YouTube's new AI agent will run your channel's back catalog

Share
YouTube's new AI agent will run your channel's back catalog

Made on YouTube 2026 is the platform politely telling creators to hand more of the job to its own AI — and two industry stories today show what the AI economy is built on when nobody reads the fine print.

YouTube is shipping an AI agent that optimizes a creator's channel in the background. Announced at the annual Made on YouTube event, the Studio agent scans your back catalog for videos gaining new relevance or trending, proposes fresh thumbnails and titles to ride the wave, and can assemble sponsor pitches from your channel's audience data. Viewers will feel it too: creators can now test three full versions of a video against small audience segments, with YouTube auto-promoting the winner after seven days, while a dynamic thumbnail tool serves different cover images to different audience segments. Amjad Hanif, YouTube's VP of creator products, frames it as "time that goes back into the creativity" — but the company declined to share any data showing the tools actually improve a channel's numbers.

The more user-facing shift: custom feeds let you describe your own algorithm. Type what you want — "relaxing commentary videos to unwind with," a podcast feed for your train commute — and Gemini builds it, pinned to the top of your home page, rolling out on web and mobile from next month. The idea Bluesky popularized, and Spotify launched the very same day, is now the default posture of every big feed: with 20 billion videos on the platform, the recommendation engine is negotiable. Meanwhile Ask YouTube, the conversational search feature, is becoming a shopping companion with product comparison tables — usage jumped more than 500 percent since December, to over 140 million people in June.


Nscale's IPO filing doesn't mention the customer that built it. According to the Financial Times, the 192-page S-1 for the Nvidia-backed AI cloud's US listing never names ByteDance — which accounted for 73 percent of Nscale's $33 million in revenue in 2025. Only an appendix nods to Spring, ByteDance's Singapore subsidiary, whose May 2025 deal put 2,304 Nvidia B200s into a converted crypto-mining data center in Glomfjord, Norway, backing a $105 million loan from Macquarie. The arrangement let ByteDance reach Nvidia chips it cannot buy in China through a gap in US export rules — legal, the FT notes, but a regulatory and reputational exposure an IPO prospectus would rather not spotlight. Nscale expects its largest customer to fall below 20 percent of revenue this year as its Microsoft and Anthropic contracts take over. We covered the pre-IPO cash pile last week — Nscale lines up $3.5B in pre-IPO cash with Nvidia at the table — and the omission tells you the rest.


Meta's Muse claims 500,000 users in week one — and Meta admits whose homework it copied. Internal data seen by The Information puts the agent at more than half a million users and 250,000 daily actives in its first week after the September 8 launch, with the app at number one on the US App Store. The awkward part: Nat Friedman, who runs product at Meta Superintelligence Labs, conceded Muse is "definitely heavily inspired as a product by OpenClaw" — the open-source agent it ships with nearly identical files of, down to the SOUL.md personality definition. When users pressed him on the copies, Friedman replied that OpenClaw's creator "got those things exactly right." The adoption numbers are Meta's own and unaudited, but a superintelligence lab shipping a consumer agent by cloning an open-source darling's personality — and bragging about it — may be the most honest product story of the year. We tracked Muse's slower climb earlier — Meta's agent Muse hits No. 2 on the US App Store — slowly.

What to watch: whether YouTube ever publishes data on what its agent actually does to watch time — and whether IPO regulators ask why Nscale's biggest customer was a footnotes-only affair.

Would you hand YouTube's background agent the keys to your channel, or does optimizing for the platform that owns both the algorithm and the tool feel like a rigged loop? Tell us in the comments.

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t