// THE AIM · PROVEN ONE FIELD TRIAL AT A TIME
Heimdall is planned farm hardware plus a managed software and service layer. The goal is accurate, farm-wide fox movement tracking: sensing points handing observations from one to the next so you can see a useful path across the property—not just one photo.
We are not there yet. The first fox-focused MVP is built and in bench testing, with a deployment near Nairne scheduled for January 2027 to test identification, node-to-node handoff, direction, timing, power and real paddock coverage. A small early field-test cohort will help us find the failures and measure what is accurate enough to be useful. There are no field results yet.
Where the system is intended to go
Show where a fox appeared, the direction it travelled and the sensing points it passed. If later welfare, neighbour-noise and field-test evidence supports it, brief directional light or sound cues may interrupt, slow or redirect a confirmed fox away from stock. Heimdall buys time and evidence; you decide what to do next.
The Pilot EOI is free, de-identified and non-binding research. It is not a customer, order, quote, paid pilot or promise that your property will be selected.
// PILOT EOI
We are looking for a small early field-test cohort: farmers who understand the system is unproven and are willing to help us test it, find what breaks and improve it. The aim is accurate, useful fox movement tracking across a working property—not a polished demo.
Nothing here promises selection or performance. A submitted EOI is evidence of interest only, never a customer, sale, paid pilot or signed LOI. Selected farms would agree a separate trial scope, access, privacy and safety plan before anything happens.
Know an SA farmer who may want to give a sober view on the proposed pilot?
https://heimdall.farm/?ref=sa-farmer#interestResearch first · callback separate
Checking pilot intake
Loading the current production forms…
// MARAUDERS, FURRY OR OTHERWISE
The goal is useful movement evidence across the farm, delivered early enough that you can inspect, verify and respond your way.
Loss can build before a producer has useful information about where a fox entered, when it moved or which corridor it used. A published Australian review reports 1–30% lamb loss on affected properties, depending on conditions.
Unusual person, vehicle or perimeter activity can leave the same evidence gap. Heimdall's intended output stays neutral—where and when something moved, never who a person is or what they intended.
Heimdall is aiming to join observations from sensing points across a property into a useful movement trail: where a fox appeared, where it headed and how its path progressed—while the information is still useful to the farmer.
That is the end goal, not a field result. The near-Nairne trial and any selected early cohort must establish identification accuracy, fox-versus-dog confusion, path handoff, coverage and failure modes before Heimdall claims it can track fox movement accurately.
// ACROSS THE FARM · A HUMAN DECIDES
The aim is for sensing points across the farm to work together. Technology builds the movement picture; it never decides intent or acts autonomously.
A low-power motion sensor notices movement and wakes the rest of the node.
Thermal sensing checks for a warm body before cameras use more power.
RGB/IR imagery captures evidence for human review: fox, dog or neutral person/vehicle presence.
The next sensing point picks up movement and the network hands the observations toward the farmhouse.
You review the time, place, direction, confidence and raw thumbnail—and decide what happens next.
Identification distance, fox-versus-dog confusion, node-to-node path handoff, fence-post-height radio performance and field power autonomy. Those measurements set the honest accuracy and coverage limits.
Brief, directional light or sound cues may be tested to interrupt, slow or redirect a confirmed fox away from stock. They must never surround, pin, chase to exhaustion or prolong distress. They ship only if welfare, neighbour-noise and field evidence support them.
// GENERATED CONCEPT SIMULATION
This scroll-driven map demonstrates a possible interface for generated sensor events. It is not telemetry, a field trial, measured coverage or evidence that the proposed hardware works. Click markers to inspect the concept; zoom before dragging the map.
Everything shown here—terrain, nodes, contacts, confidence, movement and sensor geometry—is generated for interface discussion. Thermal/IR signature imagery remains disabled pending field-captured reference frames. The proposed experiment must measure identification range, fox/dog confusion, radio handoff, power use and practical placement.
// WHAT WE BELIEVE · BEFORE WHAT WE BUILD
The principles come first. We care about animal welfare, practical right-to-repair and a rolling delayed-open model that protects active development without turning yesterday's technology into permanent lock-in.
Earlier, reviewable information should help people prevent losses and respond proportionately, reducing avoidable harm to livestock, wildlife and working animals. Observation supports responsible decisions; it does not pretend to replace lawful control.
We do not want another sealed, dealer-dependent agricultural box that becomes useless when a vendor withdraws support or permission. Owned hardware should remain understandable, repairable and useful beyond a product cycle.
The current generation may retain a three-to-four-year commercial lead. Our intention is then to publish a defined package from that generation so the advantage keeps moving forward instead of becoming permanent lock-in.
Once a generation opens, people should be encouraged to repair it, add sensors, alter software and share what works. Useful community improvements can then inform newer Heimdall generations and shorten the next development cycle.
Heimdall is a human-reviewed observation system intended to help people see earlier without surrendering control to another locked platform. The technology has to earn its place through measured field evidence.
This is the intended model, not a present licence promise. The exact release trigger, materials opened, licences, contribution terms and security boundaries must be defined before any generation ships.
Have a bounded property question? Ask the expert →// PLANNED COMMERCIAL MODEL · NOT AN OFFER
Heimdall is being designed as equipment a farmer can ultimately own, with managed software and service kept separate. Exact prices, inclusions and customer terms remain unvalidated and are not public offers.
The intended structure spreads the hardware term while preserving a clear ownership outcome. At the end of that term, the customer owns the equipment and the then-current software snapshot.
A separate managed software and service layer is planned for ongoing managed functions. It must not become a remote switch that decides whether owner-controlled equipment can keep doing its stable local job.
Owned equipment and its stable local capability continue after managed service lapses. Safety-critical patches remain free regardless of managed-service status.
Near-Nairne measurements, current supplier quotes, reviewed unit economics and legally reviewed customer terms must come before public hardware or managed-service pricing. The technical room records the architecture, evidence plan and validation gates.
// THE GOAL · THE TRIALS · THE LIMITS
What Heimdall is aiming to achieve, what an early cohort will help prove and what remains unbuilt today.
Accurate, farm-wide fox movement tracking: where a fox appeared, where it is heading and how it moved between sensing points. The aim is to give you useful evidence early enough to inspect and respond your way. The trials exist to establish how accurate and reliable that can be.
A missing lamb alone does not prove fox predation. Record when and where losses occur, photograph carcasses or tracks before disturbing them, note fence crossings and repeated movement, and consider other plausible causes. Heimdall aims to add a property-wide movement trail to that evidence, not turn one clue into proof.
The first node cluster is built and in bench testing, with a near-Nairne deployment scheduled for January 2027. There are no field results yet. The first trial and a small selected field-test cohort will help measure identification accuracy, fox-versus-dog confusion, node handoff, coverage, power and the failures that need fixing.
That is part of the later vision, not a current capability. If identification, welfare, neighbour-noise and field testing support it, brief directional light or sound cues may interrupt, slow or redirect a confirmed fox away from stock. They must never surround, pin, exhaust or prolong distress.
The intended output is a practical movement picture: time, sensing point or point-pair, direction, confidence and a raw thumbnail. As more nodes observe the same path, Heimdall aims to show how movement progresses across the farm. A person reviews it and decides what happens next.
No. Any person or vehicle output is neutral presence, location, movement and time. There is no face recognition, named biometric identity or claim about intent. For HNWI deployments, unmatched or unconfirmed human footage must purge after seven days.
Protected or native species are excluded first. Fox is the only actionable target class planned for v1; any other canid is reported neutrally as dog. Field trials must establish the fox-versus-dog confusion matrix before Heimdall claims accurate fox identification or movement tracking.