The whole thing, running
One unbroken session, from an empty map of New Zealand to a costed design and its report. Silent, and sped up where nothing is happening, so it runs to about a minute and a half. Full screen is worth it: the tool's own text is readable there.
A working beta
Neobiome Intelligence is a pre-release tool in active development. What you see here is real and running, not a mockup: every figure on this page is engine output at a real New Zealand coordinate. It runs in a controlled environment while its national datasets and hosting are being finalised, which is why it is shown here through the recording and the worked sites rather than as a live map you can click. A public release is planned.
Noted, and thank you. Knowing how many people reach for it is genuinely useful while the release is being planned.
It is not open to the public yet. If you would like to hear when it is, leave your email here, or write to research@neobio.me.
Four steps, one place at a time
Two of the artefact's three components run together here: Data Fusion, which samples a real location, and the Engine, which turns that sample into numbers. The whole tool is a single idea repeated: sample once, then let that one sample drive both the numbers and the recommendations.
Pick a place
Click a point anywhere in New Zealand. The tool reads that exact location from more than twenty national datasets: sun, wind, a nearby stream, the land, the water table, the hazards, how far it is from the grid.
It reads the site
That single sample becomes a site profile. The same profile feeds two things at once, which is the point: the engine that produces the numbers, and the recommendation layer that suggests which technologies actually fit this ground.
It recommends what fits
Because the site is already sampled, the tool can say micro-hydro is viable because there is a stream at three hundred metres, or grey out wind on a sheltered site. Every suggestion links back to the research page that justifies it.
It calculates
As you add or remove technologies, two things update live: a self-sufficiency score across energy, water and food, and a build-and-run cost. The three are combined so that a weak domain drags the whole score down and a surplus of one cannot make up for a shortage of another, since spare electricity will not feed or water you. The weakest of the three is named alongside the score as the limiting domain. The cost is an indicative ballpark, intended to be within about twenty per cent, not a quote.
The same question, five different places
| Tasman | Northland | Canterbury | Southland | West Coast | |
|---|---|---|---|---|---|
| Self-sufficiency | 1.00 | 1.00 | 0.81 | 1.00 | 0.001 |
| In a bad year | 0.69 | 0.83 | 0.48 | 0.76 | 0.001 |
| Limiting domain | None | None | Water and food, both 0.73 | None | Food 0.00 |
| Dry-summer water test | Fails | Passes | Fails | Fails | Passes |
| Verdict | Falls short in a dry summer | Self-sufficient | Falls short on water and in a dry summer | Falls short in a dry summer | Unbuildable, all conservation land |
| What the tool chose to build | |||||
| Solar PV | 200 kWp | 300 kWp | 300 kWp | 300 kWp | 100 kWp |
| Battery | 500 kWh | 250 kWh | 250 kWh | 250 kWh | None |
| Wind turbine | None | None | None | None | None |
| Wood boiler | None | None | None | 150 kWth | None |
| Grazing land used | 65 ha | 66 ha | None, all arable | 98 ha | None |
| What it would cost | |||||
| Indicative CapEx | $2.68m | $2.85m | $2.36m | $3.02m | $1.40m |
| Indicative OpEx / yr | $221k | $202k | $152k | $266k | $144k |
| Cost of energy | 40.5 c/kWh | 35.9 c/kWh | 33.0 c/kWh | 47.6 c/kWh | 40.6 c/kWh |
| Tasman | Northland | Canterbury | Southland | West Coast | |
|---|---|---|---|---|---|
| Self-sufficiency | 1.00 | 1.00 | 0.81 | 1.00 | 0.001 |
| In a bad year | 0.86 | 0.97 | 0.55 | 0.88 | 0.001 |
| Limiting domain | None | None | Water and food, both 0.73 | None | Food 0.00 |
| Dry-summer water test | Fails | Passes | Fails | Fails | Passes |
| Verdict | Falls short in a dry summer | Self-sufficient | Falls short on water and in a dry summer | Falls short in a dry summer | Unbuildable, all conservation land |
| What the tool chose to build | |||||
| Solar PV | 400 kWp | 300 kWp | 200 kWp | 200 kWp | 300 kWp |
| Battery | 1500 kWh | 1000 kWh | 1500 kWh | 1500 kWh | 1000 kWh |
| Wind turbine | None | None | 50 kW | 50 kW | None |
| Wood boiler | None | 150 kWth | None | 150 kWth | 150 kWth |
| Grazing land used | 65 ha | 66 ha | None, all arable | 98 ha | None |
| What it would cost | |||||
| Indicative CapEx | $3.77m | $3.43m | $3.75m | $4.41m | $2.81m |
| Indicative OpEx / yr | $285k | $268k | $259k | $371k | $186k |
| Cost of energy | 85.4 c/kWh | 82.9 c/kWh | 96.7 c/kWh | 122.5 c/kWh | 87.9 c/kWh |
How to read the scores. Self-sufficiency runs from 0 to 1, and 1.00 means the community covers all of its energy, water and food needs from the site itself in a typical year. "In a bad year" re-tests the same design with each need under its harshest conditions, a hard winter for heat and power, a dry year for water. The limiting domain names the weakest of the three, the one holding the score down, and "None" means no domain is short. The dry-summer water test is a separate thing again: a seasonal shortfall that no annual figure reveals, so it can fail at a site that reads 1.00 all year. Three of these five read 1.00 and only one of them holds through a dry summer. West Coast is why the limiting domain and the verdict are different questions: food reads 0.00 there, but what makes the site impossible is that the parcel is entirely conservation land.
