Quick question before we get into this:

Have you received a call in the last month where you weren't sure if the person was real?

Login or Subscribe to participate

Two months ago, Nikita Bier predicted that iMessage, phone calls, and Gmail would become "functionally unusable" within 90 days.

Bier runs Product at X.

He was looking at what I've been watching in my work with founders: open-source tools that clone a voice from three seconds of audio, deployed at scale, for almost no cost.

We're about 60 days into his timeline now.

I've been getting calls from founders trying to figure out what this actually changes about how communication works, beyond the deepfake scam angle everyone's already covering

Honestly, growing up, I hated when people would call the landline and quiz you asking “pehchana?”.

I keep thinking about something that seems kinda related.

How do we verify that anyone is who they say they are?

What we built without noticing

For twenty years, your phone number was your identity.

When Mom called, you saw "Mom" on the screen, and that was enough.

Or when a colleague appeared on video looking and sounding like your colleague, you assumed they were your colleague.

We never had to think of these as verification systems because nothing could convincingly fake them.

He did what any of us would have done.

When the channel was the trust, you never had to think about the channel.

What changes

A tool can now clone a voice from three seconds of audio.

Another can put your face on a video call in real time.

The signals we built our professional lives around are now the easiest things to generate.

The scam headlines are true but it’s more than that.

The infrastructure that made remote work possible, that let us close deals over video and build relationships with people we'd never met, is losing its load-bearing function.

Same tech, two directions

The same tools eroding trust in video are being used to replace physical presence for routine communication.

Klarna's CEO sent an AI double to deliver his quarterly update. Our PGP acceptance emails had personalised video messages for each student using the same technology.

HeyGen lets you send a clone to multiple Zoom calls where it can think, talk, and answer from your knowledge base.

Synthesia says 150,000 users created AI avatars of themselves last year.

So in the same timeline someone uses deepfake tools to steal $25 million, another CEO is using the same technology to scale presence across a global workforce.

What I’m seeing

At one recruiting firm in Dallas, in-person interview requests jumped from 5% to 30% in a single year.

The global events industry hit $1.35 trillion last year, growing nearly 10%. 67% of executives increased their meeting budgets.

Gartner predicts that by 2028, one in four job candidate profiles will be fake. Already, 41% of IT leaders say their company has hired a fraudulent candidate.

They're all pointing at the same thing.

Presence used to be inconvenient. Now it's the only verification that holds.

Where the new stack is

In January 2025, the NSA published a report recommending that government agencies adopt something called Content Credentials.

The standard was built by a coalition that includes Google, OpenAI, Adobe, and Microsoft, among 200 others.

Every digital file carries a cryptographically signed record of who made it, when, and what tools were used. And it is being fast-tracked as an ISO standard this year.

Meanwhile the identity verification market has crossed $14 billion.

Vouched launched ‘Know Your Agent’, designed to verify AI agents the same way you'd verify a human.

Because the next version of this problem is not going to be limited to fake people on video calls.

It will additionally be AI agents acting on your behalf inside systems that still trust them by default.

Both tracks are trying to close the same gap. Neither has done it yet.

Where this goes

I think the most likely outcome is a sorting.

Anything one-directional gets automated, like updates, recorded walkthroughs, and announcements.

You show up for the moments that need two people thinking together in real time, or where presence itself is what's being communicated.

This was already happening before the trust problem accelerated it. The deepfake crisis is less a cause than a forcing function.

When the channel can be faked, the room becomes the proof.

I don't know exactly where this lands, but the direction seems clear enough.

Reply and tell me what you're seeing out there.
Nandini.


Keep Reading