InnoDexis LLC

InnoDexis is a worldwide innovations AI platform that solves the problem of fragmented innovation ecosystems by integrating data from startups & universities.

[InnoDexis LLC @innodexis]

A $7 LED attachment and a smartphone detect hidden cameras in 5 seconds

Existing hidden camera detectors cannot reliably separate lens reflections from ordinary glossy surfaces.

Researchers from KAIST, Singapore Management University, and NUS developed SweepLED - an AI-powered system that turns a smartphone into a hidden camera detector using a low-cost LED attachment.

SweepLED analyses how reflections move and deform across illumination angles - a pattern unique to camera lens structures.

🔬 Key signals:
• 94% detection accuracy across 30 real-world evaluated objects
• Inspection time under 5 seconds per object
• Hardware cost under $7 (KRW 10,000) for the LED smartphone attachment
• Distinguishes camera lenses from ordinary glossy reflections using deep learning

Why this matters:
SweepLED automates the reflection analysis entirely - removing the need for trained observation or dedicated detection hardware.

The deep learning layer distinguishes genuine lens signatures from ordinary surfaces - reducing false positives conventional tools cannot filter.

94% accuracy at under 5 seconds positions this as a deployable consumer tool - not a prototype.

What's changing:
From manual visual inspection of reflective spots → to AI-automated dynamic reflection analysis accessible via smartphone

Detection becomes a routine step - requiring no prior training and no dedicated equipment.

The privacy threat from concealed recording devices has outpaced the accessibility of detection tools.

A $7 attachment and a downloadable app could shift hidden camera detection to a routine consumer behaviour.

If a $7 accessory delivers 94% detection accuracy, what does that mean for privacy standards in short-term rentals and shared spaces globally?

#PrivacyProtection #AIDetection #MobileSecurity #HiddenCamera #ConsumerAI #KAIST #InnoDexis

[InnoDexis LLC @innodexis]

What separates a product that ships from a press release that says it did

A third of global product launches contain nothing a diligence process can verify.

InnoDexis validated 1,647 corporate product launches and 537 research prototypes across 14 product categories in August 2026 - 34.1% of all validated corporate records.

Substantiation-field counting separates the signal from the noise - and it has been validated twice.

🔬 Key signals:
• Launches with 4+ substantiation fields are 5.7× more likely to be rated High investment attractiveness than those with zero
• 521 of 1,647 launches carry zero substantiation fields - a third of product-launch communication contains nothing verifiable
• API & infrastructure scores 6.29; services scores 4.05 - what is launched predicts quality more strongly than who launches it
• Only 790 of 1,647 launches assert the product can be obtained now - launch counts overstate market presence by roughly half

Why this matters:
High Bandwidth Flash, TRACE, and TESTEX Circularity each set design constraints for products that do not yet exist.

Three standards with tier-one backing - none would surface in a volume-ranked feed.

537 research prototypes name a category; only 90 name a product - the funnel narrows six to one.

What's changing:
From launch tracking built on record counts and volume → to substantiation-field counting separating signal from noise

The filter is already in the schema - it requires no new data collection.

China, Japan, and South Korea score 5.55–5.60 against the US at 4.92 - a direct consequence of any volume-weighted feed.

Standards conformance separates 0.64 points; certification just 0.09 - similar fields behave nothing alike in practice.

If a ten-field count already in the schema separates the 5.7× investment signal from the noise - what is your current product intelligence workflow counting instead?

#ProductIntelligence #InnovationIntelligence #DeepTech #LaunchTracker #AIInfrastructure #StandardsStrategy #InvestmentStrategy #InnoDexis

[InnoDexis LLC @innodexis]

The procedure did not get faster - the waiting between procedures was eliminated:

Robotic bronchoscopy, EBUS staging, marker placement, and resection already existed as separate steps.

Combining them under one anaesthetic event is a workflow decision, not a technology invention.

That is what makes this immediately replicable at centres already equipped with robotic systems.

[InnoDexis LLC @innodexis]

Early lung cancer diagnosis and surgery condensed into three hours

In early lung cancer, every week between detection and surgery costs survival.

Researchers at UCLA Health established a single-session robotic platform combining biopsy, lymph node staging, tumour localisation, and surgical resection under one anaesthetic event.

Four procedures. One anaesthetic. One operating session. No waiting between steps.

