Superpower Daily

Superpower Daily

Your reading space

Sign in
OpenAI Pilots Local Academy Trainers as Content Engagement Tops...Review Finds Only One of Five Opus 5.5 Benchmark Scores Rules O...Google Releases Gemini Speech Models With Custom Voices and Lin...Writer Says Meta’s Muse Synced His Messages After He Denied Acc...Alibaba Releases Qwen-Audio 3.1 and Cuts Voice API Prices by Up...Basecamp Research Raises $140M to Advance Six AI-Designed Thera...OpenAI Pilots Local Academy Trainers as Content Engagement Tops...Review Finds Only One of Five Opus 5.5 Benchmark Scores Rules O...Google Releases Gemini Speech Models With Custom Voices and Lin...Writer Says Meta’s Muse Synced His Messages After He Denied Acc...Alibaba Releases Qwen-Audio 3.1 and Cuts Voice API Prices by Up...Basecamp Research Raises $140M to Advance Six AI-Designed Thera...
Swipe

YouTube Will Let Viewers Create AI Video Feeds From Written Requests

Viewers will be able to save multiple feeds as home-page tabs without changing YouTube’s main recommendations. Creation on web and mobile is set to begin rolling out next month.

By 2 min read
YouTube Will Let Viewers Create AI Video Feeds From Written Requests
YouTube Will Let Viewers Create AI Video Feeds From Written Requests

Listen to this story

The audio brief

About 1:09
0:001:09
Read transcript
YouTube will let viewers turn a written request into a custom recommendation feed, saved as its own tab on the home page. The key distinction is that these feeds won’t replace or rewrite the main recommendations. Viewers can choose when to open them, and can keep several tabs for different interests or situations. Creating one starts with a prompt. You can ask for a topic, but also describe what to include, leave out, or prioritize—and the moment you’re watching for. YouTube’s examples include video podcasts for a 30-minute train commute, or relaxing commentary for winding down. Google’s Gemini model will use those instructions to shape the recommendations. The feature offers another way into YouTube’s catalog, which the company says contains more than 20 billion videos. Instead of starting a fresh search, a viewer could return to a feed built around a recurring interest or need. But the size of that catalog doesn’t tell us whether detailed prompts will produce useful results, or whether requests to exclude certain videos will be followed. YouTube says feed creation will begin rolling out on web and mobile next month. The central test is whether these tabs deliver recommendations that feel genuinely tailored to the request, rather than simply serving up the usual videos for a broad topic.

Story brief

3 key points

YouTube plans to start rolling out custom recommendation feeds on web and mobile next month, letting viewers use Google’s Gemini to create persistent home-page tabs from written instructions. Requests can specify what to include, exclude, or prioritize, and may describe viewing contexts such as a commute rather than just a subject. The feature adds opt-in destinations alongside the existing recommendation feed, not...

  1. 01

    Examples include video podcasts for a 30-minute train commute and relaxing commentary, showing prompts can describe a viewing context as well as a topic.

  2. 02

    Viewers can create multiple feeds, each saved as a separate tab pinned to the top of the home page.

  3. 03

    YouTube says its catalog contains more than 20 billion videos; the feature offers another starting point besides a fresh search.

YouTube is giving viewers a way to describe the videos they want recommended, but not to rewrite the home feed they already use. Its newly announced custom feeds will turn a written request into a separate tab, while the main recommendation feed stays in place. Viewers can choose when to open a feed built around their instructions.

From a description to a tab

Creating a feed starts with a prompt box. A viewer describes the videos they want, and Google’s Gemini AI model uses that request to build the feed. The description can go beyond a topic: YouTube says people can specify what to feature, what to leave out and what to prioritize.

YouTube’s examples show how broad those requests can be. Someone could ask for video podcasts suited to a 30-minute train commute, or for relaxing commentary to watch while unwinding. Rather than naming only a subject, those requests describe a moment or purpose. How closely the recommendations will match that intent remains to be seen.

People will be able to create multiple feeds. Each one will have its own tab pinned at the top of the YouTube home page, so a viewer can return to a particular request without rebuilding it each time. YouTube plans to begin rolling out feed creation on web and mobile next month.

The catalog behind the request
More than 20 billionVideos on YouTube

YouTube VP Emily Moxley gave this company-supplied figure while explaining the case for custom feeds.

A starting point beyond search

Moxley framed custom feeds as a way to explore that vast catalog. A viewer could make a standing request for a particular interest or situation, then return to it from the home page. That offers a different way to begin browsing than entering a fresh search query each time. It does not, by itself, show that a detailed instruction will produce more useful recommendations.

Where the viewer’s instructions end

Other services have offered ways to shape a feed. Bluesky has an AI feed-building tool called Attie; TechCrunch also points to feed-building features at Threads, Instagram, X and Spotify. The specific choice YouTube is offering is between its usual recommendations and separate feeds shaped by written requests. The named alternatives do not tell us how well YouTube’s version will follow a viewer’s instructions.

The useful test will be whether a detailed request actually produces a feed that feels different from an ordinary topic search, especially when a viewer asks it to exclude certain videos. For now, YouTube has described the instructions people can give, not shown how reliably the resulting feeds will honor them. That leaves the quality of the recommendations as the question for the rollout.

Sources

  1. techcrunch.comYouTube will let you build your own algorithm with AI | TechCrunch

Loading discussion...

YOUR READING SPACE

Notifications