Skip to content

Field Completeness and Data Quality

Established: 2026-04-05 | Data snapshot: 2026-04-04

Summary

TechPort's data quality varies significantly by program and by field. This page maps what's reliably populated versus what's missing, and documents known data anomalies. The quality map is the foundation for Phase 2 investigations — focus on programs where the data is trustworthy.

Field Coverage by Program

Overall Portfolio

Field Coverage Notes
status ~100% All projects have status
program ~100% All projects have a program
trlCurrent 73.6% 24.4% None + 2.0% TRL-0 = 26.4% missing
primaryTx 95.9% 816 projects (4.1%) have no TX assignment
destination ~65% 29.7% empty string + 5.2% None = ~35% missing
leadOrg ~98% 384 (1.9%) missing
leadOrgType ~98% tied to leadOrg
description ~95%+ some very thin records (e.g., GCD batch entries)
contacts ~96%+ most projects have at least 1 contact
contactEmails ~80%+ estimated some contacts missing email (e.g., PI emails sometimes omitted)
libraryItems unknown globally varies hugely; NIAC/SBIR completed: 2-6 typical; active GCD/FO: often 0
technologyOutcomes ~50%+ 49.1% have Closed_Out; total with any outcome unknown
trlBegin / trlEnd ~same as trlCurrent progression data tied to TRL population
benefits ~80%+ estimated most mature projects have benefits; thin records missing
closeoutDetails unknown exists for some SBIR; rare elsewhere

By Program

Program TRL Destination Description Contacts Library Items Outcomes
STRG ★★★★★ 98.7% ★★★ (not checked) ★★★★ good ★★★★ multi-PI ★★ few docs ★★★★ most tracked
FO ★★★★ 93.8% ★★★★ (good) ★★★★ good ★★★ 1-5 contacts ★★★ some docs ★★★ moderate
NIAC ★★★★ 85.6% ★★★★ (good) ★★★★★ rich ★★★★★ multi-PI w/ emails ★★★★ 3-6 items ★★★★★ 95%+
SBIR/STTR ★★★ 66.7% ★★★ moderate ★★★ variable ★★★★ PI + PM ★★★ variable ★★★★ most tracked
GCD ★★ ~55% real ★★ (unknown) ★★ many thin records ★★★ mostly 1 ★ few ★★ patchy
CIF ★★★ (not checked) ★★★ (not checked) ★★★ (not checked) ★★★★ (good) ★★★ (not checked) ★★★★★ high Transitioned_To

Known Data Quality Issues

1. Empty String Destination (29.7% of projects)

The destination field has two distinct missing-value states: a Python None/null (5.2%, shown as "(none)" in aggregates) and an empty string "" (29.7%, shown as blank in aggregates). These are separate entries in the data but both represent "not set." Total unset destination: ~35%.

Impact: Destination-based filtering will miss ~35% of projects. Moon/Mars/Earth distributions should be read as "of projects that have destination set" = 30.8% Earth, 14.9% Moon, 13.1% Mars (renormalized from 65% of total).

Confidence: confirmed (direct aggregate query separates both values)

2. TRL-0 (2.0% = 399 projects)

NASA's TRL scale runs 1-9. TRL-0 does not correspond to a defined technology readiness level. The tool documentation explicitly notes this: "~399 projects have trlCurrent=0 (NASA TRL scale starts at 1 — treat 0 as 'not set')."

Concentrated in GCD (77 projects, 16.5% of GCD) — likely a data entry artifact from batch project creation.

Impact: Never interpret TRL-0 as a real TRL. When filtering TRL, use trl_min=1.

Confidence: confirmed (per tool documentation; supported by GCD context)

3. Test Record in Production — Project 183893

Project 183893 (_TEST_REV A) in GCD program: - Name starts with underscore: clearly a test/scratch entry - Period: 2020-07-01 to 2020-07-01 (zero duration) - No TRL, no lead org, no contacts, no outcomes - Description content appears to be from a real composite materials/nozzle project — possibly copy-pasted during a test - Views: 633 — externally visible

Impact: This is a data entry artifact in the live database. Flagged as a data quality issue. It will not significantly affect aggregate counts (1 of 468 GCD projects) but represents a broader risk that other thin/malformed GCD records may exist.

Confidence: confirmed (direct project record inspection)

4. TX Human-ML Classification Mismatch

The tool reports txMismatch for projects where a human assigned a different TX area than the ML model predicts.

Mismatch rate (SBIR/STTR sample n=50): 2/50 = 4% mismatch rate

Query: techport_find_projects(program="SBIR/STTR", status=null, limit=50, fields=["projectId","title","primaryTx","mlPredictedTx","txMismatch"]) | 2026-04-05

Observed mismatches: - Project 8793 ("Accelerating ATM Optimization Algorithms"): Human = TX01.1.3 (Cryogenic Propulsion), ML = TX15.1.3 (Aeronautics). Human is clearly wrong — an ATM optimization algorithm has nothing to do with cryogenic propulsion. - Project 16664 ("Active Combustion Control Valve"): Human = TX12.3.2 (Electromechanical Devices), ML = TX01.1.1 (Chemical Propulsion). Ambiguous — the valve is a mechanism (TX12) but used in a propulsion combustion context (TX01).

Earlier examples (from session 1): - Project 184638 (GCD, H2/O2 test facility): Human = TX13.2.8 (Environment Testing), ML = TX01.1.3 (Cryogenic Propulsion). ML more accurate. - Project 13744 (NIAC, Balloon Reflector): Human = TX13.2.1 (Mechanical Testing), ML = TX08.1.3 (Optical Components). ML clearly more accurate.

Additional confirmed mismatches (session 18): - Project 158701 (Green Mountain Semiconductor, "Radiation Hard Network on Chip Neural Processor with RRAM"): Human = TX03.2.3 (Advanced Concepts for Energy Storage), ML = TX02.1.1 (Radiation-Hardened Extreme-Environment Components). txMismatch: true. This is an AI edge inference chip using compute-in-memory RRAM architecture — nothing to do with energy storage. - Project 9640 ("Statistical Decision Support Tools for System-Oriented Runway Management"): Human = TX03.1.1 (Photovoltaic Electrical Power). Air traffic management runway decision tools have no connection to photovoltaic power — an early-era classification error. - Project 102915 (Redwire MVP, "Multi-Purpose Variable-gravity Platform"): Human = TX03.3.1 (Management and Control). The MVP is an ISS dual-centrifuge for biology/fluids experiments — a life science research platform, not a power management system. Should be TX06 (Human Health) or TX04 (Robotics).

Pattern: TX mismatch tends to occur when projects span domains or when the human classifier focused on the application context (testing, systems) rather than the core technology (the sensor, the propellant). ML tends to classify on core technology; humans tend toward application domain. Early-era records (pre-2013) have the most egregious mismatches, suggesting classification rigor improved over time.

Rate estimate: ~4% of SBIR/STTR projects have explicit mismatch. The overall "primaryTx missing" rate is 4.1% (816 projects). Combined: approximately 8% of projects either lack TX or have a questionable TX assignment.

CIF TX missing rate is much worse: 26.7% of CIF projects have no TX assigned — the highest missing rate of any program.

Impact: TX-based search will miss some relevant projects (~4% mismatch + ~4% missing = ~8% blind spot for SBIR/STTR). For CIF, ~27% blind spot. Cross-check with ML classification when doing specific technology searches. The txMismatch field is available in techport_find_projects to surface questionable classifications.

