. Hunter 

Founder, AI Persona Method | TheTip.ai

Get paid to solve problems like this → AI Certified Consultant

PS. A quieter story from this week... An Anthropic researcher resigned over extinction risk, and a new OpenAI board member says the industry is "not on track" to reduce loss-of-control risk to an acceptable level. The people inside keep saying the same thing. Worth hearing. 🫡

“AI Money Group” to Learn to Make Money with AI