The web most developers interact with daily, the one optimized for search rankings and engagement metrics, is not the whole web. Running alongside it is something older in spirit and newer in parts: personal sites, hand-built pages, protocol-level experiments, and communities that predate the platform era. Kevin Boone’s article argues that this “small web” is substantially larger than most people assume, and after spending time with the communities and tools involved, that argument is more solid than it first appears.
Defining the territory
The small web lacks a single authoritative definition. In the broadest framing it includes any web presence operating outside the logic of algorithmic amplification: personal blogs running on cheap VPS instances, static sites deployed from home machines, Gopher holes, Gemini capsules, Neocities pages, and the tilde communities where users share shell access on public Unix servers. These spaces share an operational logic: the person publishing owns the content, sets the agenda, and answers to no platform.
This is worth distinguishing from the IndieWeb movement, which is a specific community with formalized standards. IndieWeb promotes owning your own domain, publishing on your own site first, and syndicating to platforms second. The community describes this as POSSE: Publish on Own Site, Syndicate Elsewhere. It has produced real infrastructure: Webmention is a W3C recommendation, Micropub is a publishing API with multiple implementations, and IndieAuth handles decentralized identity. But IndieWeb is the organized fraction of something much larger and deliberately less organized.
The actual scale
Neocities alone hosts over 700,000 sites. Not all are active, but the platform has grown steadily since its 2013 launch as a deliberate successor to GeoCities, which at its peak housed tens of millions of pages before Yahoo shut it down in 2009. Neocities is not just nostalgia; it runs a modern static hosting service with a public API, WebDAV support, and a social follow system.
The Gemini protocol, which launched in 2019 and defines a minimal alternative to HTTP with its own content format (“gemtext”) and mandatory TLS transport, now has several thousand known capsules indexed by crawlers. GeminiSpace maintains an active directory. Gopher, which many assumed died when HTTP took over in the mid-1990s, maintains a genuine community; the Gopher Lawn project maps living holes, and the count runs into the thousands. The Tildeverse is a network of publicly-accessible Unix servers where anyone can request an account and serve HTML from a ~username directory. tilde.club, tilde.town, and several dozen other servers participate, each with active IRC channels, zines, and shared games.
What makes these numbers difficult to pin down precisely is that no central registry exists for the broader small web. A personal blog running Hugo on a DigitalOcean droplet doesn’t register anywhere. RSS readers discover it; search engines may or may not index it; audiences accumulate through links and word of mouth. This is the original web topology, before platforms absorbed everyone’s content into their own databases and made discoverability a function of advertising spend.
Why conventional metrics miss it
PageRank and its descendants favor high-link-count domains. Social media algorithms favor engagement velocity. Neither metric is designed to surface a single person’s hand-written page about amateur radio or medieval Latin translation. The small web is structurally invisible to the discovery mechanisms most people use, which is a measurement problem, not an existence problem.
This has produced a secondary ecosystem of small web-specific search engines. Marginalia explicitly deprioritizes commercial content and SEO-optimized pages in favor of personal sites and small blogs. The engine’s author, Viktor Lofgren, published detailed notes on the ranking approach: sites without JavaScript, cookie banners, and tracking scripts score higher; pages with high link diversity to small domains score higher still. Wiby restricts its crawler to pages that resemble the early web by design, no heavy JavaScript frameworks, no surveillance infrastructure. These are narrow tools with small indexes, but they find things that a Google search will not surface on page one, or page ten.
Webrings have also returned, which is a more interesting development than it sounds. The mechanism is simple: a group of sites each link to the next and previous member in a ring, so visitors navigate among them without a search engine as intermediary. The xxiivv webring, maintained by Devine Lu Linvega, connects a community of artists and developers working in unusual tools and languages. Several dozen other active webrings exist across topics from amateur radio to fountain pens. The infrastructure is minimal: a small JavaScript snippet and a shared member list. The social effect is that linked communities become discoverable to each other without algorithmic curation.
The protocol layer
One reason the small web persists is that its underlying technology has low operational overhead. A static site costs nearly nothing to run at scale. Gemini’s specification fits in a few pages; a complete server can be written in a few hundred lines of C or Go. Gopher is simpler still, dating to 1991 when the bandwidth constraints were severe and protocol minimalism was a survival requirement, not an aesthetic choice.
The 512kb Club catalogs sites under 512 kilobytes of total uncompressed page weight. This sounds like a performance stunt, but it’s a real statement about what the web could be if load time were actually the priority rather than ad tracking and A/B test scaffolding. The green tier is under 100 kilobytes. A typical news site in 2026 loads several megabytes of JavaScript, calls a dozen third-party APIs, and sets dozens of cookies before rendering any readable content. The contrast is structural, not cosmetic. A Raspberry Pi on a home connection can host a Gemini capsule serving hundreds of concurrent readers without strain. It cannot serve a Next.js application with a CDN, a database backend, WebSocket connections, and a real-time analytics pipeline.
This has practical implications for who can participate. Lower infrastructure costs mean that hobbyists, academics, and writers who have no interest in monetization can maintain a presence without running a business around it. The server economics of the small web keep the barrier low enough that the population doesn’t winnow down to people who can justify the spend.
RSS as connective tissue
If the small web has one piece of infrastructure that actually scales, it’s RSS. The format is over two decades old, was declared dead when Google Reader shut down in 2013, and continues to grow in active usage. Feed readers like NetNewsWire, Reeder, and Miniflux have active development and genuine user bases. Every static site generator worth using outputs a feed by default. ActivityPub, the protocol powering Mastodon and the broader Fediverse, borrowed heavily from Atom (RSS’s close relative) and extends the federation idea to social interactions.
RSS works because it inverts the platform model. Content doesn’t go into someone else’s database to be surfaced on the platform’s terms. The feed belongs to the publisher; subscribers pull from it on their own schedule. There is no algorithm deciding which posts get promoted. For someone who builds things and wants to follow a hundred small technical blogs, a self-hosted feed aggregator is still the most reliable solution. Nothing else gives you the same combination of completeness and publisher independence.
What “bigger than you think” actually means
Boone’s point isn’t that the small web rivals Google’s index in raw page count. It’s that the assumption of marginality is wrong. The people building and reading personal sites form a substantial and intentional community, not a shrinking remnant from before social media existed.
The tooling has matured considerably. Hugo, Zola, and Eleventy make static site generation accessible to anyone who can edit a Markdown file. Cloudflare Pages and GitHub Pages make deployment free. Gemini clients exist for every major platform, including terminal clients for people who prefer not to leave the command line. The Amfora terminal client and Lagrange graphical client are both actively maintained.
More telling is the nature of the participants. The small web draws developers, writers, academics, and hobbyists who could publish on Substack or Medium but choose not to. The reasons vary: control over data, no algorithmic pressure on posting frequency, the ability to write at length about niche topics without optimizing for virality. These are not people unaware of platforms. They are people who have made a deliberate choice about where their content lives and who controls its distribution.
The web was always going to contain both kinds of content. The commercial layer grew so fast and so visibly through the 2010s that it overshadowed everything else. The small web didn’t contract during that period; it kept building, kept developing new protocols, and kept maintaining the older ones. That’s a more durable outcome than the “personal web is dead” narrative that circulated after Facebook and Twitter absorbed so much of the public’s publishing instinct. The protocols and communities sustaining it are worth understanding even if you spend most of your time building things for the other web.