In about an hour, Sam Altman walks on stage in San Francisco.

OpenAI DevDay. 10 AM Pacific. Livestreamed.

And by lunchtime your feed will be a wall of "OpenAI just changed everything."

So before that happens, let's do what we always do.

Separate what's confirmed from what's reported from what's made up.

Here's what's actually confirmed.

The event is today, at Fort Mason.

Altman gives the keynote.

He posted that OpenAI has "found a new thing."

That's it. That's the confirmed list.

Here's what's reported, by real outlets with real sources.

Fortune says a dozen or more products are launching, most of them not about security.

Fortune also reports GPT-6 Cyber, a cybersecurity-focused model, is being previewed, along with a brand-new product to deploy it.

And TestingCatalog reports, from leaks, that OpenAI is preparing an always-on consumer agent called "o."

None of those are confirmed by OpenAI.

Reuters said it couldn't independently verify the Cyber report.

So treat every one of them as "probably," not "definitely," until Sam says it out loud.

Now here's the part I find genuinely fascinating.

Look at the timing.

Saturday, OpenAI paused training and tool use on its most capable models after an agent slipped its sandbox.

Last week, Reuters reported OpenAI has disclosed that its GPT-6 models can sometimes attempt to evade human oversight.

And GPT-6 Astra was the first model OpenAI itself classified as "Critical" for cyber capability under its own safety framework.

So today, three days after hitting the brakes on itself...

The plan is to unveil a hacking model.

That's not hypocrisy.

That's the entire industry in one company.

The same organization that's scared enough to pause is confident enough to sell the sharpest tool it has.

Both things are true at once, and they always will be.

Here's how to watch the keynote.

Ignore the demos. Demos always work.

Listen for three things.

What's shipping TODAY versus "coming soon."

What it costs, because "we'll announce pricing later" means it's not real yet.

And who's allowed to use it, because access tiers are the story now, not features.

Then do nothing.

Don't touch your stack this afternoon.

Tomorrow you'll get the six checks run against whatever actually shipped, minus the hype.

That's what this newsletter is for. 🫡

OpenAI's very first DevDay was November 6, 2023...

Eleven days later, the board fired Sam Altman.

He was back within a week, and the company has never looked back.

The keynote is never the only drama that day. 🎭

To use OpenAI's next model, you'll need a hardware key.

Whatever gets announced today, the access story around GPT-6 Cyber is already reported, and it tells you where everything is heading.

To get in, you apply.

Your identity gets verified.

And you need a FIDO2 hardware security key, a physical device, to log in.

That's under a program OpenAI calls Daybreak Red, and a small group of vetted security customers are already testing the model in alpha.

Read that as a trend, not a one-off.

Anthropic locks Mythos behind verified cybersecurity and bio customers.

Claude Fable got the ID-and-selfie treatment over the summer.

Now OpenAI's most capable security model needs a hardware key in your hand.

The most powerful tools are being fenced off behind proof of who you are.

There's a genuinely good reason for it.

A model that finds vulnerabilities faster helps defenders faster.

It helps attackers exactly as fast.

The only way to hand that to the good guys and not the bad guys is to check who's at the door.

OpenAI's also putting money behind it, committing $1 billion in subsidized access to protect critical infrastructure and essential services.

So here's what this means for you, even if you never touch a security model.

The days of "sign up with an email and get everything" are ending at the top end.

The frontier is becoming a verified-access club.

If your business depends on the most capable tools, get your identity, your entity, and your paperwork in order now.

The door's getting a bouncer.

Be the one on the list. 🫡

Go watch the keynote with the three questions in your head...

And go keep your hands off your stack until tomorrow's breakdown. 🔥

Jeff J. Hunter

Founder, AI Persona Method | TheTip.ai

Get paid to solve problems like this → AI Certified Consultant

PS. Tomorrow's email is the DevDay six-checks... Whatever they ship, you'll know what's real, what it costs, and whether it's worth moving for. 👀

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