😸 Meta's AI hiring spree is over

gm legends. It’s Sunday funday.

In this edition: 

• What the latest Meta reorg means for third-party AI models
• Why a minor car crash can be good for your product launch
• How creators should protect their content from LLMs
• And, as always, the most popular products that went live this week 

Grab the carafe. We’ll bring the memes.

P.S. Launching soon? We’d love to hear about it → editorial@producthunt.co 🫶

Remember Meta’s massive AI hiring push we told you about in June? Nine-figure offers. Champagne dinners at Zuck’s house. Massage therapists for your dog (we’re speculating). Well, so much for that.

Multiple outlets reported that the hiring spree is over amid a company-wide reorg. But it’s the below-the-headline news we’re interested in: 

1. The team that built Llama, Meta’s open-source model, is being dissolved and several prominent team members have left. The failure of Llama to make waves is ostensibly what prompted the hiring push in the first place. Shortly after launch, Zuckerberg began poaching employees from OpenAI, Google, Apple, xAI, and Anthropic. In June, he hired Scale co-founder Alexander Wang to be Meta’s chief AI officer.
2. Wang and co are ditching Meta’s Llama-based frontier model, Behemoth, and starting from scratch. Per The New York Times: “Behemoth’s release was delayed last spring after disappointing performance tests.”
3. The new AI model may not be open-source. In fact, the company is reportedly even “activelyexploring using third-party artificial intelligence models” for its own products — either by building on open-source models or licensing closed ones.

The takeaway is that no one knows which building philosophy Meta will embrace as it tries to catch up in the AI race. And while Meta probably won’t try to buy OpenAI — something Elon Musk tried to team up with Zuckerberg to do earlier this year — we’re not betting our Zuck Bucks on it.

By Neha Suresh, co-founder, April (which launched this week)

It's the end of May 2025. By this point, I’ve applied to Y Combinator loads of times, but have always gotten the same answer: “No.” So when my co-founder, Akash, and I sign up for the MCP Hackathon, getting into YC is the furthest thing from our minds. We just want to build something we’d actually use, and see what happens.

We call the resultant product Inbox Zero. It’s an AI assistant that can answer emails by voice. I’m stuck in traffic most mornings, and so is Akash, so the idea of clearing our inboxes without touching a screen feels like magic. We record a demo and dent Akash’s car while recording, just minutes before the deadline. Judges love the demo. Then they announce the winner… us. And the prize? A direct YC interview.

LLMs got you(r ad traffic) down? Don’t frown. Just take a sledgehammer to some server far…just kidding. But Daniel asks: “Where should bloggers / creators pivot if ad traffic drops (due to LLMs)?” Do you up your game and put your content behind a paywall AI can’t break through, do you pass around a hat on Patreon, or try something else altogether?

The Product Hunt community is coming with ideas, from shifting “towards a multi-platform model” to changing how bloggers package their content. The answer depends on determining what (or who) the product is.