Nishi Family › Compare › Food Science and Engineering Suite
Nishi Compare · full-field SOTA · measured, not asserted
Food Science and Engineering Suite
Nishi vs the full field — every axis measured or researcher-sourced, grouped by category; each strip shows the whole field at a glance.
Nishi integrated food-science suite (39 gated organs across 14 disciplines) vs the specialist tools. Quantitative axes measured/published (Nishi measured by gates; competitors researcher-sourced); grade axes B=Best Y=Yes ~=Part n=No; every axis carries a source note. Honest: Nishi leads on sovereignty + integration + safety-by-construction and trails on measured-data depth, lab/instrument validation, and per-tool breadth.
Scope and cost (quantitative)
Nishi spans 14 disciplines in one composable suite; each incumbent is deep in 1-2 (count of the 14 suite disciplines each tool addresses)
Minitab 1 statsAspen Plus 2 processComBase 1 microbiologyFoodData Central 1 nutrition dataWolfram broad mathCOMSOL 2 physicsCompusense 1 sensoryFoodpairing 1 flavor
each Nishi organ carries a native-lane assertion gate run in the census; incumbents publish no per-feature gate count
Minitab n/aAspen Plus n/aComBase n/aFoodData Central n/aWolfram n/aCOMSOL n/aCompusense n/aFoodpairing n/a
Nishi zero-license sovereign ELF; competitor figures are published list-price ranges
Minitab ~1500Aspen Plus 30000 plusComBase subscriptionFoodData Central 0 public dataWolfram ~2150COMSOL 10000 plusCompusense quoteFoodpairing quote
Nishi researcher fetched 79 authoritative docs live over sovereign TLS, grounding the constants
Minitab vendor docsAspen Plus vendor docsComBase curated DBFoodData Central 400k foodsWolfram curatedCOMSOL vendor docsCompusense vendorFoodpairing proprietary DB
Food safety and preservation
ComBase and USDA PMP are the measured-data authority (Best); Nishi computes D and z-value plus the 12-D cook exact, never-poison gated
ComBase holds 1000s of MEASURED growth curves (Best) [combase]; Nishi gates the hazard envelope structurally, no measured curves
Nishi implements the Leistner combined-hurdle model exact [leistner2000]; ComBase and Aspen partial
Nishi HALTs a breached critical control point by construction; dedicated HACCP suites (outside this column set) own the audit workflow
Process physics and chemistry
Aspen Plus is the process-simulation Best with rigorous unit-op models [aspen-plus]; Nishi conserved mass balance only
Aspen and COMSOL evaluate a continuous equation of state (Best); Nishi tabulated saturation plus latent heat, integer-exact
COMSOL 3D CFD is Best [comsol]; Nishi analytical Biot and Fourier lumped cold-spot verdict
Nishi 2-D transient explicit finite-difference conduction (cold-spot field over time, integer-deterministic); COMSOL 3-D CFD Best; Aspen dynamic sim
Nishi Henderson-Hasselbalch buffers plus molarity plus dilution plus titratable acidity, exact
Statistics quality and sensory
Minitab is the SPC Best with 100s of tests and a GUI [minitab-docs]; Nishi computes identical X-bar R limits and Cpk sovereignly
Nishi now computes 2^k factorial and fractional designs with main and interaction effects; Minitab remains the deepest DOE (Best)
Compusense and EyeQuestion Best (panel management plus descriptive analysis) [compusense]; Nishi computes the EXACT integer binomial from first principles vs a lookup table
Nutrition flavor and rheology
FoodData Central Best (400k foods plus micronutrient panels) [fooddata-central]; Nishi computes Atwater energy plus DV coverage from first principles
Nishi now computes vitamin and mineral content plus FDA percent-DV and rich-source labels for its foods; FoodData Central Best (400k foods full panels) so Nishi partial
Foodpairing Best (proprietary aroma-compound DB) [foodpairing]; Nishi shared-compound network (Ahn 2011, the same science) [ahn2011] plus generative design, sovereign
instrument vendor software Best; Nishi computes Texture Profile Analysis plus Herschel-Bulkley yield, no instrument