Confidence: suggestive for rate; confirmed for pattern (SBIR n=50; individual mismatch examples confirmed by inspection)

5. Geographic Data Errors

Project 113005 (SBIR/STTR): Lead org VisSidus Technologies listed as "Daytona Beach, Hawaii." Daytona Beach is in Florida, not Hawaii. The multi-state field shows CA, HI, TN, TX — Hawaii is a co-investigator state (Univ. Hawaii), not the lead org location.

Impact: Location-based filtering (state, city) unreliable for small org records. Do not rely on leadOrgCity for precise location data.

Confidence: confirmed (direct record inspection)

6. Contact Email Field Anomalies

Project 106026 (NIAC, Neutrino Detector): PI Nickolas Solomey (Wichita State University) listed with email nickolas.solomey@nasa.gov — a nasa.gov email for a university PI. This suggests either the PI has a NASA guest account or the email was entered incorrectly.

Impact: Contact emails are often accurate and useful, but cannot be treated as perfectly clean. PI emails in particular may be outdated (contracts end, people move).

Confidence: confirmed (direct record inspection; single case)

7. GCD Batch Records (IDs 183xxx, 184xxx)

A large cohort of GCD projects were entered/updated 2026-03-23. Many in this batch show: - TRL 0 or no TRL - Single contact (program contacts only, no PI/PM) - No library items - No outcomes - Some with no description

These appear to be records created during a GCD portfolio review/restructuring rather than mature project records. Several were immediately "Completed" after creation.

Impact: GCD data quality analysis for recent records should treat these with caution. The older GCD projects (IDs <130000) are generally much richer.

Confidence: suggestive (observed pattern across multiple records; not exhaustively verified)

Based on the quality map above:

Trust: NIAC descriptions, contacts, outcomes, TRL, library items. STRG TRL and TX. FO TRL and descriptions.

Use with caveats: SBIR/STTR (good data for completed Phase II, patchy for Phase I). NIAC/SBIR document library (present but varying depth).

Verify before using: GCD TRL for recent records. All destination fields (normalize for empty string). TX assignments (cross-check with ML prediction for any specific technology search).

Avoid or investigate separately: GCD batch records (IDs 183xxx-184xxx, updated 2026-03-23). Location fields for small private companies. Contact emails for completed projects.

8. TRL-9 Semantic Inflation (SBIR)

NASA's TRL 9 definition: "actual system flight proven through successful mission operations." In SBIR records, TRL 9 is sometimes self-reported to mean commercial productization in a non-NASA domain.

Confirmed case: Project 113042 (Electric Power Systems Active BMS, TX03.2.1, TRL 6→9 in 11 months, 2021-2022). Document (fileId 377880) states: "The purpose of this project is to mature the active BMS concept towards productization in the EVTOL market and for use with the Bell e4D project as the commercial development partner." Non-NASA applications: Bell Nexus, Embraer Sky, Boeing Personal Air Vehicle. TRL 9 = aviation commercial maturity, not NASA space flight proven.

Pattern: When a SBIR company has an existing commercial product and NASA funds further development, the company may report TRL 9 when the technology is integrated into a commercial product — even if no NASA mission has flown. This creates a misleading tail in TRL distributions.

Impact: TX03 SBIR shows 3 projects at TRL 9; only 1 (MicroLink ELO solar cells, 112889) appears to represent genuine space hardware qualification. The other 2 (113042 EPS BMS, 102915 Redwire MVP) have data quality caveats. High-TRL counts for SBIR are likely overstated by ~30-50% relative to space-mission TRL 9.

Confidence: confirmed for the single EPS BMS case; pattern is suggestive pending broader survey.

9. outcome_path Search Filter Inconsistency

Finding (2026-04-05): The outcome_path filter in techport_find_projects does not perfectly match the technologyOutcomes array in individual project records.

Example: Project 11566 (ISRU Robotic Construction) was returned by an outcome_path=Infused_To filter query, but the full project record (retrieved via techport_get_project) has no Infused_To entry in its technologyOutcomes array.

Impact: Outcome_path filter queries may include false positives. The reported counts (115 Infused_To, 634 Transitioned_To) may be slightly overstated. When precise outcome identification is needed, verify by reading the full project record.

Confidence: confirmed (single verified false positive; broader prevalence unknown)

10. Cross-Project Description Contamination — FAME/SEADS

Finding (2026-04-06, session 46):
Project 155247 (FAME — Fluids at Altitude Mixing Experiment, Redwire, FO) contains text in its description dated "2024-03-01" that describes an EHD printing demo — clearly content from 155248 (SEADS — Space Enabled Advanced Devices and Semiconductors) accidentally appended. FAME and SEADS have consecutive IDs and share the same program (FO) and PM, suggesting the contamination occurred during a batch data entry or copy-paste operation.

Impact: FAME's description is partially incorrect — it describes technology from a different project. Any keyword search using EHD-related terms will surface FAME as a false positive. FAME's actual content (pharmaceutical fluidics, PIL-BOX DMC crystallization platform) is still present in the description; only the appended EHD text is erroneous.

Confidence: confirmed (direct comparison of FAME and SEADS full project records; timestamps in description match SEADS timeline, not FAME's)

11. Coordinated Human+ML Misclassification — Undetected by Mismatch Flag

The txMismatch flag catches disagreements between human-assigned TX and ML-predicted TX. It cannot catch cases where both sources independently arrive at the same wrong classification.

Confirmed case: PPR — Pulsed Plasma Rocket (158619) - Human-assigned TX: TX01.4.4 Solar Thermal Propulsion - ML predicted TX: TX01.4.4 Solar Thermal Propulsion - txMismatch: False (no alert raised) - Actual technology: nuclear pulse propulsion — HALEU/HEU barrel-bullet criticality mechanism, 100,000 N thrust / 5,000s Isp (document read of poster 317211 confirmed)

Why it happens: PPR's description uses language like "plasma," "propulsion," "high ISP" without leading with "nuclear pulse." TX01.4.4 Solar Thermal Propulsion has been used historically as a catch-all for exotic propulsion concepts that don't fit chemical/EP bins cleanly. The ML model, trained on descriptions and existing TX labels, learned the same association and returned the same wrong answer. Both human classifier and ML were fooled by the same mismatch between description language and the actual physics.

Significance: This failure mode is structurally harder to detect than a flagged mismatch. A search for nuclear propulsion using TX01.3.x (Nuclear Propulsion) would miss PPR entirely, with no warning from the mismatch flag. Similar coordinated failures likely exist for other niche concept categories that don't fit standard TX bins (compare: Photophoretic Propulsion classified TX01.4.3 Nuclear Thermal by ML — also completely wrong).

Extended pattern across PPR lineage (3 projects, 10 years): - PuFF Phase I 13725: human=TX01.4.4 Solar Thermal, ML=TX01.4.4 Solar Thermal — mismatch=False, both wrong - PuFF Phase II 95694: human=TX01.1.8 Warm Gas, ML=TX01.4.4 Solar Thermal — mismatch=True, both wrong (different bins) - PPR 158619: human=TX01.4.4 Solar Thermal, ML=TX01.4.4 Solar Thermal — mismatch=False, both wrong

The same underlying technology was classified into three different wrong bins across 10 years. The mismatch detector oscillates True/False but never surfaces the correct classification. TX01.4.4 Solar Thermal appears to function as a catch-all for "advanced propulsion I don't know how to classify."