One caveat on Southland. On rainwater alone no off-grid design there clears both the dry-summer test and the battery autonomy floor, so the figure above is re-sized to the next battery that could actually be built. It is a selection rather than a result the search returned.
The four sites with buildable ground get the same core kit: solar, a battery, heat pumps, solar hot water, a rainwater tank and greywater recycling, which is the piece that closes water in a typical year. What changes is the sizing and the land. Tasman needs the biggest battery and the least solar; Canterbury has no grazing at all, so its fat has to come from crops; Southland is the only one that needs a boiler. The West Coast parcel gets almost nothing, because almost nothing may be built there.
Southland is the site that changes the picture. It is the only one where the tool chooses a wood boiler, and without one that community would cover less than half its winter heat locally instead of all of it. At Tasman, Northland and Canterbury a boiler would buy insurance against a bad winter rather than a better ordinary year, for around a quarter of a million dollars, so the tool does not choose one.
Cut the connection and the designs stop looking alike. Every battery grows, and among the four sites with buildable ground one buys only solar and storage, one adds a wood boiler, one adds a wind turbine, and one adds both. Four sites, four different answers, from one instrument running one community.
This is also where wind appears, and it is worth saying why it was missing before. Wind is in the option space at three of the five sites, wherever the wind speed clears a capacity-factor threshold, and it fails that test at Tasman and on the West Coast. At the other three it qualified, was evaluated and was not bought: at Northland alone the tool ranked 3,240 candidate designs, 2,160 of them carrying a turbine, and chose none. The reason is that it ranks designs on a typical year, and solar alone already reaches 1.00 there. Adding 50 kW at Northland lifts a bad year's electricity from 0.57 to 0.93 for about $600,000, while the typical-year score does not move at all, so the ranking cannot see the gain. Off-grid, where the winter gate binds, it buys wind at Canterbury and Southland. The research reports this as a limitation of the instrument rather than of the technology, and names the fix.
Cutting the connection costs between $0.6M and $1.4M more at each buildable site and more than doubles the cost of energy at every one of them. It changes no verdict. Northland is self-sufficient either way, the two that fall short in a dry summer still do, and Canterbury still cannot meet its own water demand.
All five are modelled on rainwater only, meaning no bore. Adding a consented bore closes the dry-summer test at every site it is applied to, and it is cheaper than the greywater system it replaces, which is the opposite of what the water constraint was assumed to cost. Leaving it out is what makes the difference between these sites visible.
There is no single answer to "can a community feed and power itself here". What limits it changes with the ground: in the dry east water and food run out together, in the warm north nothing does, and on the West Coast parcel the limit is not physical at all but legal. Three of these five read a full 1.00 in a typical year and only one of them holds through a dry summer, which is the sort of thing a single headline number hides.
The honest edges
Stated up front, not buried. A tool that hides its limits is harder to trust than one that names them.
It runs on open national datasets, not on a surveyed site. It is a proof of concept for a location, not a site assessment.
National layers have a grain. Below a certain scale the tool cannot see, and it says so rather than guessing.
The tool recommends and quantifies. A person makes the call. That separation is deliberate.
Forty-seven entries, each naming a specific assumption or boundary. The full register is published with the artefact rather than here.