Sovereignty and method (the Nishi lane)
Nishi never phones home by construction; most incumbents are cloud or license-checked. EVIDENCE (nx_food_ground_gate, 8/8 GREEN, MEASURED 2026-08-14): the gate runs on the box against a durable LOCAL store (knowledge/store/food-) with no network in the path -- 26 ingredients live, 7 of 7 science rules cited to already-fetched sources. Ingestion is a separate, explicit step, so the working suite has nothing to phone home to
Nishi sovereign ELF with zero third-party deps; incumbents licensed or subscription. EVIDENCE (MEASURED 2026-08-14): nx_food_ground_gate was compiled by our own chain (nx_cc to nxasm to ELF, no LLVM, GCC or libc beneath) and RUN to 8/8 GREEN -- a 194,163 byte binary with nothing licensed underneath it. The same gate also enforces licensing on the INPUT side, which is the harder half: admissible sources (PD, CC-BY-SA, FACTS, US-Gov) are accepted while an all-rights-reserved source ingests ZERO rows, a flagged personal source stores nothing, and an UNRECOGNISED licence fails CLOSED rather than being assumed safe
unique to Nishi: never-poison and never-HARM gates that STRUCTURALLY refuse unsafe output, not a warning string
Nishi no-float determinism across the whole suite; float engines are nondeterministic by construction. EVIDENCE (MEASURED 2026-08-14): determinism is what makes the gates expressible at all, and that is the checkable form of the claim. nx_food_ground_gate asserts EXACT counts (26 ingredients, 7 of 7 rules cited, exactly 3 ingredient-pair FACTS, exactly 0 rows from each refused source), and the sibling water organs in this same suite assert exact integer physics (saturation density 304 and 173, extractable 144, yield 34560 ml, energy 208 Wh/L). A float engine cannot be tested with equality assertions; the fact that none of these are tolerance bands IS the evidence
Nishi is one suite bits-up across 14 disciplines vs a fragmented set of separately-licensed specialist tools. EVIDENCE (MEASURED 2026-08-14): the integration is exercised rather than asserted. nx_food_ground_gate walks one chain end to end -- source admission, licence adjudication, facts-only ingestion, a durable store, provenance, and science rules cited back to fetched sources -- and passes 8/8 including an additive re-ingest that leaves the store readable. The water lane composes across the same substrate in nx_water_design_test (15/15 GREEN), which imports the harvesting, quality and design organs into a single gate. Separately-licensed specialist tools cannot be composed into one gate at all
Where Nishi is honestly behind
Nishi ABSENT: it computes models but holds NO measured empirical curve or food database
Nishi ABSENT: no rheometer or HPLC or GC data acquisition
Nishi ABSENT: lumped and analytical heat transfer only, no 3D computational fluid dynamics
Nishi ABSENT: API and gate-first, no graphical UI
Nishi ABSENT: gated and grounded but NOT lab-validated or regulator-certified
Nishi ABSENT: self-host with no vendor support contract or SLA
Person · product · place — not yet measured for this domain
knowledge/compare/foodscience.ppp (rows surface|nishi or c1..c4|label|url|connect naming OUR live surface and each rival's front door), run nx_ppp_probe domain foodscience, and this section fills itself on the next beat: the same ruler on both sides — privacy and CX (third-party hosts, tracker classes, cookies, security headers), design and longevity (design hygiene, computed WCAG contrast, render-blocking resources, unsized media, script weight, theme and motion queries), findability (landmarks, skip link, on-site search, breadcrumb, headings, internal links).References
