Why we built this
A company called UMA was entered in the French register in 2025. We searched it in 14 sources and 3 document it: the legal notices, the company register, and an estimate of its size. No funding, no published paper, no public code. It has one documented tie in our record — it stood at a conference beside another robotics company — and that is the whole of what we can show you about it. This is the ordinary condition of the industry we work in: a great deal happens, and very little of it is written down anywhere you can check.
The companies are real and they last. Wandercraft was registered in 2012, Pollen Robotics in 2016 — it now describes itself as part of Hugging Face — and Inbolt in 2019; none of the three carries a closure or a procedure in the register today. Around them the map holds 1,887 organisations and 2,568 documented ties, 896 of them in the French core it draws by default.
It is not all growth, and we do not draw it as if it were. Twelve of the 77 French robotics companies the legal notices cover entered a collective procedure since 2022, and the directory marks 22 organisations as being in one today and 13 as closed.
The conversation about this ecosystem happens in posts, decks and conference rooms. None of them lets you ask the question we kept wanting to ask: who is tied to whom, on what evidence, and since when. So we built a record that answers it, and that says plainly where we have not yet looked.
What we cover
We collect companies, laboratories, funders, people, papers and patents, starting from France and following the ties wherever they lead — today as far as the American public records and the worldwide patent register. We leave out two things on purpose: products, because it is the organisations that carry the ties, and any connection between two organisations that no public record states, however well we think we know it.
How we work
Nothing we publish is entered as a bare fact. Everything starts as a dated record somebody else wrote, and every step from that record to a line on the map is one you can follow.
Events and ties
We record events — a company register filing, a press release, a funding announcement, a conference programme entry, a shared publication, a patent — and a tie between two organisations is what a group of events, taken together, supports. A tie carries the kind of relationship (commercial, funding, ownership, a shared project, a shared paper, a co-owned patent, a person who moved, a conference co-presence), whether it has a direction, the year of its first and last record, and how many events support it.
Some kinds are direct — the evidence names something the two organisations did together — and some are weak: sharing a funded project, a paper or a conference does not by itself mean a working relationship. The map draws weak ties dashed rather than solid, and the data measures each kind on its own, because “who is central?” has a different answer for each.
Evidence
A tie’s evidence is a list of titled records, each with a link where we have one. Two organisations are matched to each other either by a shared identifier (a registry number, a stable cross-reference) or by an exact name match; the map’s evidence panel says which, tie by tie, so “matched” does not read as a stronger claim than it is.
The ledger of searches
Every organisation in the record carries a coverage record: which sources we searched it in, and which of them document it. This is the difference between two very different absences: nobody looked, and somebody looked and found nothing. Every row of the directory states an organisation’s coverage in a sentence before its ties, so a gap reads as one of those two states rather than as proof that a tie does not exist.
Retractions
A record we collected is never edited or deleted. When one turns out to be wrong, we append a retraction that names it and says why, and the next build leaves it out; the wrong record and the correction both stay readable. An organisation can also carry a disputed closure notice — a public record reads as a closure, but later records show continued activity — and we report that disagreement beside the organisation rather than applying it. State what was found and let you weigh it, rather than resolving it silently on the data’s behalf.
Claim status
Every figure and every claim on this site carries one of four chips:
- Built — a mechanism exists and runs (a page, a check, a collection step), independent of what it currently reports.
- Measured — a number computed from a specific, named dataset, on a specific date.
- Hypothesis — a number computed from data with known, stated gaps (uneven search coverage, a sample rather than a census); the map’s own figures carry this chip, because centrality on collected data partly measures how well something was documented.
- Open — no number yet: a question we state rather than answer with an invented figure.
Where the record started
Before we collected anything, we needed a list of who to look for. We built it by deep research: long, repeated sessions with Perplexity, Gemini and ChatGPT, asking each for the robotics ecosystem — its companies and its research bodies — in France first, then in Europe, then in the world, and reading the answers against one another. That gave a base of 1,658 organisations: a starting point, not a finding. A list assembled that way is as good as what those tools had read, it carries no dates and no sources, and it is biased towards whoever calls themselves a robotics company.
From that base we applied the process above: we searched every organisation in the public sources we hold, recorded what they document, followed the ties to organisations the base did not name, and kept the ledger of what was searched and what was found. The base itself is cited nowhere as a fact. What it is good for is a measure: the data page reports how much of it the evidence backs — 81.7% of it today, at least one dated fact from any source — and how far the record has grown beyond it.
How to read the map
A dot is an organisation; the bigger it is, the more ties we hold for it. A line between two dots is a documented tie — solid when the evidence names something they did together, dashed when it only puts them in the same place, a project, a paper or a conference. A pale wash behind a group of dots is a region: a cluster an algorithm found in the ties, not a market or a sector. Click a dot and you get its ties and the record behind each one; nothing is drawn that you cannot open. Everything else — which ties the map keeps, how the years work, what a region is — the map page explains beside the drawing.
Who we are
Two of us keep this record.
Tell us what is missing
An organisation that is not in the directory was not collected; a tie we do not draw was either never looked for or looked for and not found, and each row says which. If you know of something we have missed, write to contact@aquia.studio and tell us where it is written down — a filing, a programme, a press page, a paper. A source we can read is worth more to us than a correction we cannot check.
The data page is the honest account of what we have searched and what we have found, source by source, so you can see the shape of the gaps before you write.
Licence and legal
You may copy, quote and reuse the text, the data and the figures on this site, including commercially, as long as you credit us: they are published under the Creative Commons Attribution 4.0 International (CC BY 4.0). The code is published under the MIT License. The plates and drawings are the one exception: they are not covered by that licence, and all rights to them are reserved.
The legal notice, the privacy policy, the terms and the cookie notice say the rest.