Counter-query: Search TX01.4.4 Solar Thermal active projects and check descriptions for non-solar-thermal technologies. PPR is the only NIAC active project with this TX; but this pattern may exist elsewhere in SBIR (where human classifiers work at scale and TX01.4.4 may absorb other exotic concepts).

Confidence: confirmed (direct document read; project record inspection; query parameters: get_project(158619, fields=["primaryTaxonomyAreas","mlPredictedTx","txMismatch"]))

Pattern discovered: PuFF Phase I 13725 has an "Advanced_To" outcome pointing to UAH ChargerSat-2 Parabolic Flight Testing 91348. ChargerSat-2 is a student CubeSat project studying nucleate boiling in microgravity — entirely unrelated to pulsed fission-fusion propulsion.

The likely mechanism: UAH (University of Alabama in Huntsville) was a listed partner in PuFF Phase I. ChargerSat-2 is also a UAH-led Flight Opportunities project running concurrently (2013-2016). The TechPort "Advanced_To" link appears to have been created based on shared institutional affiliation, not genuine technology transfer.

Distinguishing characteristics of spurious links: - Different technology areas between linked projects (TX01 propulsion → TX01.1.1 integrated systems with no propulsion content) - Different PIs (Robert Adams on PuFF; Francis Wessling on ChargerSat-2) - Different programs (NIAC → FO) - ChargerSat-2 explicitly describes its content as drag device deployment + boiling instrument verification — no propulsion component - ChargerSat-2 was ultimately Canceled (Jun 2016), with TRL 4 = TRL 4 (no advancement)

Implication for data use: The "Advanced_To" field in TechPort may contain institutional connections, not just genuine technology transfer. When tracing lineage via technologyOutcomes, verify that the linked projects share content (not just organizational partners). "Advanced_From" entries with "partner: Other" are particularly suspect — "Other" here likely means "educational institution" rather than a technology-relationship descriptor.

Counter-query: Check other NIAC projects with "Other" partner-type Advanced_To outcomes for similar spurious links. The NIAC program has a small number of such records (PuFF is the only confirmed case in the current KB).

Confidence: confirmed (direct inspection of both projects: [13725] partners include UAH; [91348] description contains no propulsion content; confirmed via get_project(91348) live call, session 56)

Cross-reference: See topics/propulsion-theme.md — PuFF Phase I section; ChargerSat-2 canceled June 2016.

13. TX Taxonomy Structural Gap — No Bin for Upper-Atmosphere Non-Space Vehicles

Discovery (session 55, 2026-04-06): Photophoretic Propulsion 182464 (NIAC Phase I, UPenn, Igor Bargatin, 2025-2027) describes vehicles that fly at 50-90 km altitude (mesosphere) using photophoretic lift — sunlight heating the underside of plate metamaterials in rarefied gas generates lift without any propellant or moving parts. The concept is explicitly for Earth's upper atmosphere, not space.

The classification problem: There is no TX area for upper-atmosphere non-space vehicles. The full NASA TX taxonomy is designed around: space propulsion, space power, space communications, Earth entry/descent, robotics, life support — all implicitly space or Earth-launch-context. The mesosphere "ignorosphere" (50-90 km, too high for aircraft, too low for orbit) is a unique environment that falls through every category.

How the gap manifested: - ML predicted: TX01.4.3 Nuclear Thermal Propulsion — completely wrong; no nuclear component; no space propulsion context - Human assigned: unclear from available data (TX not visible as a clean match either) - The misclassification is a symptom: with no correct bin, classifiers default to the nearest unusual propulsion category

Scope of gap: This is a structural gap (no correct bin exists), not a misclassification (wrong bin chosen). It differs from Issue 11 (PPR — a correct bin exists but both human and ML chose the wrong one). For photophoretic mesosphere vehicles, there is no right answer available in the taxonomy.

Related NIAC pattern: NIAC is the program most likely to surface projects that genuinely don't fit the taxonomy — by design, NIAC funds concepts at the frontier of what NASA considers its mission scope. Phase I projects in the mesosphere, interstellar medium, or biosphere may not map to any TX area.

Confidence: confirmed (direct description read; no TX area covers upper-atmosphere non-space vehicles; get_project(182464) live call, session 55)

14. Project Tethys — Mars ISRU Classified as Electrostatic Propulsion (session 58)

Project: 158665 "Project Tethys: Extracting Water from the Martian Environment" — Worcester Polytechnic Institute, STRG, TRL 2→3, 2024-2028.

Classification error: Human assigned TX01.2.2 Electrostatic Propulsion. ML predicted TX07.1.3 Resource Processing for Mission Consumables — ML is correct.

Why it's wrong: The project describes extracting water from Martian regolith for crewed mission support — ISRU (TX07), not propulsion. TX01.2.2 covers ion thrusters and electrospray thrusters. This is a complete content mismatch, not an ambiguous boundary case. Makes this project invisible in any TX07 ISRU search.

Likely cause: Human classifier data entry error — TX01 vs TX07 row selection mistake.

Confidence: confirmed (get_project(158665), session 58; txMismatch=Yes confirmed)

15. Penn State Beamed Microwave Propulsion Classified as Nuclear Thermal Propulsion (session 58)

Project: 118459 "Beamed Microwave Energy Propulsion Leveraging Lunar Resources" — Penn State, STRG, TRL 2→3, 2022-2026.

Classification error: Human assigned TX01.4.3 Nuclear Thermal Propulsion. ML predicted TX07.1.3. Neither classifier is right.

Why it's wrong: Concept is microwave beamed-energy propulsion from the lunar surface using in-situ resources — should be TX01.4.6 (Advanced Energetic Propulsion Approaches). No nuclear component exists. TX01.4.3 NTP covers fission-reactor-heated hydrogen propellant, entirely different physics.

Significance: TX01.4.3 is a small, high-signal area. A non-nuclear project classified here inflates NTP activity and will appear in NTP STRG queries.

Confidence: confirmed (get_project(118459), session 58; txMismatch=Yes confirmed)

18. txMismatch Flag Sensitivity — Cross-Area Only, Within-Area Blind (session 60)

Finding: The txMismatch flag fires only when the ML-predicted TX area differs from the human-assigned TX area at the top-level (TX01, TX02…TX17) code. When both human and ML assign the same top-level area but different sub-areas (e.g., TX05.6.1 vs TX05.6.3), the flag is NOT set.

Confirmed cases (STRG TX05 orbital debris cluster):

Project Human TX ML TX Flag set?
158717 Auburn metamaterial particles TX05.6.3 Mitigation TX05.6.1 Tracking No — same top area
156380 VA Tech JCA dust clouds TX05.6.3 Mitigation TX05.6.1 Tracking No — same top area
156377 WVU space laser network TX05.6.3 Mitigation TX08.1.5 Lasers Yes — cross-area TX05→TX08

For [158717] and [156380], the ML is arguably more accurate (TX05.6.1 Tracking is closer to the actual approach than TX05.6.3 Mitigation), but the mismatch goes undetected.

Implication: The txMismatch flag undercounts real classification disagreements. A project searching specifically for TX05.6.3 mitigation concepts will retrieve [158717] and [156380] without any indication that the ML disagrees. Within-area sub-category disagreements — which may be substantively significant (tracking ≠ removal) — are invisible. Estimated additional blind spot: unknown, but any TX area with rich sub-structure (TX01.4.x, TX08.1.x, TX12.1.x) is likely to have undetected within-area mismatches.