- [combase] ComBase (USDA ARS, University of Tasmania and Quadram Institute). ComBase -- Quantitative and Predictive Food Microbiology: the ComBase Browser of measured microbial growth and survival curves and the ComBase Predictor. Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_combase.html· pinhd40c2b2e5802160c3539da907226113a140edcd8fb854ab47340a723914ca8c7· accessed 2026-08-18 · datasetGrounds: The ComBase column: Thermal death 12-D botulinum cook and Predictive microbial growth pH temp aw are coded Best for ComBase because it holds thousands of MEASURED growth curves, which is precisely what the Nishi note concedes it lacks (structural hazard envelope, no measured curves) -- and the Measured empirical database row where Nishi is honestly ABSENT. - [fooddata-central] USDA Agricultural Research Service. FoodData Central -- a comprehensive source of food composition data with multiple distinct data types (Foundation Foods, SR Legacy, Branded). Accessed August 2026 -- mirrored once the sovereign TLS-1.2 leg was root-fixed the same day (debt 1787075987). publisher · read in our library
knowledge/fetched/cmp_foodscience_fooddata-central.html· pinh93554a35b716c148c34872ac35187449d7b383a4c918d684ae2fdccb40fabb8b· accessed 2026-08-18 · datasetGrounds: The FoodData Central column: Nutrition energy Atwater and daily value and Micronutrient content and percent DV are coded Best because FDC carries about 400k foods with full micronutrient panels; Nishi computes Atwater energy and DV from first principles for its own ingredients and is graded partial against this database. - [ahn2011] Ahn, Y.-Y., Ahnert, S.E., Bagrow, J.P., Barabasi, A.-L. Flavor network and the principles of food pairing. Scientific Reports 1, 196 (2011); arXiv:1111.6074. publisher · read in our library
knowledge/fetched/cmp_foodscience_ahn2011.html· pinh41e3d54220508e1e21276cb422d93da5db6e18aca0d0c0bddcac5854715c1074· accessed 2026-08-18 · published-paperGrounds: Flavor pairing shared-compound network -- the note says Nishi's shared-compound network is Ahn 2011, the same science; this is that paper: the flavor network of ingredients linked by shared aroma compounds and the food-pairing hypothesis it tests, which Foodpairing commercialises with a proprietary compound database (Best). - [leistner2000] Leistner, L. Basic aspects of food preservation by hurdle technology. International Journal of Food Microbiology 55(1-3):181-186, 2000. PMID 10791741, doi 10.1016/S0168-1605(00)00161-6. Confirmed by search-index read; PubMed serves a bot-challenge stub to non-browser fetchers, so no mirror is claimed. publisher · accessed 2026-08-18 · published-paperGrounds: Water activity and hurdle stability -- the note credits Nishi with implementing the Leistner combined-hurdle model exact; this is Leistner's own statement of the hurdle concept (temperature, aw, pH, redox, preservatives, competitive flora and their interactions, multitarget preservation).
- [minitab-docs] Minitab, LLC. Minitab Statistical Software support and help (control charts, capability analysis Cp and Cpk, design of experiments). Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_minitab-docs.html· pinh93a466e730a56149ca9cb3266e49a84a6eb7e5dff191458eb4d0040e6d484383· accessed 2026-08-18 · vendor-docGrounds: The Minitab column: SPC control charts Cp Cpk and Design of experiments are coded Best for Minitab (hundreds of tests, GUI, the deepest DOE); Nishi's X-bar R limits, Cpk and 2^k factorial designs are graded against what these help pages document, and Minitab is also the GUI and no-code usability Best. - [aspen-plus] AspenTech. Aspen Plus -- process simulation with rigorous unit-operation models and equations of state. Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_aspen-plus.html· pinh818a93a5ce9ba6ee0a76bce9d4913bb4b665907a596fa1d1f7c1a61809752edf· accessed 2026-08-18 · vendor-docGrounds: The Aspen Plus column: Mass and energy balance unit ops (Aspen Best with rigorous unit-op models; Nishi conserved mass balance only) and Steam tables and thermodynamics (continuous equation of state vs Nishi's integer-exact tabulated saturation plus latent heat). - [comsol] COMSOL AB. COMSOL Multiphysics -- finite-element multiphysics simulation including heat transfer and CFD. Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_comsol.html· pinh9d4818499508f7fac6625704a3ba388ac1c56b29a65c248c696bd27bfbf71b40· accessed 2026-08-18 · vendor-docGrounds: The COMSOL column: Heat transfer cold-spot Biot Fourier and 2-D transient field simulation are coded Best for COMSOL 3-D CFD; Nishi's analytical Biot-Fourier lumped verdict and 2-D explicit finite-difference conduction are graded against it, and 3D CFD simulation is a row where Nishi is honestly ABSENT. - [compusense] Compusense Inc. Compusense sensory and consumer research software (panel management, difference tests, descriptive analysis). Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_compusense.html· pinh58e7d169cc8269d8ec0d8a6bc565d183c33a2ee98826fd743ba96ab7b8ee2bdf· accessed 2026-08-18 · vendor-docGrounds: The Compusense column: Sensory difference tests triangle binomial is coded Best for Compusense (panel management plus descriptive analysis); Nishi computes the exact integer binomial for triangle tests from first principles and is graded against this vendor's documented scope, and Compusense is the only column with Lab instrument integration coded Yes. - [foodpairing] Foodpairing NV. Foodpairing -- data and business intelligence for flavor and product design on a proprietary aroma-compound database. Accessed August 2026. publisher · read in our library
knowledge/fetched/cmp_foodscience_foodpairing.html· pinh893ff319fc1ba46281094d107295c49e44495121b96a932f54465361d30515e9· accessed 2026-08-18 · vendor-docGrounds: The Foodpairing column: Flavor pairing shared-compound network is coded Best for Foodpairing's proprietary aroma-compound DB; the vendor page documents that database as the product, which is what Nishi's open Ahn-2011-style network plus generative design is graded against.
Generated by nx_swcompare_sota from knowledge/compare/foodscience.sota — quantitative axes measured/sourced; researcher-fed (nx_swcompare_research). Zero JS, zero trackers.
Where we are. Measured 2026-08-21 from foodscience.sota and the organ sources on disk. STRONG and gate-proven: 14 disciplines composed in one integer-exact suite where each incumbent is deep in one or two, with structural refusals rather than warning strings -- a low-acid food may never be water-bath canned (pv_canning_admit), the Leistner combined hurdle is proven to BEAT either single factor (pv_hurdle_beats_single), Atwater energy and FDA percent-DV are computed from first principles rather than read off a label, and nx_food_ground_gate runs 8/8 GREEN against a durable local store with no network in the path while refusing inadmissible source licences fail-closed. WEAK, and the page already says so in six rows: THE SUITE HOLDS NO MEASURED DATUM ANYWHERE. That single absence wears two masks -- no measured microbial growth or death curve on the safety half, and no measured food-composition record on the nutrition half, where nx_nutrition carries sixteen foods as hand-entered constants. Beyond it: the conduction solver is two-dimensional where the field is three, no organ owns unit operations at all, no instrument export is ever read, and the citation machinery that already grounds 7 of 7 science RULES does not reach the CONSTANTS. THE HONEST BOUNDARY, stated here rather than smuggled into a rung: commercial support and an SLA are not an engineering rung and never will be, and a GUI is owned by the surface lanes, not by this suite -- neither appears in foodscience.matrix. THE STATISTICS AND SENSORY HALF IS ALSO ABSENT FROM THIS LADDER ON PURPOSE: SPC depth against Minitab and sensory depth against Compusense are real Part grades on the sota page, but a matrix carries ONE @cols line and those rungs need Minitab and Compusense as rivals while these need ComBase and COMSOL. Naming them here against the wrong rival field would be the measured-a-different-subject defect; they are stated as owed rather than mis-scored.
