FILE
FK–2026–06
ENGINE
milestones + closed loop
CLASS.
TECHNICAL · PUBLIC
FLOW
5 touches
SCOPE
4 sectors
5 touches
from purpose to kit: the seven-step wizard was replaced by a five-touch one
4
exclusion axes: phylogenetic, co-marketed, accompanying flora and horizontal transfer
8
independent milestones, each with its own isolated golden tests
0
thresholds written into the code: they all come from the requirement profile and the configuration
+
+
+
+
What makes it unique
The negative set is not a list of relatives: it is multi-axis. What is phylogenetically close, what is marketed alongside the product, the accompanying flora of that matrix, and what travels by horizontal transfer. An assay that only excludes the clade falls over on what shares a shelf, not on what shares a tree.
On top of that, a closed-loop controller: it checks feasibility before starting, diagnoses why it is not working, escalates modality — from one amplicon to a probe over a point variant, and from there to several loci — and only declares the level the covered axes authorise. No competitor in the field — SpeciesPrimer, RUCS, varVAMP, Olivar, MFEprimer — closes the cycle from intent to purchase order with that accounting of what remains uncovered.
01
The state of the art
They all end at the primer. You assemble the kit yourself.
Turning a validated pair into something usable means calculating mixes, writing the protocol, making the operator card, contacting suppliers and documenting for the standard. Weeks of manual work, and no tool in the field goes there.
02
The four axes of the negative set
What your assay must not react to.
The requirement profile is derived from the sector and the use case: what to demand for food safety is not what to demand for plant health or human clinical use. And that demand is accounted for at the end.
The honesty guard
The controller never claims species specificity nor the top tier if any axis required by the profile was left uncovered. Instead of quietly lowering the bar, the verdict is issued with the real scope and the protocol declares which axis could not be covered, and why.
Modality escalation
If a single amplicon does not discriminate, the system neither gives up nor lies: it moves to a probe over a point variant and, failing that, to several loci. Each step makes the assay more expensive, which is why it is proposed with an explanation of what is gained.
03
How it works
Eight milestones that can be audited separately.
Every milestone is an independent entity with its input, its tool and its output, and with its own tests. Not a monolith where failure dissolves.
01
Intent into a requirement profile
What must be detected, in which matrix and for which sector becomes an explicit profile of demands, derived from the curated corpus. That profile is the contract the result will have to satisfy.
02
Organism into markers
Marker selection favouring single-copy genes over the ribosomal one, which discriminates worse at species level.
03
The negative set, in four axes
Building the exclusion set from public databases and curated knowledge, with a guard that prevents treating an unpopulated axis as covered.
04
Windows by base-by-base suitability
Majority consensus, entropy, axis-weighted discrimination and ambiguity codes to choose where design is possible. The template that reaches the designer is always the clean one, never the ambiguous one.
05
Design and thermodynamic validation
Primers bounded to the window, with a fallback to free design if there is no solution, and combinatorial validation: free energy weighted at the 3′ end, forward-by-reverse amplicon and off-target per axis.
06
Closed loop and honest verdict
The controller checks feasibility before spending compute, diagnoses failures, escalates modality if needed and stops when the profile is satisfied — or when it is time to say it cannot be done.
07
Productisation
A protocol conforming to the assay reporting guideline with the scope and any uncovered axes declared, quick cards, a technical sheet and purchase orders for oligonucleotides and reagents.
04
The contract
What goes in, what comes out, what it chains to.
In
The intent in laboratory language — what to detect, in which matrix, for which sector — and, where they already exist, the validated primers or the targets recommended by the upstream modules.
Out
A complete kit: a bilingual protocol with a calculated reagent table and cycling, a quick reference card for the operator, the assay technical sheet, the in silico validation dossier and a purchase order with specifications.
Chains to
It reuses SnapPrime's design and asks DigiCycler for a second opinion with an independent engine. When the assay is actually run, Hermes will close the circle by interpreting what the instrument measured.
05
Technical sheet
Sectors
Food safety, veterinary, plant health and human clinical, each with its own requirement profile and its intent-assisted mode.
Input
Intent expressed in laboratory language; or already validated primers and targets recommended by the upstream modules.
Output
A bilingual protocol with calculated reagents and cycling, an operator card, a technical sheet, the in silico validation dossier and a purchase order — all exportable as PDF.
Architecture
Eight independent milestones with their own data contract and isolated golden tests, governed by a closed-loop controller.
Thresholds
None written into the code: they come from the requirement profile and the module configuration, so adjusting a demand does not force a redeploy.
Multi-tenant
Isolation by organisation and facility is mandatory throughout the flow, not a deployment option.
Limit · signature
FastKit prepares the technical dossier. The declaration of conformity is signed by the manufacturer: the system documents, it does not certify.
Limit · in silico
Pre-kit validation is biophysical prediction. It replaces rounds of trial and error, not the experimental verification the standard requires before putting a method into production.
Limit · exclusion
The quality of the negative set depends on what has been deposited and curated for that matrix and that sector. When an axis cannot be populated, the kit ships with that axis declared uncovered instead of silenced.
Design benchmarked against the field's state of the art: SpeciesPrimer, RUCS, varVAMP, Olivar and MFEprimer. Reference standards: ISO 6579 for Salmonella, ISO 11290 for Listeria, ISO 10272 for Campylobacter, EPPO PM 7 protocols in plant health and CE-IVD marking in human diagnostics. Protocol issued in line with the MIQE guideline, with an explicit declaration of scope.
From purpose to purchase order, without changing system.
BIOTECNO.org · Vitoria-Gasteiz
EU + Codex