Confidence: confirmed (get_project([158717], [156380], [156377], fields=["all"]), session 60; explicit txMismatch=Yes only for [156377])

16–17. ML Lunar-Application Framing Confusion — TX03/TX12 Classified as TX07 (session 59)

Pattern: When a project's description is framed around lunar surface application contexts (permanently shadowed regions, lunar dust, resource access), the ML classifier fires TX07 (destination resource exploration/processing/contamination) regardless of the actual underlying technology. This is a consistent ML failure mode for the lunar surface application domain.

TX03 power projects (PSR framing → TX07):

Project Title Human TX ML TX mismatch
183685 Oxychalcogenide Superconducting Transmission (U. Chicago) TX03.3.2 Distribution TX07.1.1 Resource Exploration Yes
183711 Adsorbent Fuel Storage for SOFCs (CSM) TX03.2.2 Fuel Cells TX07.1.3 Resource Processing Yes

TX12 materials projects (lunar dust framing → TX07.2.5):

Project Title Human TX ML TX mismatch
182203 Crumpled Nano-ball Dust Coatings (UC Irvine) TX12 Materials TX07.2.5 Particulate Contamination Yes
182204 Hierarchical Dust-Mitigating Nanostructures TX12 Materials TX07.2.5 Particulate Contamination Yes
182220 Anisotropic Structured Surface (UCF) TX12 Materials TX07.2.5 Particulate Contamination Yes
182221 Bioinspired Dust-Repelling Textures (U Arkansas) TX12 Materials TX07.2.5 Particulate Contamination Yes

Why it happens: The ML has learned that "lunar dust" → TX07.2.5 (Particulate Contamination) and "PSR / lunar south pole" → TX07.1.x (Resource Exploration). Human classifiers correctly identify the core technology (materials science = TX12; power infrastructure = TX03). ML focuses on application context; humans focus on technology class.

Note on usual mismatch direction: In most cases documented here, ML is more accurate than humans. This is one of the few patterns where humans consistently outperform ML. It occurs because the application-context framing dominates descriptions of lunar surface projects, overriding the technology signal.

Impact: Any search for TX03 PSR power or TX12 dust-mitigation projects that relies solely on human TX classification should also check txMismatch=true results in TX07, as 6 confirmed correct-human-classification projects are ML-mislabeled there. The mismatch flag IS set for all 6 — techport_find_projects with txMismatch=true and technology_area=TX07 would surface these.

Confidence: confirmed (get_project(183685, 183711) live; find_projects(STRG, TX12, Active) batch → txMismatch=Yes for all 4 dust projects; session 59)

19. Cornell PREHEAT — Propulsion Project Classified as ISRU Consumables (session 61)

Project: 158554 "PREHEAT: Propellant Regenerative Energy for High-power Electric Adiabatic Thrusters" — Cornell University, STRG, TRL 2→3, 2024-2028.

Classification error: Human assigned TX07.1.3 Resource Processing for Mission Consumables. ML predicted TX01.2.2 Electrostatic Propulsion — ML is significantly closer to correct.

Why it's wrong: The project explicitly develops NEP (nuclear electric propulsion) and high-power SEP (solar electric propulsion) thruster technology. The description discusses "nuclear electric propulsion and high-power electric propulsion using large solar arrays" as the core technology gap being addressed. PM Lee K. Johnson (JPL) — a propulsion group contact, not an ISRU specialist. The ISRU connection is that PREHEAT proposes using ISRU-derived propellant as thruster input, but the technology being developed is the propulsion system, not the propellant production process.

Impact: PREHEAT appears in TX07.1.3 ISRU searches and is invisible in TX01 propulsion portfolio queries — inflating ISRU counts and deflating propulsion counts by 1. Destinations listed include Moon, Mars, LEO, and "Others Inside the Solar System" — consistent with electric propulsion, not site-specific ISRU.

Likely cause: Human classifier read the propellant connection as the defining feature rather than the thruster technology being developed.

Confidence: confirmed (get_project(158554, fields=["all"]), session 61; txMismatch=Yes confirmed; ML TX01.2.2 closer than human TX07.1.3)

20. Cross-Area Mismatch Rates Vary Dramatically by TX Area (session 62)

Finding: The txMismatch flag cross-area rate is NOT uniform across the portfolio. It ranges from 4% (SBIR/STTR baseline) to 18% (TX08), with TX01 at 14%. TX areas with strong application-context boundaries have the highest mismatch rates.

Data (n=50 samples per TX area):

TX Area Cross-Area Mismatch (txMismatch=Yes) Rate Primary boundary confusion
SBIR/STTR (all TX, Phase 1 baseline) 2/50 4% Mixed — no single dominant boundary
TX01 Propulsion 7/50 14% TX03 (fuel cell power-propulsion), TX07 (ISRU propellant), TX09 (EDL), TX12 (catalyst manuf.), TX13 (test facilities)
TX08 Sensors/Instruments 9/50 18% TX07 (sensors for ISRU), TX06 (medical sensors), TX05 (timing/comms), TX02 (instrument electronics), TX04 (lab robots with sensors)

TX01 cross-area cases (7/50): - [184655] + [184672] SOFC systems (TX01) → TX03 (fuel cells are power AND propulsion oxidizer source) - [106739] FORGE microfluidics (TX01) → TX09 (chemical analysis classified as EDL) - [116760] Fiber Optic Sensing (TX01) → TX09 (structural health monitoring on propulsion tanks) - [156385] Blue Origin PRO-TP (TX01) → TX07 (ISRU propellant plant is also a propulsion enabler) - [184103] SPARTAN 2 priming test apparatus (TX01) → TX13 (test facility, not propulsion hardware) - [146974] LATTICE-2 Ir catalyst (TX01) → TX12 (catalyst manufacturing for propulsion)

TX08 cross-area cases (9/50): - [106632] Honey Bubble Excitation (TX08) → TX07 (molten regolith electrolysis sensor → classified as ISRU) - [145003] STRATO tactical radio (TX08) → TX05 (comms radio ≠ sensor) - [184148] HERMES extraction (TX08) → TX06 (life sciences mixing → classified as medical) - [155242] Ultrasonic Soil Probe (TX08) → TX07 (soil sensing → classified as resource exploration) - [158680] Microgravity Lab Assistant (TX08) → TX04 (lab robot with sensors → classified as robotics) - [106606] SNAP wildfire satellite (TX08) → TX06 (fire detection sensor → classified as fire suppression) - [157668] Ultra Low Noise Clock (TX08) → TX05 (precision clock → classified as timekeeping/comms) - [105782] Housekeeping Electronics miniaturization (TX08) → TX02 (instrument electronics → classified as avionics) - (1 additional mismatch in the FO/GSFC IRAD sample)

The structural pattern — "sensor for X" → "X": The dominant failure mode for TX08 is that the ML classifies on the application of the sensor rather than the sensor technology itself. A sensor used for ISRU prospecting becomes TX07; a sensor used in a medical context becomes TX06; a timing sensor becomes TX05. The taxonomy places "sensors" in TX08 but every other TX area also uses sensors, so the boundary is inherently ambiguous.

For TX01, the pattern is "technology enabling propulsion" → "the thing it enables" (or vice versa): ISRU propellant → TX07; EDL propulsion → TX09; power for propulsion → TX03.

Implication for portfolio search: When searching TX08 or TX01, adding txMismatch=true and checking TX07, TX06, TX05, TX02 (for TX08) or TX03, TX07, TX09 (for TX01) will recover a non-trivial fraction of relevant projects. For TX08 specifically, a 18% cross-area leak is large enough to affect technology landscape analysis.