Where we need to go. Move this suite from A MODEL NOBODY HAS CHECKED AGAINST REALITY to one whose predictions are compared with measured data and whose constants each name their source. Concretely: measured composition records feed the same NxFood the hand catalog uses, predicted lethality is compared against measured curve points with the signed error printed, conduction is solved in three dimensions with the two-dimensional incumbent as its degenerate-case oracle, unit operations get an organ instead of nothing, an instrument export parses into the inputs the rheology organ already takes, and every named constant maps to a cited mirrored source with the uncited ones listed. The exceed this suite holds is that its gates assert EXACT integer equalities; every rung below must stay expressible that way or it has broken the property that makes the rest checkable.
| Rung | Closes with | Definition of done (pre-declared) | Executor | Est. |
|---|---|---|---|---|
| 12-D botulinum cook with a structural refusal (FS0A) | pv_canning_admit | LANDED. D-value z-value and the 12-D cook computed exactly, and the load-bearing part is not the arithmetic but the REFUSAL: a low-acid food routed to a water bath is refused by construction rather than warned about [@combase]. Listed so dependents carry its momentum | Organ | 0 u |
| Leistner combined-hurdle stability (FS0B) | pv_hurdle_beats_single | LANDED. The hurdle model implemented exact [@leistner2000], with the discriminating tooth being that a combined pH and aw hurdle is proven to beat EITHER single factor -- an implementation that merely returned stable for both would pass a suite that only checked the combined case | Organ | 0 u |
| Atwater energy and daily value from first principles (FS0C) | nt_kcal_per_100g | LANDED. Energy computed from macronutrients rather than read from a label, scaled to portions and whole meals with FDA percent-DV coverage [@fooddata-central]. Listed so dependents carry its momentum | Organ | 0 u |
| Measured food-composition record ingest (FS1) after FS0C | nt_fdc_ingest | nx_nutrition holds sixteen foods as hand-entered constants, which is precisely why the sota grades the measured empirical database ABSENT and FoodData Central Best [@fooddata-central]. ACCEPT RULE PRE-DECLARED: a mirrored FoodData Central record parses into the SAME NxFood struct the hand catalog uses and is scored by the EXISTING nt_kcal_per_100g -- one ruler, never a second energy computation for ingested foods. THE LOAD-BEARING TOOTH IS AN EQUALITY, not a tolerance: for a food that exists in BOTH the hand catalog and the ingested set, the Atwater energy computed from the ingested macros must equal the energy the organ already computes for its hand-entered twin EXACTLY, because a suite whose exceed is integer determinism may not accept a band here. A record missing a macronutrient yields UNKNOWN and never zero, since zero fat is a claim and a missing field is not. The partition must print and sum: records read equals ingested plus refused plus unparseable. This rung closes the sota's Micronutrient content Part grade in the same pass and the row note says so | Organ | 1.5 u |
| Predicted lethality validated against measured curve points (FS2) after FS0A | pv_curve_validate | We gate the hazard envelope structurally and hold no measured curve, which is the entire reason ComBase is Best on two rows of this page [@combase]. ACCEPT RULE PRE-DECLARED: at declared pH, water-activity and temperature points taken from a mirrored and content-pinned measured source, the model's predicted log reduction is compared with the measured value and the SIGNED error is printed per point -- signed, never absolute, so a model that is systematically optimistic shows as a bias instead of averaging to nothing. A point outside the model's declared validity envelope reports OUT-OF-ENVELOPE as its own third state and does NOT score as agreement, because an axis that cannot see must abstain rather than acquit. ANTI-VACUITY TOOTH REQUIRED: a deliberately wrong z-value must push at least one point outside the declared band, or the comparison has not been shown capable of disagreeing | Organ | 1.5 u |
