In 2026, almost everyone we talk to assumes the open web is in decline.
The signals are easy to point at. Discovery moved into platforms. AI assistants now answer half the questions that used to send people to a search result. Mobile usage is dominated by apps with no addressable URLs. Browsers are quieter than they were a decade ago, and the new generation of "AI-native" tools is being built in chat windows, not on pages.
And yet, here we are, building a company whose entire surface area is https:// URLs that work without an account. We get asked, often politely and sometimes not, why we are betting on a venue everyone else seems to be leaving. The short answer is that we think they are misreading the moment. The long answer is this essay.
A short history of where the audience went
The web was never the only place software lived, but for about fifteen years it was the most interesting one. From roughly 2005 to 2020 the browser was where the new ideas first appeared - Gmail, Google Maps, Figma, Notion, Stripe's docs, GitHub's diff view, every interesting piece of consumer SaaS. The shape of those products taught a generation what good software felt like.
Then the audience scattered. Some of it went into apps. Some went into platforms that do not let you link out. Some went into LLM chat boxes, where every answer is a re-rendering of the open web with the credits removed. The browser stopped feeling like the frontier and started feeling like a fallback for things the apps did not do well.
This is the moment a lot of people are reading as decline. We read it differently. The audience did not leave the web because the web got worse. They left because the first surface of every new product moved somewhere else. The web stayed exactly as good as it has always been at the thing it is uniquely good at: being a permanent, linkable, no-permission venue for utility.
The four properties no other venue has
If you stack the web up against the alternatives - native apps, platform feeds, AI chat - and ask what it has that they do not, you get a list that has not changed in twenty years. We list it constantly because we keep forgetting how rare it is.
It is addressable. Every page has a URL. Every URL can be linked, bookmarked, shared, archived, embedded, and indexed. You cannot do any of that with a screen inside an app. You cannot do it with a chat reply. The URL is a primitive that the rest of the internet has quietly forgotten how to live without.
It requires no permission. A user does not need an account to land on a webpage. The page does not need approval from a store to be shipped. There is no review process, no policy compliance review, no 30 percent take. The web is the only mainstream venue where the publisher and the visitor can transact directly.
It is durable. A URL written down in 2008 will, with reasonable care, still resolve in 2028. No app store has that property. No social platform has that property. No AI chat thread has that property. Durability matters more than people credit, because every other compounding effect on the internet - SEO, citations, word of mouth, internal linking - depends on it.
It is composable. A page can link to another page on a different site without asking. An author can quote, cite, embed, or extend someone else's work without negotiating an API contract. This is so basic to how the web works that we forget it is unusual. Every other venue is a closed graph.
A small number of products in the world will require the things that closed venues offer - heavy local compute, deep OS integration, a curated audience. But the vast majority of useful software does not. It just needs to be reachable, fast, and trustworthy. The open web is still the best venue in the world for that, and it has no real competition for the role.
The "AI killed search" misread
The most common 2026 argument against the open web is that LLMs have replaced it. A user asks a chatbot how to merge two PDFs, and the bot tells them - no link click required. Why publish a tool when the bot has eaten the use case?
This argument has the chain of causation backwards. LLMs are downstream of the open web. They were trained on it. They continue to require it for grounding, citation, and freshness. When a model gives a confident answer about a niche tool, it is reciting something somebody published - usually for free, usually on an indexable URL, usually with no expectation of being credited.
The open web is the substrate AI runs on. A world without an open web is a world where AI assistants get worse over time, because their training data calcifies and their grounding signal disappears. The companies building those assistants know this. That is why they are quietly investing in indexing the web, partnering with publishers, and treating "answer this from a URL" as a first-class capability. The web is not being replaced. It is being re-rendered into a new interface, and that interface still needs the underlying pages to exist.
The right way to think about LLMs as a publisher is: they are a new distribution channel for the same work. We still write the pages. We still build the tools. The bots quote us into a new context, and a fraction of those quotes turn into URL clicks, and the URL clicks land on pages we still control. The economics of that channel are different from search, but the underlying job - make a thing on the open web that is worth being quoted - is exactly the same.
What we are doing about it
The bet we are making is not abstract. It shows up in the choices we make every week.
- We publish every tool as its own URL. No SPA-style routing tricks, no hash fragments, no "single experience that handles all the things." If a tool exists, it has a page, and that page is reachable without context.
- We treat the page as the deliverable. The page renders fast, works without JavaScript for the critical content, has a real
<title>, has a real<meta description>, has structured data. We do this because it is good for users and because it is the format LLMs and search engines both reward. - We refuse to gate anything behind an account that does not need to be gated. This protects the property the web is uniquely good at - no permission required to use the thing.
- We separate the durable layer from the experimental one. The tool URLs at /apps and the brand URLs at /blogs are written to last. The marketing layer can change as often as we want, but the addressable layer of Rune is treated like infrastructure.
- We invest in cross-linking inside our own properties. Every tool links to adjacent tools. Every essay links to relevant tools. Every part of the Rune ecosystem - Apps, Hub, Learn, Career - is reachable from every other part. The graph compounds.
None of these decisions are clever. They are just the same decisions a serious publisher made in 2010, applied to a new decade. The reason they feel novel in 2026 is because most of our peers stopped making them.
The long horizon
We do not know exactly what the internet will look like in five years. We are not certain how much traffic will arrive via AI surfaces versus search versus direct. We are not certain which platforms will still be the dominant feeds. We are not certain how the economics of advertising-supported content will shake out.
We are certain about one thing: the underlying primitive - a URL, on the open web, that anyone can visit without asking permission - will still be valuable. Probably more valuable, not less, as everything around it gets more enclosed.
That is the bet. Build a lot of small, durable, useful things on the open web. Make them as good as we can. Let the rest of the internet rearrange itself around them. The pages we ship today are the foundation we will be standing on in 2031, regardless of which platform is the one everyone is panicking about that year.
If that thesis is right, we win slowly but we win for a very long time. If it is wrong, we have spent our years building a large pile of free tools that helped real people do real work, which is not a bad consolation prize.
We can live with either outcome.