Update to overall blind spot estimate: Phase 1 estimated 4% mismatch + 4% missing = 8% blind spot for SBIR/STTR. For TX08 and TX01, the mismatch component alone is 14-18%, making the effective blind spot 18-22% before accounting for missing TX.

Confidence: suggestive (n=50 per TX area; samples not stratified by program or era; sample ordering may favor more-recently-updated records which skew toward specific programs. Rates should be treated as order-of-magnitude estimates.)

Queries used: find_projects(technology_area="TX01", status=null, fields=[...], limit=50) and same for TX08. Session 62.

21. Argonne MCP Detector Classified as TX04 (Robotics) — APRA Program (session 74)

Project: 157554 "Low Thermal Coefficient of Resistance Microchannel Plates" — Argonne National Laboratory, APRA, TRL 2→4, 2023-2026.

Technology: Microchannel plate (MCP) electron amplifiers for UV photodetectors used in space astronomy missions. MCPs are detector components (TX08.1.1 Detectors and Focal Planes — the correct classification).

Classification error: Human assigned TX04 (Robotics, Tele-Robotics, and Autonomous Systems). This is a detector development grant for space astronomy UV instruments — no robotics content whatsoever.

Discovery context: Found while doing portfolio_aggregate(group_by="primaryTx", filter={"program":"APRA","status":"Active"}) — out of 30 active APRA projects, 28 = TX08, 1 = TX11 (XSTAR software), 1 = TX04 (this project). TX04 in APRA immediately flagged as anomalous given APRA's astrophysics instrumentation mandate.

Impact: This project will not appear in any TX08 search for APRA detectors. For the active APRA cohort, the real TX08 count is 29/30 (97%), not 28/30 (93%) as the aggregate reports.

Confidence: confirmed (direct record inspection; project description unambiguously describes detector technology; txMismatch flag not checked but classification is clearly wrong regardless of ML prediction)

Project: 157590 "Ultrasensitive Far-IR KID Arrays for Space" — Caltech, SAT, TRL 4→6, Completed Sep 2025.

The anomaly: TechPort records [157590] as "Advanced From" → 96368 "Toward Fast, Low-Noise, Radiation-Tolerant X-ray Imaging Arrays for Lynx" (MIT SAT, 2020-2022, X-ray CCD). Partner type: "Other."

Why it's wrong: Far-IR KID arrays (25–212 µm, KI detectors, cryogenic focal planes) do not advance from soft X-ray CCD imagers (0.2–10 keV silicon sensors). These are entirely different detector physics, wavelength regimes, materials, and readout approaches. Bradford/JPL's KID work has continuous lineage in ground/balloon KID instruments and early-stage APRA grants — not from Lynx X-ray optics.

Likely mechanism: Same "shared institutional partner" pattern as Issue 12 (PuFF/ChargerSat-2). Mario Perez is PM on both [157590] and multiple SAT X-ray projects — the link may reflect a program management connection rather than genuine technology lineage.

Impact: Confounds the lineage graph for the far-IR probe technology track. Any automated lineage tracing from [157590] back via "Advanced From" would land on the X-ray CCD program, not the correct APRA KID precursors.

Confidence: confirmed (direct record inspection; physics/technology incompatibility is unambiguous; get_project(157590, fields=["all"]), session 76)

23. THz SIS Receiver [157544] Classified as Water Recovery (TX06) by ML (session 76)

Project: 157544 "Terahertz Correlation Receiver for the Detection of Water in Planet Forming Disks" — JPL, APRA, TRL unspecified, 2023-2027.

Classification: Human assigned TX08.1 (Remote Sensing Instruments). ML predicts TX06.1.2 (Water Recovery and Management). txMismatch: True.

Why human is correct: This project develops NbTiN SIS heterodyne mixer receivers above 700 GHz for space astrophysics observations of ortho/para-H₂O and HDO in protoplanetary disks. The science application involves water molecules but the technology is a superconducting quantum detector (SIS = Superconductor-Insulator-Superconductor). TX08 (Sensors and Instruments) is correct; TX06 (Human Health and Life Support) is wrong by a large margin.

Why ML is fooled: "Detection of water" in the project title. The ML has learned that "water" + "detection" often corresponds to TX06 life support (e.g., water quality sensors, water recovery monitors for ECLSS). A THz astrophysics receiver has nothing to do with life support, but the title surface pattern matches.

Pattern context: This is a specific variant of the Issue 20 "sensor for X → X" failure mode. The ML fires on the science target (water) rather than the technology (THz SIS receiver). APRA's mission is astrophysics instrumentation, not life support, making this a high-confidence misclassification.

Confidence: confirmed (direct record inspection; get_project(157544), session 76; txMismatch=Yes confirmed)

24. LCRT NIAC Phase II [106036] ML Predicts Aeroacoustics (TX15) (session 77)

Project: 106036 "Lunar Crater Radio Telescope (LCRT) on the Far-Side of the Moon" — JPL, NIAC Phase II, TRL 2→3, 2021-2023.

Classification: Human assigned TX08.2 (Observatories). ML predicts TX15.1.4 (Aeroacoustics). txMismatch: True. granularity gap: ML provides sub-area.

Why human is correct: LCRT is an RF telescope concept — a 350m wire-mesh reflector deployed inside a 1.3km diameter lunar crater for Cosmic Dark Ages cosmology at 4.7-47 MHz. It is unambiguously an observatory (TX08.2). The human TX08 lacks granularity (no sub-code); TX08.2.2 (Structures and Antennas) would be more precise.

Why ML is fooled: Likely on structural/deployment mechanics language ("tension lift wires," "wire mesh," "crater geometry," vibration testing). TX15 (Flight Vehicle Systems) includes aeroacoustics (TX15.1.4) — structural/acoustic vocabulary in long NIAC Phase II descriptions triggers the mismatch. NIAC Phase II descriptions are longer and technically denser than most programs, providing more opportunity for domain-vocabulary confusion.

Pattern context: Long, technically dense NIAC descriptions expose the ML to more domain-specific vocabulary than metadata-thin SBIR entries. This is a structural vulnerability: the more detailed the description, the more misleading keyword patterns emerge. NIAC has a higher TX mismatch opportunity rate for this reason (Phase II = multi-year reports with engineering details).

Confidence: confirmed (direct record inspection; get_project(106036), session 77; txMismatch=Yes confirmed)

25. "Project Tethys" (WPI) ISRU Project Classified as Electrostatic Propulsion (session 78)

Project: 158665 "Project Tethys: Extracting Water from the Martian Environment" — Worcester Polytechnic Institute (WPI), STRG, TRL 2→3, 2024-2028.

Classification: Human assigned TX01.2.2 (Electrostatic Propulsion). ML predicts TX07.1.3 (ISRU Extractive Tech). txMismatch: True.

Why ML is correct: "Project Tethys: Extracting Water from the Martian Environment" is unambiguously an ISRU project — it extracts water from Martian regolith or atmosphere. The technology is extraction/processing, not propulsion. TX07.1.3 (Acquisition and Processing of Resources) is correct.

Why human is wrong: The proposed application of extracted water is likely as rocket propellant (hence the TX01 assignment). The propellant application of water does not make the extraction technology TX01 propulsion. This is a consistent failure mode: classifying by downstream application (propellant production) rather than the technology being developed (resource extraction).

