The feed doesn't know you, and YouTube refuses to let you browse

There are tens of thousands of talented people on YouTube right now making the most careful, deep, useful videos that have ever existed in their respective subjects. Restoration channels rebuilding 1980s computers from corroded boards. Bench scientists explaining their actual research. Cabinetmakers narrating a 14-hour build. Linux specialists with 800 subscribers writing better tutorials than the books I bought in 2003. The good content is out there, and most of it has fewer views in its lifetime than the slop on my home page got this morning.

I cannot find it.

I do not mean I have to search harder for it. I mean the platform is structurally hostile to finding it. There is no browse on YouTube. There is a search box and there is a funnel, and once the funnel has decided what kind of viewer you are, the search box is the only escape hatch. It's a bad one, because search is keyword-driven and the things I'm trying to find don't have the keywords I'd think to type.

That is not personalisation. That is a trap. And the trap exists because the platform's incentives and my incentives are different things, and the platform won.

How the funnel actually works

You click one video. Maybe it was a tired moment, or a strange recommendation, or a thing a friend sent you. The recommender system records the click and the watch-time and feeds you three more in that flavour. You watch one of those, half-attentively, and now the next batch is six in that flavour. Two weeks later your home page is a uniform smear of the thing you didn't even mean to be interested in, and the careful video about C compilers you were watching last month has fallen off the edge of the world.

The narrower version of the same trap is more insulting. I watch a single video, say a fifteen-minute piece on the restoration of a fifteenth-century painting, and the recommended sidebar instantly converts into a multiple-choice quiz drawn from the same channel. Not "other people working on conservation," not "adjacent crafts I might also care about," not "what else has this presenter been thinking about." Just more of the same channel's back catalogue, sorted by whatever the engagement model thinks I'll click next. There is no sense that I watched the video twenty minutes ago or eight months ago. There is no recognition that the channel's other content might be nothing like what I just watched, or that I might have landed on this video because someone linked it, not because I have decided to become a viewer of fifteenth-century-painting-restoration content for the rest of my life. One video is enough. The funnel narrows immediately, and if I don't want to spend the next three weeks being assumed to be a conservation enthusiast, my options are exactly two: clear the watch history and start over, or block the channel outright so the system stops treating it as a signal, which punishes the creator for the platform's bad inference. Neither is a reasonable interface. They are both nuclear options dressed up as preferences.

The system does this because it's optimised for engagement: watch-time, return frequency, completion rate, comment volume. None of those are preference. They're measurable proxies for what kept you on the platform last time you were there, and the reason any of it matters is straightforward. More time on the platform means more ad impressions, and more ad impressions means more revenue. That is the entire mechanism. YouTube doesn't make money when you find what you were looking for and leave. YouTube makes money when you keep watching, regardless of what you're watching, and the funnel is the most efficient instrument ever built for keeping people watching. The product is your attention. The metric is engagement. The customer is the advertiser. The system is structurally optimised to maximise the thing being sold to the customer, which is not the same as the thing you went there for. Once you see the incentive plainly, every other behaviour of the home page falls into place.

This is well-trodden ground. Tristan Harris and Eli Pariser and Cal Newport have all said versions of it. What I want to land here is a sharper point: the funnel does not just narrow what you see. It removes the verbs. You cannot browse. You cannot survey a topic. You cannot ask "who else is doing careful work in this space" and get a list. The platform has no surface for that question. The Subscriptions tab is a fig leaf because most consumption flows through the home page, and the home page is the funnel. The "Trending" tab is engagement metrics with a different label. There is no "show me good videos in this niche, ranked by quality, that I haven't seen yet" button, because building one is structurally adversarial to the business model.

The good stuff is buried, on purpose

Here is the part that genuinely upsets me. There is no shortage of quality on YouTube. There is a creator drought as the algorithm sees it, which is to say a shortage of videos that maximise watch-time per dollar of ad inventory, but there is a thunderous abundance of careful, slow, accurate, beautiful work. Hundreds of thousands of videos with under a thousand views, made by people who clearly know what they are doing, sitting unindexed by anything except their own titles.