| Transient conduction in the third dimension (FS3) | hc3d_run | hc2d_run is a real explicit finite-difference transient solver and it is two-dimensional, which is why the sota grades 2-D transient Part and 3-D CFD absent against COMSOL Best [@comsol]. EXTENDS THE INCUMBENT ORGAN -- a second conduction solver is refused by construction. ACCEPT RULE PRE-DECLARED, and its two teeth must BOTH hold because either alone is passable by a wrong implementation: with the third dimension one cell thick the new solver reproduces hc2d_run's centre temperature EXACTLY at every step (the degenerate-case oracle, an equality not a band), AND for a genuinely three-dimensional geometry -- a cube against an infinite prism of the same cross-section -- the centre must reach the surface temperature SOONER, because a cube loses heat through six faces and a prism through four. A solver that simply returned the 2-D answer would pass the first tooth and fail the second, which is what makes the pair a discrimination test rather than two spot checks | Organ | 2 u |
| Unit-operation mass and energy balance (FS4) | uo_balance_solve | The suite conserves mass and stops there, against Aspen Plus rigorous unit-op models at Best [@aspen-plus]. NEW ORGAN DECLARED IN THE ROW rather than bolted onto a thermal organ, because no organ in the suite owns unit operations and single responsibility is cheaper to keep than to restore. ACCEPT RULE PRE-DECLARED: for a declared unit operation both balances close EXACTLY in integer units -- mass in equals mass out plus accumulation, energy in equals energy out plus accumulation -- and a stream whose composition is under-specified REFUSES with the missing quantity named rather than closing the balance by inventing the remainder, which is the one failure mode that makes a balance solver worse than no balance solver. NEG-CONTROL: a deliberately unbalanced fixture must be refused and must name which balance failed | Organ | 2 u |
| Instrument export parsed into the rheology inputs (FS5) | rh_instrument_ingest | Texture Profile Analysis and Herschel-Bulkley yield are computed from values typed by hand; the sota grades lab instrument integration ABSENT. ACCEPT RULE PRE-DECLARED: a banked instrument export parses into the inputs nx_rheology already takes and reproduces the organ's EXISTING hand-entered fixture EXACTLY -- the parser is proven by agreeing with a known answer, not by producing a plausible one. A malformed export is refused NAMING the field that failed, never partially accepted. SCOPE DECLARED: this rung reads a file an instrument wrote and drives no hardware, so it closes the data half of that row and leaves acquisition open, and the row note says so rather than letting the flip imply more | Organ | 1 u |
| Every named constant maps to a cited source (FS6) | fd_constant_cite_map | fd_sci_cited and fd_sci_cite_of already ground 7 of 7 science RULES; the CONSTANTS -- the botulinum D-value, the z-value, the water-activity thresholds, the Atwater factors, the daily values -- are not covered, and the sota grades published validation ABSENT. ACCEPT RULE PRE-DECLARED: every named constant in the suite's organs maps to a cited mirrored source or is listed UNCITED by name and file, the partition of constants prints and sums, and the uncited count is a SHRINK-ONLY ratchet that self-seeds on first sight -- never a fresh absolute bar, because a detector that is red on the day it ships is one everyone learns to ignore. THE DIRECTION NOBODY AUDITS IS THE ONE THIS RUNG MUST GET RIGHT: an omitted constant flatters the number, so the enumeration is of the SOURCE and the coverage is computed against it, never the other way round. This rung claims coverage and never claims certification | Organ | 1 u |
Milestones
| Milestone | Rungs | Cumulative |
|---|---|---|
| F0 · The suite answers to measured data | FS1,FS2 | 3 u |
| F1 · Physics depth | FS3,FS4 | 7 u |
| F2 · Instruments and provenance | FS5,FS6 | 9 u |
Risk register
| Risk | Likelihood x impact | Mitigation |
|---|---|---|
| Ingesting a measured database makes the suite look validated when only its inputs changed | possible x high | FS1 changes where numbers come from and nothing about the science. The row it closes is the measured DATABASE row; the validation row is FS6 and the curve-agreement row is FS2, and each states its own narrower claim. No rung is permitted to cite another rung's flip as its own evidence. |