Confidence: confirmed (find_projects(program="STRG", technology_area="TX01", status="Active"), session 78; txMismatch=Yes confirmed)

26. "Beamed Microwave Energy Propulsion" (PSU) Classified as Nuclear Thermal Propulsion (session 78)

Project: 118459 "Beamed Microwave Energy Propulsion Leveraging Lunar Resources and Dual-Use Infrastructure Systems" — Pennsylvania State University, STRG, TRL 2→3, 2022-2026.

Classification: Human assigned TX01.4.3 (Nuclear Thermal Propulsion). ML predicts TX07.1.3 (ISRU). txMismatch: True.

Why both are wrong: "Beamed Microwave Energy Propulsion" involves beaming microwave power to a spacecraft from lunar infrastructure to drive propulsion — neither nuclear nor ISRU extraction. The correct TX is TX01.4.2 (Solar/Laser Thermal or Beamed Energy Propulsion) or TX01.4.6 (Advanced Energetic Propulsion).

Why human picked TX01.4.3: Artemis context and "lunar resources" language may have triggered nuclear/lunar connection in the classifier's mind. Why ML picked TX07.1.3: "leveraging lunar resources" and ISRU vocabulary in the title. Both are fooled by contextual framing.

Pattern context: Beamed energy propulsion (laser, microwave) falls in a taxonomy gap — TX01.4.x lacks a clear bin for non-nuclear advanced propulsion, causing recurring misclassification. This is the third known beamed/non-nuclear propulsion misclassification.

Confidence: confirmed (find_projects(program="STRG", technology_area="TX01", status="Active"), session 78; txMismatch=Yes confirmed)

27. Purdue RDRE Geometry Optimization Classified as Nuclear Thermal Propulsion (session 78)

Project: 156358 "Optimization of Device Geometry for Rotating Detonation Engines" — Purdue University, STRG, TRL 2→3, 2023-2027.

Classification: Human assigned TX01.4.3 (Nuclear Thermal Propulsion). ML predicts TX01.3.4 (Airbreathing Pressure Gain Combustion = RDRE). No explicit txMismatch flag in the query results (possible threshold not triggered).

Why human is wrong: This is explicitly an RDRE (Rotating Detonation Engine) experimental project — a variable-geometry modular RDRE is being built. TX01.3.4 is exactly correct. TX01.4.3 (NTP) is for nuclear reactor-powered propulsion — entirely different technology and physics.

Why ML is correct: "Rotating detonation" = pressure gain combustion = TX01.3.4, consistent with the other 3-4 STRG RDRE projects.

Why human picked TX01.4.3: Unknown — possibly data entry error, possibly confusion with TX01.3.4 sub-area codes. This is the clearest misclassification in the TX01 STRG cohort.

Confidence: confirmed (find_projects(program="STRG", technology_area="TX01", status="Active"), session 78; ML predicts TX01.3.4)

28. JANUS (GA Tech) Incorrectly Classified as Predominately Black Institution (session 79)

Project: 118384 "Joint Advanced Propulsion Institute (JANUS)" — Georgia Institute of Technology-Main Campus, STRG, Active, 2021-2026.

Classification: leadOrgMsiCategory = ['Predominately Black Institutions (PBI)']

Why wrong: Georgia Institute of Technology is not a PBI. PBIs are institutions where ≥40% of enrollment is Black/African-American. GA Tech is a flagship state research university with ~6% Black enrollment — far below PBI threshold. This is a clear data entry error.

Impact: GA Tech appears in any MSI/PBI portfolio query, inflating PBI-associated project counts. Since JANUS is a high-profile $5M+ consortium, this error would materially distort PBI program statistics.

Confidence: confirmed (get_project([118384]), session 79; leadOrgMsiCategory: ['Predominately Black Institutions (PBI)'] returned explicitly)

Issue 30: UT Austin AR&D [118518] filed as Optimetrics (TX05.1.6) — should be RPOD Navigation (TX04.5.2)

Program: STRG | Project: 118518 | Session: 80

Filed TX: TX05.1.6 (Optimetrics)
ML predicted TX: TX04.5.2 (Autonomous Rendezvous, Proximity Operations, and Docking)
txMismatch: Yes

Why wrong: "Autonomous Rendezvous and Docking with the Aid of Optical Maneuver Detection" develops algorithms for detecting relative spacecraft motion using optical sensing, enabling both-maneuvering RPOD (rather than the traditional non-maneuvering target + maneuvering chaser approach). The technology is RPOD navigation (TX04.5.2). The word "optical" caused a human to file it under TX05 (Comms/Nav optical systems), but this is proximity operations, not communications.

Impact: Appears in TX05 queries; absent from TX04 RPOD queries. Inflates TX05.6.x count artificially.

Confidence: confirmed (session 80; txMismatch field returned True; description clearly describes RPOD)

Issue 31: ESM [13633] — ARC flagged as AANAPISI (federal NASA center cannot be MSI)

Program: GCD | Project: 13633 | Session: 81

Field: leadOrgMsiCategory = ['Asian American Native American Pacific Islander (AANAPISI)']
Lead Org: Ames Research Center (ARC) — federal NASA center

Why wrong: MSI (Minority Serving Institution) categories (HBCU, HSI, PBI, AANAPISI, AANAPISI, TCU, ANNH) are classifications for educational institutions. NASA Centers are federal agencies operated by a federal contractor (ARC = operated by NASA directly). ARC cannot qualify as an AANAPISI. Same type of error as Issue 28 (GA Tech as PBI). This appears to be a data entry error where a partner institution's MSI status was incorrectly applied to the lead NASA center.

Impact: ARC/ESM appears in MSI portfolio queries under AANAPISI, inflating count of AANAPISI-affiliated projects. ESM [13633] is one of the highest-viewed active GCD projects (7,647 views); any MSI report that includes it would significantly distort statistics.

Confidence: confirmed (batch get_project([13633]), session 81; leadOrgMsiCategory: ['Asian American Native American Pacific Islander (AANAPISI)'] returned)

Issue 33: SBIR Phase III and Phase II-E projects systematically lack outcome records

Program: SBIR/STTR | Session: 89

Pattern: SBIR Phase III contracts and Phase II-E extensions do not reliably generate TechPort outcome records — not even Closed_Out. Standard Phase II completions almost always receive at least a Closed_Out; Phase III and Phase II-E completions frequently have zero outcomes.

Evidence (TX08 TRL 9 cohort, session 89): 10 TX08 SBIR projects completed at TRL 9. Three have 0 outcomes: - 102733 Vista Photonics — explicitly labeled "SBIR Phase III" in project description. TRL 5→9. 0 outcomes. (This is the highest TRL in the cohort with no tracking at all.) - 102379 Black Swift Technologies — Phase II-E per description ("adapt existing fixed wing aircraft for VTOL"). TRL 7→9. 0 outcomes. - 154431 Space Environment Technologies (ARMAS-DM) — Phase II-E per description ("funding applied toward the 'dual' aspect"). TRL 6→9. 0 outcomes.

The remaining 7 TRL-9 projects (standard Phase II with commercial/transition outcomes or CCRPP designation) all have ≥1 outcome record.

Root cause: Phase III SBIR is by definition the commercialization phase — companies take NASA-funded technology to market using non-NASA funding (or any federal agency's funding). A Phase III project may have no separate SBIR award number, or may be funded through a procurement contract rather than a SBIR grant. TechPort's outcome model is designed around SBIR Phase I→II→Transition; Phase III bypasses this funnel. Phase II-E is an amendment to the Phase II award, not a new project — the extension work may not trigger a separate outcome entry.

