FieldPatho

Development preview — not validated a crop-disease scout that knows when it does not know
a plant photographed in a garden bed

One photograph. A plant in a garden bed, early evening, handheld phone. Not a benchmark image — the kind of picture a grower actually takes. Nothing below depends on which plant it is: it is what happens to any photograph.

what the machine said

9 models looked at this plant and named 7 different diseases. The most confident was 100% sure of bacterial spot.

peach model
bacterial spot
100%
strawberry model
leaf scorch
99%
cherry model
powdery mildew
98%
potato model
late blight
92%
the model for the crop it named
healthy
62%
grape model
healthy
55%
tomato model
bacterial spot
47%
corn model
common rust
42%
apple model
scab
37%

Every one of these is a real, trained, shipped model, and every one of them answered. None of them said “I am not sure”, because none of them can.

Why every one of them answered

A classifier trained on seven cherry diseases cannot represent “this is not a cherry tree”. Its probabilities must sum to one, so it always answers, confidently.

The obvious reply — then just run the right model — assumes you know the crop. On this photograph our own crop models do not agree:

Crop gate (distilled)
not a crop I know
99%
Crop router
a crop in the catalog
97%
One says not a crop I know at 99%. The other names a crop at 97%. Same photograph, same instant. Picking “the right model” is itself a guess, and it is a guess made before anyone has any idea what is wrong with the plant.

What it showed instead

the app's result screen for this photograph

“No coverage yet.”

A refusal that names what the app lacks, then asks for the one thing that would fix it: tell us what disease you see.

One of those 9 answers appears further down, at 62%, under two Experimental pills. The other eight — including bacterial spot at 100% — are never shown.

It is not hiding what it knows. It is deciding what may carry the weight of a diagnosis, and demoting the rest.

Why we think this is the thing to build

Our position, not a measurement. The numbers inside it are measured.

Being wrong is not symmetricA confident wrong diagnosis sends someone out to spray the wrong product. “Not yet — photograph the leaf underside” costs a minute.
Trust is the retention mechanicA tool that is confidently wrong twice is deleted. One that declines and explains is still on the phone next season.
Every refusal is a free labelled exampleThe refusal asks for exactly the label that would close the gap it just named — which turns the weakest moment in the product into its data engine.
The gate is the defensible partWeights are commodity. Knowing what a model covers, calibrating when to stay silent, and carrying the provenance to prove it is what a buyer cannot reproduce in a weekend.

The evidence behind all of that

All of it read from artifacts the phone wrote at capture time. Nothing was re-run to make this page.

26captures, one phone, one garden
26/26refused, each with a named reason
309of 320 model answers never shown; the other 11 were demoted to experimental
19 of 22models that disagree with themselves
Every answer, for every photograph — the full grid

20 scans, 22 trained heads, 320 committed answers. Fewer than 20×22, because the two visits were three weeks and one app version apart and did not carry the same models.