| A comparison against measured curves is reported as agreement when the model is simply inside a wide band | likely x medium | FS2 prints the SIGNED error per point and no aggregate pass mark, and its anti-vacuity tooth requires that a deliberately wrong z-value be visible. A comparison that could only ever agree has measured nothing. |
| A third conduction dimension is added by wrapping the 2-D solver in a loop and reads as 3-D | likely x high | FS3's second tooth is the whole defence: a cube must cool faster at the centre than a prism of the same cross-section. The degenerate-case oracle alone would pass a wrapped 2-D solver, which is why both teeth are mandatory and neither may be dropped for being redundant. |
| The constant-coverage ratchet is satisfied by deleting constants rather than citing them | possible x medium | FS6 prints the partition and the total, so a falling uncited count against a falling total is visible as such; a constant that moved into a literal is a magic number and nx_magic already owns that detector, so the two are read together rather than one being widened. |
Watch contracts (measured)
| Axis | Organ | Symbol | Status | Note |
|---|---|---|---|---|
| Thermal death 12-D botulinum cook | runtime/nx_preservation.nx | pv_canning_admit | PRESENT decl | LANDED. D-value z-value and the 12-D cook computed exactly with a STRUCTURAL refusal -- a low-acid food may never be water-bath canned. ComBase is Best because it holds the measured curves we do not [combase]. Listed so dependents carry its momentum |
| Water activity and hurdle stability | runtime/nx_preservation.nx | pv_hurdle_beats_single | PRESENT decl | LANDED. The Leistner combined-hurdle model implemented exact [leistner2000] -- the tooth that matters is that a combined pH and aw hurdle is proven to beat either single factor, which a lookup table cannot express |
| Nutrition energy Atwater and daily value | runtime/nx_nutrition.nx | nt_kcal_per_100g | PRESENT decl | LANDED. Atwater energy and FDA percent-DV computed from first principles rather than read off a label. FoodData Central is Best on breadth with about 400k foods [fooddata-central] |
| Measured empirical database | runtime/nx_nutrition.nx | nt_fdc_ingest | WATCHING decl | OPEN -- rung FS1. The suite's largest self-declared absence: it computes models and holds NO measured food record. nx_nutrition carries a hand-entered 16-food catalog as constants. The contract is ingesting a mirrored FoodData Central record into the SAME NxFood shape. Closes the sota's Micronutrient content Part grade in the same pass [fooddata-central] |
| Predictive microbial growth pH temp aw | runtime/nx_preservation.nx | pv_curve_validate | WATCHING decl | OPEN -- rung FS2. We gate the hazard envelope structurally and hold no measured growth curve, which is exactly why ComBase is Best here [combase]. The contract is validating the model against measured curve points with the signed error printed per point |
| 3D CFD simulation | runtime/nx_heat_conduction2d.nx | hc3d_run | WATCHING decl | OPEN -- rung FS3. hc2d_run is a real explicit finite-difference transient solver in two dimensions; the third is missing, which is why the sota grades 2-D transient Part and 3-D absent against COMSOL Best [comsol]. Extends the incumbent conduction organ rather than adding a second solver |
| Mass and energy balance unit ops | runtime/nx_unitop_balance.nx | uo_balance_solve | WATCHING decl | OPEN -- rung FS4. Conserved mass balance only today against Aspen Plus rigorous unit-op models at Best [aspen-plus]. NEW ORGAN DECLARED: no organ in the suite owns unit operations, so this contract names the file it will live in rather than bolting a second concern onto a thermal organ |
| Lab instrument integration | runtime/nx_rheology.nx | rh_instrument_ingest | WATCHING decl | OPEN -- rung FS5. Texture Profile Analysis and Herschel-Bulkley yield are computed from hand-entered values; no instrument export is read. The contract is a parser into the inputs the organ already takes, not a driver for hardware we do not have |
| Published validation and certification | runtime/_hdl_build/nx_food_provenance.nx | fd_constant_cite_map | WATCHING decl | OPEN -- rung FS6. fd_sci_cited and fd_sci_cite_of already cite 7 of 7 science RULES. The contract extends that to every named CONSTANT in the suite with the uncited ones listed -- coverage is claimable and certification is not |
watch rows=9 landed=0 watching=6 present=3 missing=0 absent=0 (partition sums)