The internet did not get boring. Our routes through it did.

YxYo — Sat, 08 Aug 2026 10:50:14 +0000

Disclosure: I’m the independent creator of Web Wander. I’m sharing the design idea behind it, not presenting an independent review.

There is a peculiar contradiction in modern browsing. The internet contains more pages, tools, archives, essays, games, experiments, and personal projects than any person could visit in a lifetime. Yet “going online” can feel like opening the same handful of doors.

That is not necessarily a failure of the internet. It is a consequence of the routes we use through it.

Feeds are built to reduce uncertainty. Search engines are built to answer questions we already know how to phrase. Both are useful. Neither is especially good at the moment before the question—the moment when someone is simply curious.

Serendipity used to be an ordinary side effect of browsing. A personal homepage linked to an essay, which linked to a tiny archive, which led to a tool made by one person for an audience of twelve. The route was inefficient, occasionally confusing, and often memorable.

The open web still contains those experiences. What disappeared was not the material; it was the default path.

Web Wander is one attempt to add that path back. It is a free Chrome and Firefox extension that opens an unexpected original website with one click. People can browse broadly, choose from more than 50 categories, or deliberately enter a keyword. The current editions draw from nearly two dozen discovery providers, then filter and de-duplicate candidates locally.

The point is not to argue that prediction is bad. Prediction is excellent when you want the next episode, the fastest route, or a relevant answer. The point is that not every online moment needs to begin with a known objective or end inside a feed.

Sometimes a good interface should ask a different question: what would you like to encounter?

Designing for that question requires more control than a pure random URL generator and less control than conventional search. Categories give curiosity a direction. A deliberate keyword provides a loose theme. Provider controls let people shape the source mix. Recent-history filtering reduces the frustration of repetition.

The result will not always be important. It should not pretend otherwise. A useful side quest, an unusual piece of art, a tiny game, an old archive, or a page about something absurdly specific can be enough. Discovery is valuable partly because its value is not fully known in advance.

The internet is not getting smaller. But without alternative routes, it can feel that way. A healthier browsing ecosystem may not require abandoning feeds or search. It may simply require keeping a few doors open to the unexpected.

Try Web Wander

Web Wander is free, and no Web Wander account is required.

Designing Useful Randomness for an Open-Web Discovery Extension

YxYo — Wed, 29 Jul 2026 18:03:35 +0000

Most “random website” buttons have the same failure mode: they are random for about three clicks, then the experience becomes repetitive, noisy, or unsafe.

I wanted to explore a different question:

Can serendipity feel surprising without feeling careless?

That question became Web Wander, a free Chrome and Firefox extension for discovering unexpected corners of the open web.

Randomness is not the product

Pure randomness is easy. Useful randomness is a design problem.

A good discovery session needs several kinds of variation:

different providers and provider types
different destination domains
broad wandering and focused wandering
safeguards against search pages, login gates, unsafe downloads, and obvious spam
enough memory to avoid immediately repeating what someone just saw

Web Wander currently rotates across 24 discovery providers. Those sources include independent-web indexes, curated collections, archives, academic resources, media discovery services, and small-web projects.

The extension also offers 56 categories, custom keyword wandering, web-era browsing, and dedicated video and music modes. The point is not to predict the perfect next page. It is to give someone a useful way to leave the same feed for a few minutes.

The interface has to explain the randomness

A mystery button can be fun, but it also creates uncertainty. Where will it send me? Why did I get this result? Can I influence the next one?

I ended up treating control as part of the discovery experience:

Browse all rotates across the enabled provider mix.
Browse by category narrows the journey without turning it into ordinary search.
Custom keyword mode lets a passing curiosity guide the next destination.
Web-era modes make time itself a browsing filter.
Provider controls let people enable or disable discovery sources.

That balance matters. Too little control makes randomness frustrating. Too much control turns the product back into a search engine.

Repetition is a state problem

A random choice can still repeat patterns. Even if two URLs are different, they may come from the same domain, the same kind of provider, or the same narrow corner of the web.

Web Wander keeps preferences, provider queues, a wander count, and recent destinations locally in the browser. It varies recent providers, provider types, and destination domains before navigating.

The useful lesson was that “don’t repeat the previous URL” is not enough. Repetition has several layers:

exact URL repetition
domain repetition
source repetition
source-type repetition
category repetition

Treating these as separate signals makes the experience feel much larger.

Privacy changes the architecture

Web Wander does not require an account. It does not send browsing history, destination history, cookies, referrers, advertising identifiers, or an extension-specific user identifier to a recommendation service.

Most wandering uses provider routes, public catalogs, or fixed non-personal discovery requests. When someone deliberately uses custom keyword mode, that keyword may be sent to compatible search and discovery sources so candidate results can be assembled. The useful candidate pool and ranking remain local.

This constraint removes a familiar growth shortcut: building a progressively richer behavioral profile. But it also keeps the product aligned with its purpose. A tool for escaping predictive feeds should not quietly build another predictive feed.

Discovery should send people outward

Another design choice was to open discoveries on their original websites.

The destination is not converted into a card inside Web Wander. It is not wrapped in an engagement feed. The extension is a doorway, not a replacement for the open web.

That makes the product less “sticky” by conventional metrics, but stickiness is not always the right goal. Sometimes the most honest success metric is whether the tool helped someone leave the tool.

What I am testing now

The project is live on Chrome and Firefox.

The questions I care about next are:

Does controlled randomness stay useful across a longer session?
Which categories produce genuine curiosity rather than novelty clicks?
Is the privacy explanation clear before installation?
Where does the provider mix feel uneven?

If you try it, I would value blunt feedback on discovery quality, repetition, and whether the controls make the experience clearer.

Disclosure: I built Web Wander. It is free, and this post explains the product and the design decisions behind it.

DEV Community: YxYo