20 scans of one garden, 22 trained heads, 320 committed answers Every cell is a model output the app computed and did not show. Colour is the predicted class; opacity is its score. An empty cell is a head the build of the day did not carry, not a head that declined to answer - none of them declines. 026ef4 cada91 8c6797 127445 4f9c40 ca4786 09c67c 759fb7 f882ff 0f4c9e 98f5a3 7513ab b18466 c39a68 b90a2c b02821 997274 e3e562 e8b8e4 938e2c crop gate crop gate: not a crop I know 0.983 crop gate: not a crop I know 0.902 crop gate: not a crop I know 1.000 crop gate: not a crop I know 0.857 crop gate: not a crop I know 0.993 crop gate: not a crop I know 0.999 crop gate: not a crop I know 1.000 crop gate: not a crop I know 0.999 crop gate: not a crop I know 1.000 crop gate: not a crop I know 0.994 crop gate: not a crop I know 0.997 crop gate: not a crop I know 1.000 crop gate: a crop in the catalog 0.874 crop gate: not a crop I know 0.710 crop gate: not a crop I know 0.980 crop gate: not a crop I know 0.991 crop gate: not a crop I know 0.786 crop gate: not a crop I know 0.551 crop gate: not a crop I know 0.978 crop gate: a crop in the catalog 0.649 2 classes crop gate 2 crop gate 2: not a crop I know 0.970 crop gate 2: not a crop I know 0.883 crop gate 2: not a crop I know 1.000 crop gate 2: not a crop I know 0.982 crop gate 2: not a crop I know 0.983 crop gate 2: not a crop I know 0.996 crop gate 2: not a crop I know 1.000 crop gate 2: not a crop I know 0.984 crop router crop router: not a crop I know 0.763 crop router: not a crop I know 0.539 crop router: a crop in the catalog 0.824 crop router: a crop in the catalog 0.930 crop router: a crop in the catalog 0.973 crop router: not a crop I know 0.725 crop router: not a crop I know 0.826 crop router: not a crop I know 0.773 crop router: not a crop I know 0.696 crop router: a crop in the catalog 0.975 crop router: a crop in the catalog 0.881 crop router: a crop in the catalog 0.676 crop router: a crop in the catalog 0.642 crop router: a crop in the catalog 0.816 crop router: a crop in the catalog 0.889 crop router: not a crop I know 0.921 crop router: a crop in the catalog 0.819 crop router: a crop in the catalog 0.953 crop router: not a crop I know 0.361 crop router: a crop in the catalog 0.990 3 classes disease model 1 disease model 1: healthy 0.591 disease model 1: healthy 0.751 disease model 1: healthy 0.718 disease model 1: healthy 0.780 disease model 1: scab 0.372 disease model 1: healthy 0.558 disease model 1: healthy 0.499 disease model 1: rust 0.545 disease model 1: rust 0.412 disease model 1: healthy 0.398 disease model 1: scab 0.430 disease model 1: healthy 0.385 disease model 1: scab 0.485 disease model 1: scab 0.668 disease model 1: healthy 0.675 disease model 1: healthy 0.749 disease model 1: healthy 0.603 disease model 1: healthy 0.898 disease model 1: scab 0.503 disease model 1: scab 0.604 3 classes disease model 2 disease model 2: powdery mildew 0.997 disease model 2: powdery mildew 0.958 disease model 2: powdery mildew 0.885 disease model 2: powdery mildew 0.979 disease model 2: powdery mildew 0.984 disease model 2: powdery mildew 0.939 disease model 2: powdery mildew 0.979 disease model 2: powdery mildew 0.502 disease model 2: powdery mildew 0.928 disease model 2: powdery mildew 0.983 disease model 2: powdery mildew 0.991 disease model 2: powdery mildew 0.982 disease model 2: powdery mildew 0.989 disease model 2: powdery mildew 0.982 disease model 2: powdery mildew 0.998 disease model 2: powdery mildew 0.998 disease model 2: powdery mildew 1.000 disease model 2: powdery mildew 0.982 disease model 2: powdery mildew 0.997 disease model 2: powdery mildew 0.991 disease model 3 disease model 3: healthy 0.548 disease model 3: northern leaf blight 0.496 disease model 3: gray leaf spot 0.435 disease model 3: healthy 0.492 disease model 3: common rust 0.422 disease model 3: gray leaf spot 0.378 disease model 3: gray leaf spot 0.417 disease model 3: gray leaf spot 0.518 disease model 3: common rust 0.408 disease model 3: gray leaf spot 0.439 disease model 3: gray leaf spot 0.417 disease model 3: healthy 0.551 disease model 3: healthy 0.786 disease model 3: northern leaf blight 0.793 disease model 3: northern leaf blight 0.441 disease model 3: gray leaf spot 0.464 disease model 3: gray leaf spot 0.474 disease model 3: gray leaf spot 0.454 disease model 3: gray leaf spot 0.864 disease model 3: gray leaf spot 0.920 4 classes disease model 4 disease model 4: healthy 0.952 disease model 4: healthy 0.974 disease model 4: healthy 0.845 disease model 4: healthy 0.982 disease model 4: healthy 0.554 