From 4–6 October 2026, the global open science hardware community met at Heritage Garden in Lalitpur, Nepal, hosted by Karkhana Samuha, for the GOSH 2026 Gathering. Sponsors included the Open Science Hardware Foundation and Seeed Studio, which published its account of the three days on 26 October — which is how it reached us, three weeks after the event.
We are flagging the gap up front: this is a report on something that has already happened, there is no open call attached, and no date has been announced for the next gathering. We are covering it anyway because one of the workshops is the best teaching exercise on hardware documentation we have come across, and it costs nothing to run.
The premise
The case for open science hardware, as Seeed puts it:
Scientific tools should not be limited to well-funded laboratories or specialized institutions. Open science hardware makes it possible for more people to build, adapt, repair, and share the tools they need to investigate the world around them.
Across three days, sessions covered water testing, soil health monitoring, homemade solar cells, disaster shelter construction and accessible hardware.
What is notable structurally is the format, and it will be familiar to anyone who has been to a good art-and-technology gathering rather than a conference:
Rather than focusing solely on finished products or formal presentations, the gathering creates space for people to share work in progress, learn from one another, and experiment with ideas that may develop into new tools, research methods, or community-led projects.
Work in progress, not results. That is a deliberate choice and a rarer one than it should be.
The paper hat workshop
Here is the exercise, as described. In under 30 minutes, each team:
- designed a paper hat
- wrote its manufacturing document, so a stranger could reproduce it
- pitched the design to the group
- then replicated the chosen hat following that document — with the original creator acting as quality control
The debrief mapped those four steps onto real electronics manufacturing: engineering verification, bills of materials, pilot verification, and mass production.
Read step 4 again, because it is the one doing the work. You do not grade your own documentation. Somebody else builds from it while you watch, and you are the one who has to decide whether what they produced is the thing you designed. Every ambiguity in your document becomes a physical object in front of you, made by a person who followed exactly what you wrote rather than what you meant.
That is the entire problem of open hardware, staged in half an hour with paper.
It is also, and this is why it caught our attention, the same test we have spent this week applying to other people’s work. NVIDIA shipped 2,575 3D assets this month with a hashed bill of materials and a per-asset licence file, and declared none of it where the platform could read it. Scenario published a dataset whose README matched its file tree exactly, including the one scene that was broken. The difference between those two is not effort, care or resources. It is whether a stranger, following only what was written down, arrives at the right thing.
The paper hat workshop teaches that in thirty minutes and the industry keeps relearning it at scale.
”Vibe Coding with Wio Terminals”
The other session worth noting put AI-assisted coding into a hands-on hardware setting — writing code for Wio Terminals with model assistance, in a room, with the devices present.
Seeed’s framing of it is more careful than the session title suggests:
The goal is not to replace the process of learning or engineering, but to give more people a starting point for experimentation—and a way to turn curiosity into something tangible.
Which is the defensible version of the claim. The gap between having an idea and beginning to build is where most projects die, and it is a different gap from the one between beginning and finishing. A tool that lowers the first barrier is useful without needing to be good at the second.
For a community whose core value is that people should be able to build, adapt and repair their own instruments, there is a real tension here that the write-up does not explore: generated code you do not understand is not obviously repairable by you, and reproducibility by a stranger — the paper-hat standard — is harder when part of the method was a prompt. Both sessions ran at the same gathering, and holding them together is the interesting problem.
For the calendar
GOSH is now on our conference list, and we had not covered it before. The specifics as verified on the Open Science Hardware Foundation’s own page: 4–6 October 2026, Heritage Garden, Milap Road, Lalitpur 44600, Nepal, hosted by Karkhana Samuha. The page describes the event as being “in Kathmandu”, and the venue is in Lalitpur, which adjoins it.
No participant count, no country count, no next-gathering date and no open call are stated. If the pattern of previous gatherings holds, an announcement would be the thing to watch for; we will check periodically rather than guess at a cadence.
Try the exercise
You do not need a gathering. Two people, paper, 30 minutes:
- Design something simple enough to make in five minutes.
- Write the document. No photographs of your own build — words and drawings only.
- Hand it over. Say nothing while they work.
- Inspect what comes back and decide whether it is your design.
Then apply the same standard to the README of whatever you are about to publish.
Sources
- Seeed Studio at GOSH 2026 in Nepal: Making Open Science More Accessible
- GOSH 2026 — Kathmandu, Nepal — Open Science Hardware Foundation
- Open Science Hardware Foundation
- Regional GOSH, Karkhana, Nepal — GOSH community forum