A search is people and aircraft working on one problem: teams on the hill, a control room, people who ring in, and a swarm overhead. Meshfare is being built to listen to all of them, weigh who said what, mark the ground that matters and share out the drones, with a person deciding anything that matters.
Ground team 2 · their phone
Fresh boot prints going up the gill.
Seen first-hand, on the ground · strong
Ranger · by radio
A walker saw a torch near the gill at about nine.
Second-hand, agrees with the team · backs it up
Member of the public · report desk
Someone online says he is at the far lake.
Nobody vouches for it, nothing agrees · weak
2 of 5 drones go to the gill
3 drones stay on the general search
The far lake waits for the coordinatorCoordinator: not on that evidence. It stays on hold.
Possible find by drone 2Ground team 2 is nearest and is shown the place.
one search · an illustration with made-up reports
Reports arrive
From different people, by different routes, in their own words.
Each is weighed
By who sent it, how they know, and whether anything else agrees.
The ground is marked
The place the strong reports point at becomes likelier on the shared map.
Drones are shared out
Enough go to the new place. The rest keep searching everywhere else.
People decide
A weak report waits for the coordinator. A find goes to the nearest team.
This is an illustration of what Meshfare is being built to do, with made-up reports and a made-up search drawn over a satellite picture of Easedale in the Lake District. It is not a recording of the app. Today the app does these steps in simulation only: no aircraft has flown under it.
People and drones, each doing what they are good at
People know things and carry the responsibility. Drones cover ground quickly and do not tire. Meshfare sits between them and keeps both working from the same picture.
PeopleSay what they knowTeams, rangers, callers and the coordinator, in plain words or by circling the map.
MeshfareWeighs it, proposes a planReads every report beside the state of the search and suggests what the swarm should do.
Code, then a personCheck and approveCode checks every instruction. Anything out of proportion waits for the coordinator.
DronesSearch the groundEach goes where a look is worth most, and reports what it has covered and seen.
EveryoneSees the same mapThe control room sees all of it. Teams on the hill see what has been searched.
↺ and round again: what people see on the map changes what they report, and what they report changes where the drones look.
How it works
A model proposes. Code decides.
A language model is good at reading what people write and poor at being trusted with aircraft. So it is given one job: turn a report into a short plan from a fixed menu of nine instructions. Everything after that is ordinary, tested code.
A report arrives, with its source
From a control room, a ranger, a ground team's phone, or typed by the coordinator. Who it is from is fixed by the link it came through, never by what the sender claims.
A plan is proposed
A model, or keyword rules when no model is available, reads the report beside a brief of the search and proposes instructions from the menu. It sees nothing the search itself does not know.
Every instruction is checked
Shape, place and possibility are validated in code. Limits such as how many drones must stay on the general search are enforced there, not left to a prompt.
Out of proportion? It waits for a person
A plan that would move most of the swarm on weak evidence is held, with the reasons, until someone approves or dismisses it.
control room · demonstration
What it does
One picture of the search, for everyone who needs it.
An operations console on the real ground
The search on a satellite map or the 3D terrain of the place: a shared probability map that every look updates, the drones with their state and battery, the reports and the plans they led to, and a log of everything.
The fleet on one page
Every aircraft with its battery, link and position. Recall or release one, several or all.
Reports in words, or on the map
Type what a coordinator would say, or circle an area and write about it. A vague place, "by the lake" on a map with three, is asked about before anything is planned.
A search anywhere on Earth
Pick a point and the app builds the map to search, from open data: its heights, land cover and place names.
Fits what you already run
Other systems read the search as GeoJSON and send reports as JSON with a latitude and longitude.
Field teams
The people on the hill see the same search.
A team on foot opens a link on the phone they already carry. No account, no app store.
Searched ground, never "cleared". How thoroughly each piece of ground has been looked at, in three steps, because a search can miss.
One large button to report. With a circled area or the phone's position. It arrives in the control room under the team's name.
Works with a poor connection. A report written without signal is kept and sent when there is one.
Shown only what a team needs. Never the probability map, the plans or anyone else's reports.
Ground team 1● live
looked at lightlysearched wellsearched thoroughly
Report something
Built to be trusted
Limits live in code, where they can be tested.
01
A fixed menu of nine
A model can only produce instructions from the menu. It cannot write code, name a waypoint outside the map, or reach an aircraft. With no model at all, the search still runs.
02
A record that shows tampering
Every command, report and find is written as it happens, each entry carrying the hash of the one before, with who was signed in when it was done.
started→report→plan→accepted→found
03
Your network, encrypted
Devices connect over HTTPS with a certificate made on the coordinator's own computer. No outside service sits in the middle, and the control room stays on that computer.
Evidence so far
Measured, with the caveats attached.
Everything below was measured in simulation, on 60 real lost-person probability maps or on small test sets written by the author.
20of 20
Everyday wording understood
Twenty reports in loose, everyday words: Claude Opus 5.5 read all 20 correctly. The keyword rules, with no model, read 8.
35% found, against 26%
The missing person is found more often when a model plans
Each simulated search was sent one true and one false report of where the person was. In 65 searches of 30 minutes, the person was found in 35% when Claude Opus 5.5 decided what to do with the reports, in 26% when fixed rules did, and in 8% when the reports were ignored. That puts the model 9 percentage points ahead of the rules; with only 65 searches, the 95% interval for that lead runs from 1 to 17 percentage points.
~2times
Camera strip against hovering
Drones that search with a camera strip while moving find roughly twice the people of drones that stop and hover over each cell.
What this is not. No aircraft has flown under Meshfare. The test sets are small and the camera model's values are placeholders. It is an early demonstration and a research tool, not something to run a real search with today.
The founders
Who is building Meshfare.
Sam Struthers
Co-founder, Meshfare · MSc Data Science and Artificial Intelligence, TU/e
I founded Meshfare to solve one problem: in a search, information arrives from many people at once, and the aircraft should act on it within minutes, inside limits that can be tested.
I designed and built the software shown on this page: the planner, the simulator, the control room and the field teams' page. I am now showing it to search-and-rescue teams, researchers and drone operators, to shape what it becomes with the people who would use it.
Co-founder, Meshfare · MSc Data Science and Artificial Intelligence, TU/e
I have joined Sam as co-founder. From here on we will develop the software together and share its main projects, starting from what he has built so far.
Meshfare is an early demonstration, not a finished product. It runs searches in simulation, and we are showing it to search-and-rescue teams, researchers and drone operators who want to see where it is going and tell us what it should do.