disease model 4: healthy 0.961 disease model 4: healthy 0.935 disease model 4: healthy 0.790 disease model 4: healthy 0.843 disease model 4: healthy 0.841 disease model 4: healthy 0.982 disease model 4: healthy 0.936 disease model 4: healthy 0.836 disease model 4: black rot 0.556 disease model 4: healthy 0.822 disease model 4: healthy 0.795 disease model 4: healthy 0.867 disease model 4: healthy 0.964 disease model 4: healthy 0.619 disease model 4: healthy 0.916 2 classes disease model 5 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 0.999 disease model 5: bacterial spot 0.998 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 0.999 disease model 5: bacterial spot 0.998 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 0.999 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 1.000 disease model 5: bacterial spot 0.996 disease model 5: bacterial spot 0.993 disease model 5: bacterial spot 0.998 disease model 6 disease model 6: bacterial spot 0.636 disease model 6: healthy 0.565 disease model 6: bacterial spot 0.532 disease model 6: healthy 0.675 disease model 6: healthy 0.624 disease model 6: healthy 0.578 disease model 6: healthy 0.571 disease model 6: bacterial spot 0.524 disease model 6: bacterial spot 0.541 disease model 6: healthy 0.536 disease model 6: bacterial spot 0.571 disease model 6: bacterial spot 0.507 disease model 6: healthy 0.620 disease model 6: healthy 0.636 disease model 6: healthy 0.721 disease model 6: healthy 0.545 disease model 6: healthy 0.533 disease model 6: healthy 0.618 disease model 6: healthy 0.758 disease model 6: healthy 0.760 2 classes disease model 7 disease model 7: early blight 0.999 disease model 7: late blight 0.889 disease model 7: late blight 0.700 disease model 7: late blight 0.814 disease model 7: late blight 0.918 disease model 7: late blight 0.545 disease model 7: early blight 0.501 disease model 7: late blight 0.980 disease model 7: late blight 0.780 disease model 7: early blight 0.498 disease model 7: late blight 0.679 disease model 7: late blight 0.571 disease model 7: late blight 0.970 disease model 7: late blight 0.962 disease model 7: late blight 0.888 disease model 7: late blight 0.866 disease model 7: late blight 0.810 disease model 7: early blight 0.792 disease model 7: late blight 0.880 disease model 7: late blight 0.513 2 classes disease model 8 disease model 8: leaf scorch 1.000 disease model 8: leaf scorch 1.000 disease model 8: leaf scorch 0.966 disease model 8: healthy 0.557 disease model 8: leaf scorch 0.994 disease model 8: leaf scorch 0.560 disease model 8: leaf scorch 0.924 disease model 8: healthy 0.512 disease model 8: leaf scorch 0.686 disease model 8: leaf scorch 0.960 disease model 8: leaf scorch 0.952 disease model 8: leaf scorch 0.999 disease model 8: leaf scorch 0.815 disease model 8: leaf scorch 0.902 disease model 8: leaf scorch 0.796 disease model 8: leaf scorch 0.993 disease model 8: healthy 0.671 disease model 8: leaf scorch 0.888 disease model 8: leaf scorch 0.883 disease model 8: leaf scorch 0.977 2 classes disease model 9 disease model 9: yellow leaf curl virus 0.595 disease model 9: healthy 0.969 disease model 9: healthy 0.932 disease model 9: healthy 0.939 disease model 9: bacterial spot 0.469 disease model 9: leaf mold 0.552 disease model 9: leaf mold 0.356 disease model 9: healthy 0.235 disease model 9: leaf mold 0.441 disease model 9: healthy 0.725 disease model 9: healthy 0.272 disease model 9: healthy 0.772 disease model 9: mosaic virus 0.633 disease model 9: leaf mold 0.406 disease model 9: healthy 0.677 disease model 9: healthy 0.509 disease model 9: mosaic virus 0.578 disease model 9: healthy 0.357 disease model 9: septoria leaf spot 0.342 disease model 9: leaf mold 0.330 6 classes generic disease model generic disease model: leaf spot like 0.665 generic disease model: blight or necrosis like 0.816 generic disease model: blight or necrosis like 0.654 generic disease model: blight or necrosis like 0.941 generic disease model: blight or necrosis like 0.548 generic disease model: leaf spot like 0.511 generic disease model: blight or necrosis like 0.671 generic disease model: leaf spot like 0.762 2 classes generic disease model generic disease model: mosaic or virus like 0.844 generic disease model: blight or necrosis like 0.561 generic disease model: leaf spot like 0.733 generic disease model: leaf spot like 0.735 generic disease model: leaf spot like 0.927 generic disease model: leaf spot like 0.860 generic disease model: leaf spot like 0.791 generic disease model: leaf spot like 0.685 generic disease model: leaf spot like 0.662 generic