BLOG - 2021-09-10


Hypergraph lessons and future

One year has passed since we launched Hypergraph (Beta) and it is time for us to take stock, share what we learned from this experiment, and let you know where we're going next.

Almost a year ago we introduced a way to publish research modules with Hypergraph (Beta).

Research modules are a publishing unit where the researcher decides what gets published with links between subsequent modules. These can be complete manuscripts but also datasets with linked code, audio files with linked transcripts, 3D models with linked revisions, or anything else you may produce during your research.

One year has passed since we shared that work with you publicly and it is time for us to take stock, share what we learned from this experiment, and let you know where we're going next with Hypergraph.

Taking stock

In the past year, you downloaded the Hypergraph desktop application over 600 times, with 157 profiles created and 50 research modules shared. We are grateful for the support you've given the project and the patience as we worked through many bugs.

This is liberate science's first big project and we're learning a lot. Especially with respect to product strategy, we'd like to share some, despite some of those lessons sounding super obvious in hindsight. Knowing lessons and actually learning them are two vastly different things.

Some lessons learned

Lesson 1: Desktop is the past

Installing a desktop application is a major hurdle for most researchers. We underestimated this because of our conviction in what we were building.

Institutional IT-restrictions may prohibit researchers from installing anything at all, or operating systems throw warnings because we are a new player with no clear trust mechanism. You might've gotten a "Windows protected your PC" warning for that reason.

Moreover, we did not fully consider what it means to meet researchers where they are. A desktop application is only a fraction of where researchers are --- a lot of work happens elsewhere. In the field, in wetlabs, and other environments where a PC is not the device of choice. The way we built Hypergraph did not provide us with established paths to provide a mobile experience either, providing no perspective to meet researchers where they are in the future.

Lesson 2: Invest in product, not technology

We built Hypergraph on cutting edge technology: the Hypercore protocol. This is a promising peer-to-peer protocol that is in active development. We owe the core inspiration of Hypergraph to this technology.

In the wake of this cutting edge technology, we overlooked that we needed to tread new ground in every direction as we built on top of it. We had to partner with GEUT to create our own Content Distribution Network (CDN) --- mobile applications are not (yet) feasible with this technology.

As Alan Cohen frames it in his book Prototype to Product, we committed the deadly sin of investing in technology instead of investing in the product. After launch a year ago, Hypergraph started stagnating as we tried to resolve issues of the technology instead of investing in our product (e.g., account recovery mechanisms).

Additionally, we would have to keep committing this product sin if we further develop Hypergraph on top of this cutting edge technology. We do not have the scale to do so, nor is it our mission to be a technology company.

Lesson 3: Rolling releases all the way

Another issue with desktop applications is the overhead in updating.

In 2021, we updated Hypergraph once a month, even though we might've implemented improvements the day after the last update. This means there is a lag in getting features to you. We would prefer to roll out any improvements to you as fast as possible by using rolling releases. Including bug fixes.

Additionally, for each update, each researcher would have to download the Hypergraph installer of circa 100MB. This means that with the total 600 downloads, we used up circa 60GB of bandwidth --- and that's without using the application at all.

As we start considering our climate impact with all our business choices, a desktop application has linear impact: More users and more releases resulting in higher bandwidth use, and in a bigger climate impact.

We want to move towards a decreasing marginal climate impact as more researchers start publishing modules.

What next

We are even more convinced than before that publishing research modules is worthwhile. We've spoken to many researchers in the past year and it's clear publishing modules resonates with you. We've shown it is possible, now it is up to us to take the next step and reduce the hurdles to do so.

We admit: The desktop version of Hypergraph has not proven as frictionless as we had imagined.

In light of the above, we are announcing that we will transition Hypergraph to a web based platform in the next five months. We are taking this opportunity to evolve the brand name as well (Hypergraph was already trademarked...).

By February 2022 at the latest, we will be ready for you to publish research modules on the web. If you'd like to help us in those five months, sign up for testing here.

We bring the same vision, mission, and values to publishing research modules on the web version as we did to the desktop version: Privacy first, maximizing knowledge distribution, and 100% free to publish and read research modules. Our manifesto remains our unchanged guiding document.

In light of recent discussion regarding surveillance capitalism, we also want to highlight we are not looking to produce another platform to collect data about users. This is about nurturing researchers, not exploiting them. We will share our source code and decisions on how to safeguard your privacy and how we build our web platform so you can make up your own mind whether we are worthy of your trust.

Hypergraph desktop will be supported until the end of the calendar year, but without any additional features as we transition to the web application that will be launched February 1st, 2022 at the latest.