WhoKnows.
← All briefings
3 stories

Daily Briefing — June 22, 2026


01

Kevin Roose is leaving The New York Times to build an AI media company

Substack →
Career & skills

Kevin Roose, one of the most prominent tech journalists in the world, is leaving The New York Times after more than a decade. He and Casey Newton, his co-host on the Hard Fork podcast, are launching their own media company. Hard Fork will air its final episode in August 2026, and then the two of them are going independent. Newton will keep writing Platformer, Roose will keep doing books, and together they will build something new that they actually own.

Roose and Newton are framing this as building a media company that takes AI seriously — one that covers AI progress as a real editorial beat, with both enthusiasm and skepticism, rather than just riding the hype cycle. Two of the best-known voices in AI journalism looked at their options and decided the move was to leave institutional media and build independently. That says a lot about where they think the leverage is going.

The broader pattern: top talent in media keeps concluding that ownership and creative control matter more than institutional backing. The tools to go independent have never been better, and the cost of staying inside a structure that limits what you can build keeps going up. Roose himself put it well, quoting his old mentor: say yes to adventures or you will lead a very dull life.

SO WHAT

The best people in media are choosing ownership over institutional safety. If you are in any knowledge work field, the calculus between staying and building your own thing is shifting fast, and it is worth honestly evaluating which side of that equation you are on.


02

Epic wants to let you bring your Fortnite skins to other games

The Verge →

Epic Games just announced that Unreal Engine 6 will support cross-game skin interoperability, meaning your Fortnite cosmetics could show up in other games built on the engine, and third-party developers could create skins that work inside Fortnite too. This is the most concrete step Epic has taken toward the metaverse idea it has been pitching for years without much to actually show for it.

The skins are the demo, but the real problem Epic is trying to solve is infrastructure. Getting digital assets to move cleanly between different games — respecting ownership, rendering correctly — is hard. Epic is using Fortnite as a live stress test for a system it hopes the broader developer ecosystem will adopt. If it works, this becomes a proof of concept for something much bigger.

Who owns in-game purchases and where those assets can travel has big implications for how platforms get built, how players engage across products, and how revenue models evolve. Epic is betting that respecting player purchases across games is a competitive advantage — and that bet tells you where the industry is going.

SO WHAT

If you work anywhere near gaming, virtual goods, digital identity, or platform design, the technical and business model questions this raises are about to become very real conversations in product and engineering rooms.


03

The Atlantic created a searchable database of the music used to train AI

The Verge →

The Atlantic just made it a lot easier to see whose music ended up in AI training datasets without anyone's permission. Reporter Alex Reisner dug into four datasets being used to train AI models and built a fully searchable public database out of them. Two of those datasets are enormous, we're talking 12 million and 9 million tracks respectively. The other two are smaller but still clock in at over 100,000 songs each. These things have been downloaded thousands of times, and both Google and Stability AI have confirmed using them in published research papers.

The licensing situation is a mess. Some of these sources, like the Free Music Archive, are perfectly legal to stream for personal use. Commercial use requires proper licensing. That distinction gets glossed over when engineers are racing to build the next model. "Freely available on the internet" and "free to use as AI training data" are not the same sentence, even if they look similar at a glance.

The same logic applies to written content, images, and code — not just music. The legal and ethical frameworks around what AI can legitimately train on are still being written in real time. People who understand that nuance will be more valuable than the ones who assume anything downloadable is fair game.

SO WHAT

If you work anywhere near AI products or data pipelines, understanding the difference between publicly accessible data and legally usable training data is quickly becoming a core professional competency, not just a legal team problem.

ACTION ITEM

Search your own name or your organisation's content in The Atlantic's database to understand firsthand what this exposure actually looks like, then bring that context into your next conversation about how your team sources training data.