Today we’re launching a new website, which represents something much bigger: a new way of building infrastructure that matches the evolving needs of knowledge institutions. Applying lessons learned from nearly a decade of building and maintaining knowledge tools, we’ve completely re-architected the way we work over the last year. From the policies and locations of the underlying cloud providers to the structured, portable, and preservable knowledge storage layer to the inexpensive-to-maintain, self-hostable applications themselves, the way KF builds infrastructure today is simply more sustainable, cost-effective, and innovative.
Taken together, we believe these changes represent a methodology for creating a new class of knowledge systems, built for purpose.
Our updated site reflects this way of working. It centers what we build and how we build it, in the hope that you’ll see opportunities for your own community, including solutions to problems that have long been too expensive to take on. Here’s what’s behind it.
From platform to purpose
Until very recently, the only financially feasible way to build knowledge infrastructure was to build one big application and ask everyone to use it. That’s how PubPub has worked for nearly a decade: one platform, thousands of communities, each one bending its work a little to fit the affordances and limitations of the tool.
That financial reality is changing. Much of what made custom work expensive was infrastructural: migrating and aligning data, auditing performance, untangling legacy code, wiring up hosting. Over the last eighteen months, new tools have made that kind of boilerplate work much easier to accomplish. Jobs that used to take a team months now take days. That means more of every budget goes to the parts that make a project yours: the editorial, design, and research decisions that only the people doing the work can make. A small budget can now get a community what it actually needs. Work that used to require venture-scale capital is within reach of the libraries, societies, labs, and publishers who actually steward public knowledge.
For us, these changing economics offer a powerful new path toward our mission to make information useful. We can now build every community we work with a system shaped around its specific needs. And every one of those systems is built on, and adds to, a shared foundation of open-source tools and infrastructure that keeps getting cheaper, more stable, and more accessible to the next public-minded group that wants to build something.
That’s what we mean by knowledge systems, built for purpose. Purpose-built, because each one is shaped around the community using it. Built for a purpose, because each one adds to a larger public record: open, portable, and accessible to whoever needs it next.
Start with the knowledge, then build the tools
Every new project we build starts in the same place: a structured, versioned collection of the knowledge itself, stored in Underlay, our open knowledge sharing protocol and registry. All of the articles, datasets, and metadata start as versioned, permanently addressable records described by a reusable schema and accessible via an enterprise-grade API.
Everything else — the website, journal, search interface, accessible edition, the exportable archive — are outputs built on top of that structured collection. Building on a structured foundation lets those outputs stay fluid. They can change as the community’s needs change, and there can be as many of them as there are audiences. All while the collection underneath stays simple, stable, and durable.
One of the first projects we built this way is the Open Encyclopedia of Cognitive Science. Because the site is built on well-described structured data, we rebuilt it in weeks rather than months. We’ve begun moving other PubPub communities to the same system, and it’s giving them far more room to align their outputs with their work while benefiting from and contributing to open infrastructure.
Building many specific tools also teaches us what they have in common. When we see the same need come up across projects, we pull it into shared infrastructure that lowers the cost for everyone. Underlay is the first and most central of those pieces, but there will be more.
This specific-to-general approach to building is deliberate. The only way we know to build infrastructure grounded in people’s needs is to start by solving those needs. We like abstractions as much as anyone, and we know how satisfying it is to imagine the perfect system, build it, and wait for everyone to come. Like many of you, we’ve seen how rarely that works. So we work the other way around: solve real problems for specific communities, then look across many of them for what they share. If the goal is to make information useful, you first have to understand what “useful” means, in all its variety.
This approach has another benefit, too: it allows other builders to seamlessly plug into and expand our work for their own purposes. Too often, we’ve seen open projects become bogged down and bloated by the coordination required to build integrations that serve only a small percentage of users. By starting with open structured data, building specific tools to modify and display it, and releasing everything open-source, we’re creating tools that allow for the kind of modularity that our ecosystem has always strived for in theory, but struggled to build in practice. We’re already collaborating with other infrastructure builders on joint projects using this approach, and are excited to explore more opportunities.
Why choose us
The communities we work with could hire a contractor, or sign up for a platform from a large tech company. They come to us because they want something that lasts, and that doesn’t close them off from the rest of the ecosystem. In practice, that means:
- Custom, but connected. Built for your community’s specific needs, and part of a shared public knowledge ecosystem.
- Hosted, but not proprietary. We’ll run it for you, on privacy-respecting infrastructure, and everything we build is self-hostable on generic hosting providers. You’re never dependent on us to keep it running.
- Built for your community, and whoever comes after. Data is structured, versioned, and preservable from day one, not as an afterthought.
- Built by infrastructure experts. We’ve run public knowledge infrastructure at scale for nearly a decade, and we’ve reworked it under pressure.
Built on difficult lessons
That last point has come to define our new way of working. In June 2025, an unexpected funding withdrawal left us with a runway short enough to feel existential. We responded with a deep technical overhaul of PubPub’s infrastructure: paying down nearly a decade of technical debt, optimizing expensive queries, and bringing outside services into the main codebase. Together, those changes cut PubPub’s server costs by more than 99%.
We then launched the PubPub Sustainability Fund, with a goal of $300,000 to cover PubPub’s core costs for the foreseeable future. We haven’t reached that number, and it turns out we don’t need to. After several months on the new infrastructure, PubPub’s real operating costs are low enough that the fund as it stands can cover PubPub’s servers indefinitely, assuming a nominal 4% annual return on the fund’s balance. Thank you to everyone who contributed.
The lessons learned re-architecting PubPub under pressure completely changed what sustainability means for us, and directly led to our new methodology. Sustainability is no longer a fundraising target, it’s a design constraint. Every system we build and host has to have a sustainability model for its server costs that can outlast the ups and downs of any particular funding cycle.
As a result, we’re concluding the PubPub Sustainability Fund in its current form. In its place, we’re opening the Infrastructure Longevity Fund. It supports the shared infrastructure that works across communities and underpins the custom work built on top: Underlay, the open-source tools, and the common services every project relies on, PubPub included. It’s how we make sure the work communities entrust to us has a very long runway ahead.
No lock-in, by design
As a result of these lessons learned, we’ve made portability a core part of our methodology, so our partners are never forced to scramble for alternatives again. With a project’s open-source code repository and its Underlay collection, everything we build can be reproduced on generic hosting infrastructure in its entirety. Not a static export, not a partial backup, but the entire project is redeployable by anyone willing to take on its technical maintenance.
This approach keeps the power where it belongs: with the communities whose work we support. They decide how it grows, where it lives, and who it’s shared with. We plan to be good stewards of that work alongside them for a very long time. But nothing we build requires anyone to rely on the values or longevity of our organization in perpetuity.
This is true of every new project we take on. It’s also available to communities that have been on PubPub, as their interest and needs demand. If that’s you, get in touch.
What’s next
We’re working on several projects and demos that we’ll be sharing publicly over the coming months on our new site and through our newsletter. We hope you’ll see ideas in them that help you picture solutions to your own challenges.
We’d love to help you build those solutions. And if they simply encourage you to demand technology that serves your needs, from us or anyone else, that’s a good outcome too.
If you have something you’d like to build, write to us at partnerships@knowledgefutures.org.