🔬 Key signals:
• Median delay to surgery: 57 days at UCLA, 70 days in Veterans Affairs data
• Survival decreases beyond 8 weeks from detection - recurrence increases beyond 12 weeks
• Single combined procedure completed in 3–4 hours under one anaesthetic event
• Real-time intraoperative pathology prevents unnecessary resection if lymph node spread is identified

Why this matters:
Multi-week delays between detection and surgery are documented contributors to worse survival and higher recurrence.

Condensing four procedures into one session eliminates waiting periods - live staging decisions protect patients from unnecessary surgery.

Intraoperative pathology ensures resection only proceeds when staging confirms no lymph node spread.

What's changing:
From sequential multi-specialist visits over weeks → to a single robotic session combining diagnosis, staging, localisation, and resection

Time lost between specialist appointments is the documented driver of worse outcomes - this removes it.

The bottleneck in early lung cancer outcomes is not surgical capability - it is time lost across fragmented care pathways.

A single-session model using existing FDA-cleared robotic infrastructure could redefine the standard of care.

If the infrastructure already exists, what is the clinical cost of lacking only the integrated workflow?

#LungCancer #RoboticSurgery #PrecisionOncology #ThoracicSurgery #ClinicalInnovation #EarlyDetection #InnoDexis

[InnoDexis LLC @innodexis]

https://t.co/vhuFW6IIKQ

[InnoDexis LLC @innodexis]

Graphene synthesis at 300°C - a third of what conventional CVD needs

Conventional graphene CVD requires temperatures up to 900°C - limiting structural control and feedstock compatibility.

Researchers from Tohoku University and Queen Mary University of London developed a low-temperature CVD method synthesising graphene at 300°C using acetylene gas over a cerium oxide catalyst.

Reaction temperature alone controls the output morphology - quantum dots, aggregated graphene, or porous graphene.

🔬 Key signals:
• Synthesis temperature reduced by two thirds - from up to 900°C to 300°C
• Temperature-tunable output: quantum dots at 300°C, aggregated graphene at 450°C, porous graphene at 600°C
• Compatible with recycled plastics, biomass, and industrial waste gases as carbon feedstocks
• Slower reaction kinetics at low temperature enables morphology control not possible at high-temperature CVD

Why this matters:
High-temperature CVD moves too fast for fine structural tuning - lower temperature is what makes precise control possible.

The same mechanism that enables structural tuning also removes the feedstock constraint - two problems solved by one variable.

Acetylene decomposition initiates at 113°C - within range of standard industrial processing conditions.

What's changing:
From high-temperature CVD at fixed morphology from virgin feedstocks → to low-temperature synthesis producing tunable structures from recycled carbon

This opens graphene manufacturing to feedstocks and facilities that high-temperature CVD excludes.

The barrier to graphene manufacturing has been energy cost and structural unpredictability - this method addresses both simultaneously.

If scalable, it could integrate graphene production directly into industrial waste stream processing.

If graphene morphology can be tuned by temperature alone and powered by waste gases, what does that mean for carbon nanomaterial economics at scale?

#Graphene #CircularEconomy #SustainableMaterials #CarbonNanomaterials #Decarbonisation #AdvancedMaterials #InnoDexis

[InnoDexis LLC @innodexis]

Standard care left the clot - enVast retrieved it.

The outcome difference between arms is not marginal - zero strokes and zero deaths versus 1.3% and 2.6% in control.

That safety gap, alongside the infarct size reduction, is what makes this trial clinically significant.

A device already in 60+ countries now has the randomized data that protocol adoption requires.

[InnoDexis LLC @innodexis]

A clot retrieval device cut infarct size 26% - zero strokes, zero deaths

Mechanical thrombectomy in STEMI has lacked high-quality randomized trial evidence - until now.

Vesalio and the Cardiocentro Ticino Institute announced positive results from the NATURE randomized controlled trial evaluating the enVast Coronary Clot Retriever - a device already used in more than 20,000 patients across 60+ countries - in STEMI patients with large thrombus burden.

The NATURE trial met its primary endpoint across 154 patients randomized 1:1 at 11 centers.

🔬 Key signals:
• 26% reduction in infarct size versus standard of care via CK-MB AUC (p=0.001)
• 25% relative reduction in infarct size on cardiac MRI at day 3
• 0% strokes and 0% deaths in enVast arm versus 1.3% and 2.6% in control
• 1.3% MACE rate in enVast arm versus 3.8% in control group

Why this matters:
Standard primary PCI alone often leaves significant residual clot and myocardial damage in high-burden STEMI cases.

