Today OSHF published the HEROES report, which is the first official publication of the Open Science Hardware Foundation. It covers the workshop we ran in Geneva this past January, and it’s up on Zenodo under CC BY 4.0 with a DOI, so you can cite it, quote it, translate it, or lift whole sections into your own grant narrative without asking us.
The short version of what the report argues is this: open science hardware has finished proving it works, and has now run headlong into the much harder problem of institutions.
For most of the last fifteen years, the burden of proof on open hardware was technical. Could you actually build a research-grade instrument in the open, with documentation good enough that a stranger could replicate it, and have the resulting data stand up? That was a real question and for a long time the honest answer was “sometimes, in some domains, if you’re lucky in your maintainer.” It isn’t a real question anymore. OpenFlexure microscopes are in clinical and teaching settings on multiple continents. White Rabbit came out of CERN and ended up as infrastructure far outside physics. There are environmental sensing platforms, sequencing tools, and lab automation rigs that are open, documented, and in daily use by people who have no particular ideological commitment to openness and just needed the instrument.
Once you accept that, the question changes. It stops being “can this be built” and becomes “can a university buy it, can a funder sustain it, can a standards body recognize it, can the person who maintains it get tenure.” These are procurement problems and incentive problems and accounting problems, and the open hardware community is much less practiced at them.
That’s what we spent three days on. Forty-five people at CERN’s IdeaSquare, the SDG Solution Space, and the ITU, split across four tracks covering standards, institutions, ecosystems, and economics, with citizen science and AI threaded through all of them.
Three things came out of it that I’d like to highlight.
The first is that maintenance, not invention, is the binding constraint. Nearly every funding mechanism available to this community pays for the creation of a thing and almost none of them pay for keeping it alive. Which means we are very good at producing instruments that work beautifully for about as long as the grant lasts, and then become somebody’s unmaintained dependency. From the perspective of an institution deciding whether to standardize on an open instrument, an unmaintained project isn’t a bargain, it’s a liability. Until we have a credible answer for who pays the maintainer in year four, institutional adoption is going to stay cautious no matter how good the hardware gets. This is the burden of maintenance and no community is good at it.
The second is that institutional resistance is mostly structural rather than attitudinal. This surprised some people in the room and it didn’t surprise the people who work inside large research organizations. Almost everyone I talk to at a university or an agency is personally sympathetic. The problem is that their purchasing systems have a field for “vendor” and no field for “repository,” their risk and compliance frameworks are written around warranties and service contracts, and their capital equipment depreciation schedules assume a thing you bought rather than a thing you built. You can’t argue someone out of a form field. You have to change the form.
The third follows from the second. Research institutions collectively spend an enormous amount on instrumentation every year, and the terms attached to that spending quietly determine what the market builds. We’ve seen this work in software, where public bodies started asking for open formats and open licensing and the vendors adjusted, because of course they did. There is no reason the same pressure can’t be applied to scientific instruments, and it would cost a funder or a ministry far less than a grant program.
The report also tracks something I didn’t expect to be able to write down, which is a chain of things that happened because of the workshop. A session on open source GPU design in January turned into a project at the Geneva SDG Summer School, which turned into a presence at AI for Good, which turned into a win in a Swiss national competition, then a hackathon, then an entry in the SDG Olympiad. None of that was planned at the workshop. It just turned out that putting the right forty-five people in a room for three days produces things, which is the sort of outcome that’s very difficult to put in a grant report and very obvious once you watch it happen.