Impact: The highest-TRL SBIR projects (those that have reached commercialization) are the least tracked. A portfolio researcher counting SBIR TX08 commercialization successes from outcome records will undercount by at least 30% in the TRL-9 cohort. The true commercialization rate is higher than TechPort's outcome fields suggest.

Confidence: confirmed (get_project([102733]) live API, description text contains "SBIR Phase III"; 0 technologyOutcomes; same pattern in [102379], [154431]; 7 of 10 TRL-9 projects in the same cohort have outcomes; session 89)

Issue 32: Made in Space AMF [12902] filed as TX08.1.1 (Detectors) — should be TX12.2 (Manufacturing)

Program: SBIR/STTR | Project: 12902 | Session: 85

Filed TX: TX08.1.1 (Detectors and Focal Planes)
Correct TX: TX12.2 (Manufacturing/Materials Processing) — arguably TX12.2.2 (Additive Manufacturing)

Why wrong: "ISS Additive Manufacturing Facility" (AMF, Made in Space Inc., PI Michael Snyder) is an on-demand 3D printer that flew to ISS. It has nothing to do with detectors or focal planes. The SBIR contract (O3.02-9753) targeted manufacturing capability — the briefing chart describes Phase II objectives as CDR → ETU build → ETU qualification → flight-ready ETU. This is a confirmed SBIR→ISS hardware pathway.

Impact: This is one of the clearest SBIR-to-flight success stories in TechPort, but it is invisible to any search for SBIR manufacturing successes because it is classified under TX08 (sensors). A researcher querying TX12 SBIR TRL 8 completions would miss it entirely. The outcome field says "Advanced From" (not Infused_To or Transitioned_To), compounding the invisibility: two independent classification failures cause the AMF to disappear from every relevant portfolio query.

Confidence: confirmed (document fileId 356086 read session 85; briefing chart explicitly describes additive manufacturing hardware for ISS)

Issue 29: MIT SuperLimbs [156352] filed as Pressure Garment (TX06.2.1) — should be Wearable Robotics (TX04.4.1)

Program: STRG | Project: 156352 | Session: 80

Filed TX: TX06.2.1 (Pressure Garment)
ML predicted TX: TX04.4.1 (Robotic Systems, Wearable)
txMismatch: Yes

Why wrong: "Design, Control, and Human-Robot Coordination of Space Suits Integrated with Supernumerary Robotic Limbs" is about wearable robotic extra limbs that attach to astronaut bodies and coordinate with spacesuit motion. The technology being developed is a robotic control system — TX04 (Robotics and Autonomous Systems). The spacesuit integration context caused the human classifier to file it under TX06 (Life Support and Habitation). ML correctly identifies the technology as wearable robotics.

Impact: Appears in pressure garment portfolio queries; absent from robotics (TX04) queries. Minor distortion — STRG has only 2 TX06.2.1 projects; [156352] is one of them, so removes 50% of that sub-area if corrected.

Confidence: confirmed (get_project([156352]), session 80; txMismatch field returned True)

Issue 34: SMD Research Grant Programs Have Near-Zero Outcome Tracking (sessions 90-93)

Programs confirmed: H-TIDeS, PICASSO, APRA, MatISSE, DALI, COLDTech, HOTTech, SAT, EXEP, PCOS — all SMD grant programs.

Complete outcome audit across 10 programs (sessions 91-93):

Program Total Projects Completed Infused_To Transitioned_To Closed_Out Total Outcomes
H-TIDeS 128 121 1 0 0 1 (0.8%)
PICASSO 124 103 0 0 0 0 (0%)
APRA 246 216 0 0 0 0 (0%)
MatISSE 33 33 0 0 — 0 (0%)
DALI 39 28 0 — — 0 (0%)
COLDTech 32 32 0 — — 0 (0%)
HOTTech 19 19 0 — — 0 (0%)
SAT 91 81 0 — — 0 (0%)
EXEP 2 2 0 — — 0 (0%)
PCOS 4 4 0 0 — 0 (0%)
TOTAL 718 639 1 0 0 1 (0.16%)

One outcome record across 639 completed SMD grant projects. The single record (CODEX [95780] Infused_To from H-TIDeS) is almost certainly an anomaly — CODEX was an exceptionally high-profile ISS instrument and may have been manually entered by a diligent project manager.

For comparison: SBIR/STTR has Closed_Out on ~80%+ of completed projects and Infused_To/Transitioned_To on ~5-10%. NIAC has outcomes on ~95%+ of completed projects.

Root cause: SMD grants are PI grants, not procurement contracts. Contractors (SBIR companies, mission teams) report outcomes because they have contractual requirements. ROSES PIs submit publications and follow-on proposals — these are the "outcomes" but they're not entered in TechPort. TechPort's outcome model (Closed_Out, Infused_To, Transitioned_To) maps to the SBIR/STTR commercial pathway and simply doesn't apply to academic instrument development grants.

The PICASSO→MatISSE pipeline: Montbach is PD for both programs and they're explicitly connected (PICASSO feeds MatISSE). Yet no PICASSO project has an "Advanced_To MatISSE" or "Infused_To MatISSE" link in TechPort. Cross-program instrument continuity (LIfE: COLDTech→MatISSE; SESAR: MatISSE→DALI) is traceable only by reading project descriptions, not outcome records.

A specific anomaly — H-TIDeS flight missions with 0 outcomes: Multiple H-TIDeS LCAS-track projects flew to orbit (CURIE [92216], MinXSS-2 [92211], LLITED [92224], C-REX-2 [92221]) and returned direct science data — actual spaceflight missions. Only CODEX [95780] has an outcome record. Four confirmed spaceflight missions from a single program have zero TechPort outcome documentation.

Impact: All SMD research grant programs look like dead-ends in TechPort's outcome graph. A researcher using outcome records to trace planetary science or astrophysics technology maturation will find essentially no signal. The real pipelines (PICASSO → MatISSE → mission; H-TIDeS → HPD mission proposal; COLDTech → MatISSE → ocean worlds instruments) are completely invisible in TechPort's outcome data model.

Note on EXEP: Unlike ROSES grant programs, EXEP TRL records DO advance (NEID: TRL 4→8; Roman CGI: TRL 6 confirmed). EXEP is not a standard ROSES solicitation — it's a NASA program-level funding mechanism for exoplanet technology. Issue 34 (zero outcomes) still applies; Issue 35 (frozen TRL) does NOT apply to EXEP.

Pioneers Program audit (session 94): find_projects(program="Pioneers Program", outcome_path="Infused_To/Transitioned_To") → 0 each. However, Pioneers has only 1 TechPort project total (The Landolt Mission [157580], Active, TRL 7→9). No outcomes possible for an active project. Pioneers is essentially absent from TechPort, likely because it funds science investigations (CubeSats, SmallSats, major balloon missions), not technology development. The program description explicitly states: "primary review criterion for selection is the merit of the proposed science investigation." This makes Pioneers a category-boundary case — borderline whether it belongs in TechPort at all. Issue 34 technically applies (0 outcomes) but the more significant finding is the near-total absence of program data (1 project for a multi-year, multi-award APD program).

Note on HFORT, HLCAS: Not yet audited but expected to show the same outcome pattern.

Queries used (sessions 91-94 extended audit): - find_projects(program="APRA/MatISSE/DALI/COLDTech/HOTTech", outcome_path="Infused_To") → 0 each - find_projects(program="SAT", outcome_path="Infused_To") → 0 (session 92) - find_projects(program="EXEP/PCOS", outcome_path="Infused_To/Transitioned_To") → 0 each (session 93) - find_projects(program="Pioneers Program", outcome_path="Infused_To/Transitioned_To") → 0 each (session 94, 1 total project) - Data snapshot: 2026-04-04

Confidence: confirmed for all 10 programs listed above.