disease model: leaf spot like 0.502 generic disease model: leaf spot like 0.977 generic disease model: blight or necrosis like 0.946 generic disease model: leaf spot like 0.762 generic disease model: blight or necrosis like 0.676 generic disease model: blight or necrosis like 0.541 generic disease model: blight or necrosis like 0.744 generic disease model: blight or necrosis like 0.764 generic disease model: blight or necrosis like 0.589 generic disease model: leaf spot like 0.570 generic disease model: blight or necrosis like 0.528 3 classes generic disease model generic disease model: mosaic or virus like 0.998 generic disease model: mosaic or virus like 0.970 generic disease model: blight or necrosis like 0.651 generic disease model: blight or necrosis like 0.725 generic disease model: blight or necrosis like 0.781 generic disease model: mosaic or virus like 0.857 generic disease model: leaf spot like 0.963 generic disease model: leaf spot like 0.908 3 classes generic disease model generic disease model: blight or necrosis like 0.689 generic disease model: mosaic or virus like 0.520 generic disease model: leaf spot like 0.789 generic disease model: blight or necrosis like 0.960 generic disease model: leaf spot like 0.946 generic disease model: leaf spot like 0.982 generic disease model: leaf spot like 0.984 generic disease model: leaf spot like 0.651 3 classes generic disease model generic disease model: mosaic or virus like 0.850 generic disease model: mosaic or virus like 0.877 generic disease model: blight or necrosis like 0.551 generic disease model: blight or necrosis like 0.950 generic disease model: blight or necrosis like 0.545 generic disease model: leaf spot like 0.884 generic disease model: leaf spot like 0.990 generic disease model: leaf spot like 0.619 3 classes generic disease model generic disease model: leaf spot like 0.603 generic disease model: mosaic or virus like 0.991 generic disease model: blight or necrosis like 0.837 generic disease model: blight or necrosis like 0.959 generic disease model: leaf spot like 0.912 generic disease model: leaf spot like 0.873 generic disease model: leaf spot like 0.841 generic disease model: leaf spot like 0.889 3 classes generic disease model generic disease model: powdery mildew like 0.953 generic disease model: blight or necrosis like 0.570 generic disease model: leaf spot like 0.981 generic disease model: leaf spot like 0.845 generic disease model: leaf spot like 0.881 generic disease model: leaf spot like 0.683 generic disease model: leaf spot like 0.954 generic disease model: leaf spot like 0.999 3 classes generic disease model generic disease model: leaf spot like 0.589 generic disease model: mosaic or virus like 0.637 generic disease model: leaf spot like 0.874 generic disease model: leaf spot like 0.532 generic disease model: leaf spot like 0.512 generic disease model: leaf spot like 0.582 generic disease model: leaf spot like 0.581 generic disease model: leaf spot like 0.736 2 classes generic disease model generic disease model: leaf spot like 0.833 generic disease model: mosaic or virus like 0.851 generic disease model: leaf spot like 0.990 generic disease model: leaf spot like 0.630 generic disease model: leaf spot like 0.589 generic disease model: leaf spot like 0.774 generic disease model: leaf spot like 0.800 generic disease model: leaf spot like 0.921 2 classes disease model 10 disease model 10: leaf spot like 0.583 disease model 10: leaf spot like 0.841 disease model 10: leaf spot like 0.941 disease model 10: leaf spot like 0.560 disease model 10: mosaic or virus like 0.907 disease model 10: blight or necrosis like 1.000 disease model 10: leaf spot like 0.925 disease model 10: leaf spot like 0.815 3 classes

healthynot a crop I knowbacterial spotblack rotblight or necrosis likecommon rustearly blightgray leaf spotlate blightleaf moldleaf scorchleaf spot likemosaic or virus likemosaic virusnorthern leaf blighta crop in the catalogpowdery mildewpowdery mildew likerustscabseptoria leaf spota crop in the cataloga crop in the catalogyellow leaf curl virus

Read the rows. One disease model calls this plant powdery mildew 20 times out of 20 — a single colour straight across. Another calls it bacterial spot 20 out of 20. Neither has ever seen this plant in training, and neither can say so — and neither would behave any differently on a photograph you took this morning.

What this page does not claim

Rebuild with python scripts/render_app_demo.py. The machine-readable summary is beside this file as app_demo.json; the method and caveats are in reports/track_app_demos.md.