We have so many talented people in the world. Why does the public have to consume the slop the platforms hand us? The answer is structural and not flattering. Quality is not a metric YouTube can measure. Engagement is. So engagement is what the system rewards, and quality is what gets buried. Not because the platform hates quality, but because quality and engagement are different signals and the platform only knows how to look at one of them.

The slot-machine framing is right but incomplete. The slot machine is in the home feed. The real damage is the absence of any other interface. There is the funnel and there is the search box. That's it. A platform with eight hundred million hours of content and no browse layer is not a library. It's a casino with one game.

TikTok, Shorts, Instagram, leave entirely

YouTube is the focus of this rant because YouTube has good content underneath the slop. The funnel is hiding something worth finding. The other platforms aren't hiding anything. They are the slop.

TikTok's For You Page is the purest form of the engagement-funnel ever shipped. Subscriptions barely matter. Account history barely matters. The algorithm decides what you watch, full stop, and the format (vertical, sub-minute, autoplay) is engineered to bypass the part of your brain that decides whether to keep watching. YouTube Shorts is the same product, copied. Instagram Reels is the same product, copied again. None of these have a discovery interface because none of them want one. Watch the next one is the entire user experience.

There is no "use these mindfully" advice that survives contact with these formats. Get them off your phone. The reach you give up by leaving is real, and the agency you reclaim is bigger.

What to do, explore the indie web

This is the part where I'm supposed to give you a clean "here's how to fix it" answer, and the honest version of that answer is: the structural problem is not fixable by user discipline. The platforms are designed to win this loop. But there are things that help, and most of them are old.

Read RSS. Personal blogs, newsletters, technical writing. They still exist, and they're better than they were in 2008. NetNewsWire on Mac and iOS, Feedly or FreshRSS in a browser, a hundred small clients on Android. Subscribe to people, not platforms. The feed is what you put in it.

Read books. Not e-books on a platform that tracks your reading speed and recommends the next purchase. Actual books, from authors you found through the indie web rather than through Goodreads's recommendation engine.

Listen to podcasts in a podcast app. Not a recommendation app pretending to be a podcast app. Overcast, AntennaPod, Pocket Casts. Subscribe to a feed. Skip the ones you don't want. There's no algorithm here.

Use YouTube on your terms, minimally, until something better exists. Subscriptions tab, not home page. Search by exact creator name. When you find a careful channel with 700 subscribers, watch the back catalogue. Refuse the recommended sidebar.

And because the structural problem is real, I am building something. It's called Content Curator. It's a discovery layer over YouTube that doesn't use view counts or engagement signals as ranking inputs. Videos are placed into a concept graph by what they're actually about, and surfaced by experiential signals (informative versus fluffy, clear versus confusing, calm versus high-energy, beginner-friendly versus complex, up-to-date versus outdated) derived from comment analysis and structured user impressions, not popularity. You browse a node, see the videos and the connected concepts, follow a link sideways into an adjacent topic, and decide where to go next. The graph is the interface. There is no home feed.

It's at concept-and-design stage right now, no code shipped yet, and it is going to take a while. But the answer to "there is no browse on YouTube" is "build a browse." If you'd like to follow along, the system concept document is going up on this site soon, and the first deliverable, a single-video classification tool, will follow.

Closing

The next time you scroll your YouTube home page and think "why is the algorithm so good at knowing what I want", stop. The algorithm doesn't know what you want. It knows what you'll click. It built the home page by extrapolating from clicks you barely remember making, and it's serving you a portrait of your worst, laziest reflexes. The talented people are out there. The work is out there. The platform has decided you don't get to find it without paying for the privilege of being marketed at first.

We can do better than this. The platform isn't going to ship it. We'll have to build it ourselves.