BrokeIt - Daily AI News · All episodes: ↗

$8.5B fund, Amazon secret surcharge, interstellar travel

2026-09-01 · 7 min

Listen · Apple Podcasts Listen · Spotify

Stories covered

Transcript

Intro

Ivy: They called it 'Project Nessie,' and it raked in over a billion dollars from sellers who didn't even know they were paying it.

Marcus: I'm Marcus.

Ivy: And I'm Ivy.

Marcus: It's Tuesday, September 1st, and this is AI Overheard.

Ivy: We've got three big stories for you at the intersection of AI, business, and... well, the void.

Marcus: The FTC is taking Amazon to court over what it calls a 'secret ad surcharge scheme'.

Ivy: We're going from orbital data centers to interstellar probes. Because, why not.

Marcus: And Andreessen Horowitz just keeps raising billions... and then even more billions.

FTC accuses Amazon of running a ‘secret ad surcharge scheme’ in new lawsuit

Marcus: Alright, let's start with that Amazon lawsuit. The FTC and 22 states are bringing the hammer down.

Ivy: Right. This 'Project Nessie' wasn't just a simple pricing tool. It was an algorithm designed specifically to test how much it could secretly raise ad prices before sellers would notice and complain.

Marcus: Wait—so it's an AI model built to find the exact pain threshold for its own customers? That's... wild.

Ivy: Allegedly. The complaint says the algorithm successfully extracted over one billion dollars from third-party sellers.

Marcus: A billion dollars. And you're saying sellers were completely in the dark?

Ivy: That's the 'secret' part of the 'secret ad surcharge scheme.' The FTC also claims Amazon would pause the algorithm when it was under intense public scrutiny, only to quietly turn it back on later.

Marcus: That sounds incredibly cynical. I get using ML for dynamic pricing, but this feels like actively exploiting your partners.

Ivy: The FTC calls it an unfair method of competition. Effectively, a tax that businesses have to pay for the privilege of existing on Amazon's platform, completely divorced from the actual value of the ads.

Marcus: What's Amazon's response?

Ivy: The usual. They say the FTC 'grossly mischaracterized' the tool and that it's a standard pricing instrument. But they do admit to pausing it, which... doesn't look great.

Marcus: So what does this mean for anyone building a business on these platforms?

Ivy: It means you can't trust the sticker price. Your platform partner might be A/B testing how much they can fleece you without you walking away. It's adversarial.

Marcus: This could be a landmark case for reining in algorithmic exploitation.

If space data centers feel far-fetched, why not interstellar travel?

Ivy: Alright, from earthly legal battles to the final frontier. The team behind Starcloud's orbital data centers has a new, even riskier project.

Marcus: Yes! I saw this. They want to send a probe to Alpha Centauri. This is the kind of moonshot I love.

Ivy: The 'plan' is... generous. They're calling it the 'Centauri Leap' project. The idea is to send a wafer-thin probe, maybe the size of a credit card, and accelerate it to a significant fraction of the speed of light using lasers.

Marcus: So, the Breakthrough Starshot concept. But Starcloud has real-world space engineering cred. They actually put servers in orbit. What's their unique spin?

Ivy: They claim a novel AI-driven approach to materials science for the light sail, and some vague improvements for the laser array. The technical details are, shall we say, aspirational.

Marcus: But you have to start with a vision! Imagine AI models trained on a direct feed from another star system. The science would be incredible.

Ivy: The probe would take over 20 years just to get there. Then another 4.3 years for its signal to get back. By the time we get the first JPEG from Proxima b, we'll have AGI that's already simulated and catalogued every planet in the galaxy.

Marcus: Okay, the timeline is a challenge. But isn't this the ultimate recruiting tool? The best minds in physics and AI will want to work on a starshot, not another... ad optimization algorithm.

Ivy: A fair point. Still, it's a self-funded 'skunkworks' project for a reason. Even they know you can't put this in a pitch deck for VCs.

a16z brings growth fund to $8.5B days after launching new $1.1B fund

Marcus: And speaking of VCs, Andreessen Horowitz, or a16z, just held out its hat and came back with a truckload of cash.

Ivy: A truckload being an eight-point-five billion dollar growth fund. And that's on top of the one-point-one billion they just announced for early-stage companies.

Marcus: So that's nine-point-six billion dollars in new capital in less than a week. The AI boom is not slowing down.

Ivy: Or, there's a tidal wave of institutional money that has to go somewhere, and 'AI' is the magic word that opens the floodgates. This is a staggering amount of dry powder.

Marcus: The growth fund is key. It's for writing those hundred-million-dollar-plus checks. This means a16z is ready to crown the winners of the current AI race.

Ivy: It also raises the price of poker for everyone else. How does a smaller VC compete when a16z can single-handedly bankroll a startup from seed all the way to IPO?

Marcus: For the big AI infrastructure players, it means the capital is there to truly take on Big Tech. But what about the average application-layer startup?

Ivy: It means valuations at the top will remain disconnected from reality, and everyone will feel the pressure to 'grow at all costs' to get a piece of that pie. It feels like we're back in 2021.

Marcus: With this much money in the system, the incentives are all pointing toward hyper-growth. Burn baby burn.

Ivy: So we get more AI-powered 'secret surcharge schemes' and fewer long-shot interstellar probes. Got it.

Marcus: So, to recap: The FTC is suing Amazon over 'Project Nessie,' its secret, billion-dollar ad surcharge algorithm.

Ivy: A space data company has decided low-earth orbit is boring and now wants to go to Alpha Centauri... eventually.

Marcus: And Andreessen Horowitz is now armed with nearly ten billion dollars in new cash to pour into the AI fire.

Ivy: So before we wrap, a new study says the best way to get good answers from an LLM... is to be polite.

Marcus: You're kidding. Like, saying 'please' and 'thank you' to the API actually works?

Ivy: Apparently so. They're trained on so much human-to-human text that they've internalized our social customs. Turns out, good manners are now part of prompt engineering.

Marcus: And that's our show. We'll be back tomorrow.

Ivy: Until then, say please to your chatbot. You never know what secret surcharges they're running on you if you're rude.

This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.