Search for a way to browse the web without a query and you will find a scattering of small tools. Some are lovingly maintained, some are a weekend project someone never came back to, and a few have been quietly running for years. What almost none of them are is big.
For the broader context behind this business guide, read What We Lost When The Blogroll Died.
The usual reading of that is that these projects failed to take off. A better reading is that the smallness is doing the work. This piece surveys the categories fairly, then argues the constraint.
A note on how to read the tool names below. Every tool named here was opened and confirmed working on 6 September 2026, and each is described from what its operator publishes about it rather than from anything we assumed. Anything we could not open ourselves on that day is not named at all. We state no figures about anybody's size, collection or traffic, and where a maintainer does not publish a number we do not estimate one. The rest of our sourcing rules are at our source ledger.
What this is, and who it is for
A survey for someone who has hit the limits of query-based search and wants to know what else exists, plus an argument about why the alternatives share a shape.
It is not a ranking. This site does not rank anything, including its own category, and a leaderboard of small volunteer projects would be a strange thing to build while claiming that curation is hard. Several of the tools described below do things this site does not do and has no plan to do.
The history behind all of these is in the history behind all of these.
The tools people actually use, described fairly
Five categories. Most real tools sit in one and borrow from another.
Curated random-discovery buttons. A maintainer assembles a collection of sites they consider worth seeing, and a control sends you to one of them. The collection is the product; the button is the interface. indieblog.page is a tool of this kind: a personal project by Andreas Gohr which, in its own words, lets you randomly explore the IndieWeb by opening a random post from a personal blog. Checked 6 September 2026.
Search engines running their own index with different priorities. Rather than reproducing a general-purpose index, these crawl a deliberately different slice of the web and rank it by criteria the maintainer has chosen and often documented. Marginalia Search and Wiby are independently built indexes of this kind. Marginalia states that it runs its own index and crawler software, prioritises non-commercial content and uses no AI; Wiby says it is building a web of pages as it was in the earlier days of the internet, and puts a "surprise me" control beside its search box. Both checked 6 September 2026. The value here is that a different index returns genuinely different results, which is not something a wrapper around an existing index can offer.
Pure random-link toys. A button that sends you to something amusing from a hand-assembled list, with no pretence of usefulness. The Useless Web, made by Tim Holman, is the recognised example: one button, a browsable catalogue of what is in the collection, and a separate archive of the entries that have gone. Checked 6 September 2026. These are honest about being entertainment and are usually better maintained than they need to be.
Rings and link lists, in their current form. Themed rings and hand-maintained lists of personal sites continue to be built and run, generally by individuals or small groups, and generally around a specific subject or community. The XXIIVV webring describes itself as an attempt to inspire artists and developers to build their websites and share traffic amongst each other, and takes new members by pull request. The Hotline Webring accepts sites that show an individual personality and publishes what it refuses, including hate speech, bigotry, advertisement-focused sites and sites promoting LLMs. Ye Olde Blogroll, maintained by Manuel Moreale, is a hand-curated list of personal and independent blogs and discloses its sponsor on the page. All three checked 6 September 2026. What a webring was covers the mechanism.
Launch and showcase surfaces. Sites where makers post new work and an audience looks at what was posted recently. These serve a real need and have a structural property worth naming plainly: a launch is an event, and an event surface records that something was announced rather than that it still exists. That is a difference in purpose, not a criticism.
What each one optimises for
| Category | Optimises for | Best when you want |
|---|---|---|
| Curated random button | The quality of a single unexpected encounter | To be shown something good with no criteria of your own |
| Independent index search | Returning results a general index would not surface | To search, but away from the mainstream result set |
| Random-link toy | Delight per click | Five minutes of entertainment |
| Rings and link lists | Community coherence and mutual consent | To go deep in one subject among people who know it |
| Launch surface | Awareness of what is new | To see what has just been made |
Reading across that table is more useful than reading down it. These are not competitors offering different qualities of the same product. They are answers to different questions, and the reason nobody has consolidated them is that a consolidated version would answer all the questions worse.
Two places where other tools clearly do something we do not. An independent index will search, and this site has no search over the open web at all. A ring gives its members mutual, consensual membership of a community; a register makes a decision about a business without that business necessarily being involved, which is a weaker social contract and we should say so.
The constraint they share
Every tool above depends on a collection that someone assembled by hand, and every one of them is limited by the same thing: how much attention the maintainer has.
That is not a funding problem that a grant would fix, or a technology problem awaiting better tooling. It is a property of what curation is. To include something in a curated collection, a person has to look at it. Looking takes time. Time is fixed. The collection can only grow as fast as somebody looks.
And inclusion is only half of it. The other half is the part almost nobody does.
Why curation cost scales with the collection, not with the audience
This is the load-bearing argument, so here it is concretely rather than as a slogan.
Suppose a collection holds some number of entries, and suppose you want each entry to be true — the site still exists, still does the thing the description says, still belongs in the collection. Then:
- The cost of serving readers is roughly flat. Ten readers and ten thousand readers cost approximately the same, because serving a page is cheap and the collection does not change based on who is looking at it.
- The cost of keeping the collection true is linear in the collection. Every entry must be re-examined periodically. Double the entries and you double the recurring work, permanently.
- The two costs are paid by different things. Audience growth might eventually pay for something. Collection growth pays for nothing at all; it only adds obligation.
