Meta's AI Agent Play + Unmasking Anonymous Accounts

General Catalyst seeks 10 billion dollars across multiple investment vehicles, including

growth and early-stage vehicles.

The firm has expanded beyond traditional VC into broader financial services, and this

would rank among the largest capital raises by a venture firm in recent years.

10 billion is like a massive war chest.

I think the signals are trying to compete with the mega-funds at every stage, right?

Early to late.

Yeah, exactly.

The question is whether they can actually deploy that capital effectively.

AI infrastructure bets need longer time horizons and bigger check sizes.

If this goes through, it's a clear bet that the venture landscape is consolidating around

a few massive multi-stage players.

Totally, and the fact that they're spreading it across several vehicles gives them flexibility,

but it also means they need deal flow at scale.

General Catalyst in talks to raise around 10 billion dollars across multiple investment

vehicles, including growth and early-stage vehicles.

The firm has expanded beyond traditional VC into broader financial services, and this

would rank among the largest capital raises by a venture firm in recent years.

10 billion is like a massive war chest.

I think the signals are trying to compete with the mega-funds at every stage, right?

Early to late.

Yeah, exactly.

The question is whether they can actually deploy that capital effectively.

AI infrastructure bets need longer time horizons and bigger check sizes.

If this goes through, it's a clear bet that the venture landscape is consolidating around

a few massive multi-stage players.

Totally, and the fact that they're spreading it across several vehicles gives them flexibility,

but it also means they need deal flow at scale.

Swedish legal AI startup LaGora just raised $550 million at a $5.55 billion valuation led

by Excel.

This is their third raise in the past year.

They're opening new U.S. offices in Houston and Chicago and targeting 300 U.S. employees

by year end.

Third raise in a year?

That's wild.

They're clearly in land grab mood.

Absolutely.

And it's part of a bigger trend.

European AI startups have raised over $9 billion in just two months of 2026, following

a record $21.7 billion last year.

They're seeing billion dollar rounds for companies like NScale and Jan LaKoon's New Venture,

Advanced Machine Intelligence Labs.

I think what's interesting here is that legal departments are moving from pilots to full

deployment.

That's a huge shift.

LaGora is racing to dominate before the big software incumbents wake up.

Yeah, and the pace suggests investors see a narrow window.

If this pans out, they could own enterprise legal workflows.

Third rolled out Ford Pro AI, a chatbot embedded in its telematic software for commercial

fleet managers.

It analyzes vehicle data like speed, seat belt use, engine health, and can recommend ways

to cut fuel costs, provide vehicle specific insights, or even draft summary emails for

supervisors.

Okay, so Ford's targeting over 840,000 paid subscribers with this.

That's a huge install vase.

And for fleet managers, operational efficiency translates directly to savings.

Right.

Ford says they built it on manufacturer-grade vehicle data using a multi-agent architecture

designed to reduce hallucinations.

They didn't name the underlying AI model, but they claim it's more reliable than standard

large language models because it's using structured, fleet-specific data.

I'm curious how well it actually works in practice.

Like, can it really prevent hallucinations or is that just marketing?

Good question.

But Ford's joining other automakers racing to embed AI everywhere from design tools to

consumer apps.

The commercial fleet play is smart because it's a paid market with clear ROI.

Robotics startup ROTA AI raised $450 million at a $1.7 billion valuation to build AI models

trained on millions of online videos instead of expensive teleoperation data.

They've already tested their system with off-the-shelf parts in a major automotive factory

and plan to license the tech while building their own humanoid robots.

So the big idea is that training on internet videos gives robots better generalization

than teleoperation, where humans remotely control the robot.

That makes sense, right?

More diverse data.

Yeah.

CEO Jagdeep Singh says if a phone's orientation changes, teleoperation models might fail.

But their model has seen so many examples at different orientations that it can adapt.

The round was led by PremG Invest with backing from Coastal Adventures, Tomasik and John

Doerr.

PremG sees potential for robots as a service model, which could be huge for U.S. manufacturing

reshoring.

If you want sophisticated manufacturing back in the U.S., you need flexible automation like

this.

Totally.

The real world is messy and wrote as betting they've cracked the generalization problem.

If they're right, this could unlock a lot of industrial use cases.

Researchers from ETH Zurich and Anthropic built an AI agent system that correctly identified

up to 68% of anonymous accounts with 90% precision.

The system analyzes writing quirks, biographical details, and posting patterns to match accounts.

And it cost less than $2,000 to run, identifying each profile for just $1 to $4.

Though anonymity online might be dying?

Pretty unsettling.

Well, the researchers warn that pseudonymous posting may no longer protect identity as AI systems

get more capable and access larger data sets.

Performance improved with more data points.

For example, the system linked Reddit accounts mentioning one movie, only 3% of the time,

but jumped to nearly 50% success when users mentioned 10 or more films.

The low barrier to entry is what worries me.

Like journalists, dissidents, activists relying on anonymity now face real risks.

And it's cheap enough that way more actors can try this at scale.

Yeah, the economics have totally changed.

Every clue and AI found could have been found by a human investigator.

But now it's automated end-to-end.

That's the shift.

That's today's Pivot 5.

Like and share Pivot 5.

Thanks for listening.