Benchmark, pilot, instrument
A short, complimentary benchmark on your own frames, with the acceptance threshold agreed before any data is taken. It is the first of three steps: the benchmark; then the application pilot, engineering of your part on our demo configuration ($4,800 = $2,400 on signing + $2,400 at the signed readout, both credited in full to the instrument); then the instrument itself, the bench surface profiler, $34,900, with a second-wavelength option at +$2,400, 50% on order and the balance on delivery, built to your order (sample beam width, emitter colour, focal point). The $10,000 software / $25,000 hardware design partnerships remain for labs that need custom work.
Why a benchmark first
Every measurement with us is a comparison against a reference you already trust, at a threshold fixed in writing before the data is seen. The benchmark produces that record for your part: your frames, your reference measurement, and our reconstruction, all three committed. It is the shortest route to a pilot that is scoped on evidence instead of a demonstration.
What you provide
- Roughly 20 frames from your own rig. Not a curated best-case set — a representative one. No rig yet? Tell us; for a lab heading to the Hardware partnership we agree at scoping a sample or target you send us instead.
- One independent reference measurement of the same target, from an instrument or method you already trust. This is the part that makes it a validation rather than a demonstration. Without it there is nothing to compare a reconstruction against.
- Your optical parameters — illumination wavelength and the pixel pitch at the plane your sensor actually sampled.
If you cannot produce an independent reference measurement inside the validation window, say so early. We would rather postpone than run a validation that cannot produce a comparison.
What we do
- Scoping call. We agree the target, the frame count (at most about 20), one rig configuration, and the reference measurement. We also agree the acceptance threshold — what result would count as a pass — and we fix it here, in writing, before any of your data has been seen. A threshold chosen after seeing the data is not a threshold.
- Reconstruction. Your frames through the same API anyone else uses. No special build, no hand-tuning that you could not reproduce yourself.
- Comparison and report. Our reconstruction set against your reference measurement, with the outcome stated against the threshold agreed in step 1 — including when the outcome is a miss.
What you get
- Every reconstruction, with its provenance record and content-addressed report ID.
- A comparison against your reference measurement, reported against the pre-agreed threshold.
- A result whose input, settings and software you can check yourself, and publish if you want to.
If the validation misses its threshold, the report says so. That is the point of setting the threshold first, and we would rather hand you a committed negative result than a flattering one you cannot reproduce.
Timeline
10 business days from the day your frames, parameters and reference measurement are complete. Long poles are almost always on the input side, not ours.
What it costs, and its bounds
Nothing. In exchange it is bounded: about 20 frames, one rig configuration, one independent reference measurement, 10 business days. It includes no hardware, no custom code and no NDA-scoped work; those belong to the paid step.
We also agree, in writing and before any data, what happens if the threshold is met: the next step is Phase 1 paid pilot engineering ($4,800 in two $2,400 milestones, P-card or single-signer PO, credited in full to the instrument), or, for a ready buyer, a deposit on the bench surface profiler ($34,900) or a Software ($10,000, ~90 days) / Hardware ($25,000, ~4 months) design partnership, with a decision date and a named signer. If the threshold is missed, the report says so and nothing else is owed.
Data handling
Your frames are used for the validation and for nothing else. If you require an NDA, it is signed before any frame is transferred — not at contract stage, not afterwards.
Raw uploads are not retained unless you opt in, and you can delete your results at any time (see data retention). Validation data handling is agreed explicitly at scoping, in writing, including how long anything is kept and what happens to it at the end. Retention is policy-driven and class-keyed, and unclassified records route to a named owner rather than defaulting silently.
Getting started
Have ready: what you are measuring, what your reference measurement is, and whether you can produce both inside the window. That is the whole qualifying conversation.
Pilot intake
You can send those facts without an account. POST /public/pilot-intake takes
a small JSON body: tier set to 1, plus specimen, field_of_view,
reference_method, cycle_time and email, each a short text answer. Tier 2
also carries your run settings: device, wavelength_nm, pixel_pitch_um,
z_mm and geometry (reflection or transmission). An accepted request
returns {"ok": true, "id": "PI-..."}; quote that ID when we reply. A rejected
one returns {"ok": false, "errors": [...]} with one line per problem, saying
what to fix. Requests are limited to 10 per hour from one address.