That asymmetry is the whole explanation. A project can grow its audience indefinitely at almost no marginal cost, and cannot grow its collection at all without acquiring recurring work that nothing pays for. So collections stay the size of one person's sustainable attention, and audiences vary wildly. The economics behind this, and why business models fund intake instead of accuracy, are in the cost of intake.
There is a real counterargument: automation. Software can check whether a URL responds, whether a page has changed materially, whether a domain has lapsed. That genuinely reduces the per-entry cost of noticing something is wrong. What it cannot do is decide whether a thing belongs — whether a site is a real independent operation or a shell with good typography — because that decision comes from reading it. Automation widens the funnel and does not remove the judgement, which means it lowers the constant and leaves the linear term standing.
What happens to the ones that try to scale anyway
There are only three ways past the constraint, and all three change what the tool is.
Drop the review. Accept submissions automatically and let volume solve the coverage problem. The collection gets large and the inclusion of any given entry stops carrying information, because nothing was decided about it. This is how a curated collection becomes an index.
Charge for inclusion. Now the recurring cost is funded, and the meaning of a listing has changed: it records that somebody paid, and a reader cannot tell from the outside which listings would have been included anyway.
Let it rot. Keep the entries, stop the checking, and let the collection describe a web that existed at the moment somebody last cared. This is the most common outcome by a distance, because it requires no decision at all and it is invisible until someone clicks.
Each of these is a rational response to a real cost problem. None of them is a moral failure, and describing them that way would be both unfair and inaccurate. They are simply what happens when a linear recurring cost meets no recurring revenue.
Why small is a feature, and what it costs the reader
If a collection is small because someone is refusing entries and removing dead ones, then smallness is evidence of the process working. That is genuinely a feature, and it is measurable: a small collection where every entry was decided is more informative per entry than a large one where nothing was.
Here is the cost, stated plainly, because the argument is dishonest without it.
A small collection will not have what you are looking for most of the time. It cannot be your only route. Its coverage reflects whatever its maintainer happens to know about, which means whole categories are missing for no better reason than that nobody looked at them yet. And its judgement is one person's or one team's, with all the taste and blind spots that implies.
A reader who understands that gets real value from a small collection. A reader who expects coverage will be disappointed and should be. Why random is not the same as curated goes further into what a curated surface is actually promising.
Where we sit, and what we do differently
Briefly, and last, because this is a survey rather than a pitch.
This site keeps a register of small independent businesses. What it does that is not standard in the categories above: it publishes an inclusion bar before deciding anything, at our entry bar; every entry names the person who reviewed it and the date of the last check; entries that stop trading become published tombstones rather than quiet deletions; and nothing about a listing can be purchased.
What it does not do: search, community membership, mutual consent from listed operators, or comprehensive coverage. A manually checked register stays small for exactly the reason this whole piece describes.
If a different tool in this survey suits you better, use that one. Several of them are good, and a route list that mostly is not us is at where else people find independent work. The rest of this cluster is at the independent web.
In Plain English
The tools for finding websites without a search query are all small, and that is not a sign they failed.
Curating costs time per entry. Serving readers costs almost nothing per reader. So a project's audience can grow freely while its collection cannot, because every added entry is a permanent recurring obligation to check that it is still true.
There are three ways around that constraint. Stop reviewing, and you have an index. Charge for entry, and a listing now tells you who paid. Stop checking, and the collection slowly starts describing a web that no longer exists. Most projects take the third route without ever deciding to.
Small collections are worth using if you understand what they are: a sample of somebody's attention, not coverage of a category. They will usually not have the thing you want. When they do have it, somebody looked.
Sources
Each entry below is recorded in the source ledger under the identifier shown.
- S-001 · indieblog.page — indieblog.page, Andreas Gohr. https://indieblog.page/ Checked 2026-09-06. Supports: the description of a curated random-discovery button, what its control does and who runs it.
- S-007 · Marginalia Search — Marginalia Search. https://marginalia-search.com/ Checked 2026-09-06. Supports: that Marginalia runs its own index and crawler software, prioritises non-commercial content and states it uses no AI. The older address https://search.marginalia.nu/ redirected here on the day of the check.
- S-017 · Wiby — Wiby: Search Engine for the Classic Web. https://wiby.me/ Checked 2026-09-06. Supports: the presence of a "surprise me" control beside the search box.
- S-016 · Wiby, about — About, Wiby. https://wiby.me/about/ Checked 2026-09-06. Supports: Wiby's stated aim of building a web of pages as it was in the earlier days of the internet.
- S-015 · The Useless Web — The Useless Web, Tim Holman. https://theuselessweb.com/ Checked 2026-09-06. Supports: the random-link toy description, and that the site keeps both a catalogue and an archive of departed entries.
- S-019 · XXIIVV webring — Webring, XXIIVV. https://webring.xxiivv.com/ Checked 2026-09-06. Supports: the ring's stated purpose and that members join by pull request.
- S-003 · Hotline Webring — Hotline Webring. https://hotlinewebring.club/ Checked 2026-09-06. Supports: its membership basis and its published list of refusals.
- S-006 · Ye Olde Blogroll — Ye Olde Blogroll, Manuel Moreale. https://blogroll.org/ Checked 2026-09-06. Supports: that it is a hand-curated list of personal and independent blogs with a sponsor disclosed on the page.
Try the stumble if you want to see what one deliberately small collection shows you.