What happens when an AI agent starts doing more than answering questions?

What if it can file paperwork, sign documents, pay invoices, negotiate contracts, or interact with government systems on your behalf?

Estonia thinks we’re about to find out.

The country is exploring a new framework that would allow AI agents to receive their own digital identities and operate inside government and business systems as recognized actors.

Think of it as giving AI agents their own version of a digital ID card.

Not citizenship.

Not human rights.

But an official identity.

Why does this matter?

Today, AI agents can perform many tasks.

But legally, they don’t really exist.

They can’t sign documents.

They can’t authenticate themselves.

They can’t take responsibility for actions.

And they can’t be audited in the same way people and companies can.

Estonia’s proposal aims to change that.

Under the concept, every AI agent could receive a unique digital identity linked to a person or organization.

That would make it possible to:

• Verify which AI performed an action
• Track decisions and transactions
• Define permissions and limits
• Create accountability for mistakes or abuse

In short, AI agents would become traceable digital actors rather than anonymous software.

The bigger picture

This might sound futuristic.

But we’re rapidly moving toward a world where millions of AI agents will book meetings, manage finances, file reports, negotiate purchases, and interact with other software without human involvement.

The question isn’t whether these agents will exist.

The question is how society identifies and governs them.

Estonia may be the first country trying to answer that question at a national level.

What I’m watching

For years, digital identity was built around humans.

The next decade may require digital identity for machines.

And if Estonia succeeds, future AI agents may have something that sounds surprisingly familiar:

A name. An ID number. And a trail of responsibility.

See you tomorrow,

TSOA

Hey there,

One of the biggest limitations of AI agents is that they can recommend products but they can’t actually buy them for you.

That may be about to change.

Visa and OpenAI have announced a partnership that will allow ChatGPT to make purchases on your behalf.

The idea is simple:

You tell ChatGPT what you want.

“Find me the best noise-canceling headphones under $150.”

Instead of just showing recommendations, ChatGPT could:

• Compare products
• Find the best deal
• Add the item to your cart
• Complete the purchase using your Visa card

All without you leaving the conversation.

How does it work?

According to Visa, users will remain in control through:

• Spending limits
• Approval requirements
• Merchant restrictions
• Fraud protection and monitoring

The goal is to make AI-powered shopping both convenient and secure.

Why this matters

This isn’t really a payments story.

It’s an AI agents story.

For the past two years, AI has mostly helped us create things:

• Text
• Images
• Code

Now AI is starting to take actions on our behalf.

Booking flights.

Ordering groceries.

Buying products.

Managing subscriptions.

And eventually handling many of the repetitive digital tasks we do every day.

The bigger picture

The real competition in AI may no longer be about who has the smartest model.

It may be about who owns the transaction.

If millions of people start shopping through ChatGPT instead of Google, Amazon, or traditional websites, the entire e-commerce ecosystem could change.

The companies that control AI agents could become the new gatekeepers of online commerce.

And Visa wants to make sure it remains the infrastructure powering every transaction.

We’re moving from AI that answers questions to AI that gets things done.

That shift may be bigger than most people realize.

See you tomorrow,

TSOA

Hey there,

While most AI companies are busy building better chatbots, Jeff Bezos is aiming at something much bigger.

His new startup, Prometheus, just raised $12 billion at a $41 billion valuation to build what Bezos calls an “Artificial General Engineer.”

The goal isn’t to write emails, generate images, or answer questions.

The goal is to help design and manufacture real-world products like:

• Jet engines
• Medical devices
• Consumer electronics
• Robots
• Spacecraft

In other words, Prometheus wants AI that can help engineers build physical things faster and better.

Why this matters

Today, creating a new jet engine or medical device can take years or even decades.

Not because engineers are slow.

Because the process is incredibly complex.

Bezos believes AI can compress that cycle dramatically, allowing companies to move from idea to production 10x faster or more.

The bigger shift

For the past few years, AI has mostly been about generating digital content.

Text.

Images.

Code.

Prometheus is betting that the next trillion-dollar opportunity is applying AI to the physical world.

Not replacing writers.

Not replacing designers.

Augmenting engineers.

If it works, future products could be designed, tested, optimized, and manufactured in a fraction of the time it takes today.

What I’m watching

The most interesting part isn’t the funding.

It’s the vision.

Most AI companies are building assistants.

Prometheus is trying to build a co-founder for engineering itself.

And if they succeed, the companies that make airplanes, cars, chips, robots, and medicines may never operate the same way again.

See you tomorrow,

The Son of Anton

Subscribe now

Hey there,

One of the biggest AI stories of the year just dropped.

Anthropic has been forced to shut down access to its newest AI models, Fable 5 and Mythos 5, after the Trump administration imposed export controls citing national security concerns. The directive prevents foreign nationals from accessing the models, including people living inside the United States and even some of Anthropic’s own employees.

What’s especially remarkable is that Anthropic didn’t just block international users.

The company completely disabled access to the models while it figures out how to comply with the order.

Why the concern?

According to government officials, Mythos 5 has advanced cybersecurity capabilities that could potentially be used to discover and exploit software vulnerabilities at an unprecedented scale. Reports suggest the government became concerned after learning about possible methods for bypassing some of the model’s safeguards.

Anthropic strongly disagrees with the decision.

The company argues that the alleged vulnerabilities are limited, already understood, and not meaningfully different from capabilities available in competing frontier AI models. Still, it complied immediately with the order.

Why this matters:

• This is one of the first major AI export restrictions targeting models themselves rather than chips.
• It creates a new precedent for governments directly controlling access to frontier AI systems.
• It could accelerate the global AI race as countries push to develop sovereign AI capabilities.
• It raises difficult questions about who gets access to the most powerful AI tools in the future.

For years, AI regulation focused on GPUs and computing power.

Now the models themselves are becoming strategic assets.

And that may change the AI industry more than any model release.

See you tomorrow,