Issue 35: TRL Current Field Frozen at Starting TRL for SMD ROSES Grant Programs (session 92)

Summary: The trlCurrent field is not updated for completed SMD ROSES grant projects. Projects show trlCurrent = trlBegin regardless of actual technical progress. This extends Issue 34 (outcome records) — both outcome documentation AND TRL advancement recording are effectively absent for ROSES grant programs.

Evidence — PSD instrument program TRL achievement distribution (completed projects):

Program Mandate Completed trlCurrent=6 trlCurrent=trlBegin Source
DALI TRL 4→6 28 0 (0%) ~32% show TRL 4 = trlBegin aggregate(trlCurrent, DALI, Completed)
MatISSE TRL 4→6 33 0 (0%) ~48% show TRL 4 aggregate(trlCurrent, MatISSE, Completed)
COLDTech TRL 2→4 32 0 (0%) ~41% show null aggregate(trlCurrent, COLDTech, Completed)
HOTTech TRL 2→6 19 1 (5.3%) 3 key projects at trlBegin aggregate(trlCurrent, HOTTech, Completed)

All four programs had a mandate to develop instruments to TRL 4-6. TechPort records show near-zero TRL advancement.

Specific frozen TRL cases (HOTTech — where TRL 6 was expected from context):

Project trlBegin trlCurrent trlEnd Expected outcome
HOTLINE Power 92299 4 4 6 "TRL 6" stated in KB session 91
InnoSys UHF 116070 3 3 6 "TRL 6" stated in KB session 91
Honeybee Venus Actuator 116074 4 4 6 "TRL 6" stated in KB session 91
GRC SiC Memory 92917 3 3 6 "TRL 6" stated in HOTTech page

DALI geochronology case (mission coupling adds complexity): - CDEX 96939 (SwRI, trlBegin=4, trlCurrent=4): Quad chart says "TRL (4) to (5-6)" and explicitly shows CDEX on an Astrobotic lander with text "We plan to mature CDEX for flight to the moon in 2023-2024." Peregrine Mission One (Astrobotic) failed January 2024. CDEX's TRL stall is partially a mission failure story — the intended TRL 6 demonstration platform was destroyed.

Root cause (analogous to Issue 34): ROSES PI grants have no contractual TRL reporting requirement. For contractors (SBIR, GCD), TRL updates are part of milestone reporting. For ROSES PIs, TRL advancement is documented in journal papers and mission proposals — not TechPort records. TechPort's trlCurrent reflects what the PI entered at project creation and is rarely updated.

Why the HOTTech page contained errors: Session 91 analysis relied on project descriptions suggesting TRL 6 milestones (quad chart text, technical accomplishment statements) without verifying against trlCurrent. The descriptions are likely accurate about what was attempted — the TechPort trlCurrent field simply doesn't record what was achieved.

Corrected HOTTech count: 1 TRL-6 project (Makel Engineering SiC chemical sensor [92279], Gen 1, last updated 2020-01-14). This predates most DALI/HOTTech projects and may have been manually updated by a diligent project manager — the same pattern as the single H-TIDeS outcome record (CODEX, Issue 34).

Quad chart evidence for the frozen-TRL mechanism (session 93 document reads):

Project Quad Chart States TechPort trlCurrent
InnoSys UHF Transmitter 116070 "Starting TRL: 3, Ending TRL: 6" 3 (frozen at start)
Honeybee Venus Actuator 116074 "Starting TRL: 4, Ending TRL: 6" 4 (frozen at start)

The quad chart is the project's work plan — it states the intended TRL endpoint. TechPort shows the unupdated starting value. This confirms the mechanism: TRL was planned (and likely attempted) but TechPort field was never updated. File IDs: 315655 (InnoSys), 315713 (Honeybee).

Extension to PCOS (session 93): The same TRL underreporting pattern appears in PCOS — NASA's ESA mission contribution program. All 4 projects show trlCurrent=4 (target was 5-6):

Project trlBegin trlCurrent trlEnd Notes
LISA Laser 183289 3 4 6 Partial advance (3→4)
LISA Telescopes 183290 3 4 6 Partial advance (3→4)
UV LED CMS 183291 4 4 6 Frozen at start
Athena X-IFU 183297 4 4 5 Frozen at start

PCOS is NOT a ROSES solicitation — it's a program-level funding mechanism for ESA mission contributions. The LISA Laser and Telescope did advance one TRL step (3→4), which is different from completely frozen ROSES grants. But none reached their target TRL. Issue 35 applies in a milder form to PCOS.

Impact on KB pages: HOTTech page "TRL 6 projects | 4 (21%)" claim has been corrected to "1 (5.3%)." PCOS page corrected from "TRL 6 (LISA) / TRL 5 (Athena)" to "trlCurrent=4 for all." DALI page includes TRL Achievement Gap section (0/28 result).

Impact on data use: When investigating SMD grant programs, do NOT rely on trlCurrent to assess technology maturation. Read project descriptions and quad chart documents to understand what milestones were attempted. Only SBIR/STTR and contracted programs (GCD, FO) have reliable TRL record updates. EXEP is an exception — TRL records DO advance for EXEP projects.

Queries used: - portfolio_aggregate(group_by="trlCurrent", filter={"program":"DALI","status":"Completed"}) → TRL5:11, TRL4:9, none:8 (no TRL6) - portfolio_aggregate(group_by="trlCurrent", filter={"program":"MatISSE","status":"Completed"}) → TRL4:16, none:13, TRL3:3, TRL5:1 (no TRL6) - portfolio_aggregate(group_by="trlCurrent", filter={"program":"COLDTech","status":"Completed"}) → none:13, TRL4:9, TRL3:6, TRL2:4 (no TRL6) - portfolio_aggregate(group_by="trlCurrent", filter={"program":"HOTTech","status":"Completed"}) → TRL3:8, TRL2:5, none:3, TRL4:2, TRL6:1 - find_projects(program="HOTTech", trl_min=6, status=null) → 1 project: Makel [92279] - get_project([183289,183290,183291,183297]) → all trlCurrent=4 - Data snapshot: 2026-04-04

Confidence: confirmed for all program TRL distributions. Frozen-TRL mechanism is confirmed for 6 specific cases (HOTTech ×4, PCOS ×2 complete freeze + ×2 partial) and confirmed via quad charts for InnoSys and Honeybee. Pattern is suggestive for DALI/MatISSE/COLDTech (consistent with pattern, not individually verified per project).

Open Threads

  • Library item coverage: Now covered in topics/document-availability.md. Summary: NIAC ~100%, SBIR/STTR ~35%, FO ~15% (cohort effect), GCD ~5% (batch records), STRG ~90% count but mostly boilerplate for old cohort.
  • Description quality by program: STRG descriptions not sampled yet. CIF descriptions not sampled.
  • Duplicate entries: "10 meter Sub-Orbital Large Balloon Reflector" appears twice in NIAC (IDs 13744 and 14325). This may be Phase I + Phase II duplication — normal and expected for NIAC. Verify count implications.
  • Advanced_From count: Not queried. Would reveal the SBIR Phase I→II transition rate, a useful quality metric.
  • Spurious "Other" partner-type Advanced_To links: ChargerSat-2/PuFF is the confirmed example (Issue 12). Check other NIAC projects for this pattern.