The NATURE trial provides the randomized evidence base that has been missing for mechanical thrombectomy in this patient group.

The improvement holds across two independent measurement approaches - CK-MB AUC and cardiac MRI.

What's changing:
From standard primary PCI as the default → to adjunctive mechanical thrombectomy with randomized evidence of reduced infarct size

This moves mechanical thrombectomy from an optional adjunct to an evidence-backed protocol candidate.

Mechanical thrombectomy has faced adoption barriers due to limited high-quality randomized trial data - NATURE directly addresses that gap.

Success across 154 patients at 11 centers provides a replicable trial design for the broader interventional cardiology community.

If randomized evidence shows 26% infarct size reduction and zero strokes, what is the clinical cost of not adopting mechanical thrombectomy as standard?

#InterventionalCardiology #STEMI #MechanicalThrombectomy #CardiovascularMedicine #MedicalDevices #ClinicalTrials #HeartAttack #InnoDexis

[InnoDexis LLC @innodexis]

Geometry replaces electronics in this light-powered jumping soft robot

A geometric angle - not code or electronics - determines how this robot moves.

Researchers at North Carolina State University developed a light-driven soft robot made from liquid crystal elastomers that continuously jumps or crawls under infrared light - with no onboard electronics or manual resetting required.

One angle change switches the robot between crawling, jumping, and vertical leaping.

🔬 Key signals:
• 50-degree V-angle produces vertical leaping; 90-degree forward jumping; 120-degree crawling
• Eliminates onboard batteries, complex controllers, and mechanical actuators entirely
• Self-resetting mechanism stores and releases torsional energy automatically under continuous light
• Demonstrated locomotion across grass, sand, rocks, mulch, slopes, and hurdles

Why this matters:
The self-resetting mechanism is the enabling breakthrough - most soft robots require manual intervention to reset between cycles.

Light-induced contraction stores torsional energy and releases it autonomously - enabling perpetual locomotion.

That makes this design viable at scales where onboard electronics cannot be miniaturised.

What's changing:
From battery-dependent, electronically controlled soft robot actuation → to geometry-driven, light-powered locomotion with autonomous continuous resetting

A single physical parameter replaces the entire electronics stack.

The ability to tune locomotion mode through a single geometric parameter opens a new design language for passive micro-robots.

Ambient light actuation and steering control are the remaining steps toward battery-free swarm robotics in environmental monitoring.

If a single geometric angle determines locomotion mode and light provides the power, what becomes possible when steering control is added?

#SoftRobotics #LightDrivenRobotics #AutonomousRobots #MaterialsScience #SwarmRobotics #Biomimetics #InnoDexis

[InnoDexis LLC @innodexis]

This dataset does not describe the genome - it systematically breaks it:

The Human Genome Project told us what genes exist. The Human Cell Atlas showed how cells read them.

This study measures what happens when each gene is switched off - one by one, in real human immune cells.

That shift from cataloguing to causal perturbation is what makes this dataset actionable for AI.

[InnoDexis LLC @innodexis]

22 million immune cells mapped gene by gene - AI gets the rulebook

A genome-scale perturbation dataset in primary human immune cells has not existed - until now.

Researchers at Gladstone-UCSF, Stanford, and UCSF knocked out nearly 12,800 genes one by one in primary human T cells - generating 22 million high-quality cells freely available as a functional lookup table for immune gene circuit discovery.

The largest contribution to date to Biohub's Billion Cells Project - generated in primary human cells, not artificial cell lines.

🔬 Key signals:
• 12,800 genes knocked out one by one across primary human T cells from real blood donors
• 22 million high-quality cells analysed from 33.4 million total cells screened
• Captures gene circuit behaviour across both resting and infection-fighting cell states
• Performed on primary human T cells - not artificial cell lines - preserving real patient immune variation

Why this matters:
Previous genomics efforts catalogued gene expression - this study measures what happens when each gene is switched off across dynamic immune states.

Real patient immune variation is preserved - making findings directly comparable to disease states seen in clinical settings.

Released globally to support AI virtual cell model training - the dataset functions as a lookup table for the field.

What's changing:
From static observational gene cataloguing in simplified cell lines → to causal functional mapping of gene circuits at genome scale

This shifts genomics from describing what exists to measuring what each gene actually does.

The life sciences field is entering a third wave of genomics - from sequencing and cataloguing toward understanding what genetic changes actually do.

A dataset of this scale could accelerate target discovery for cancer immunotherapy and autoimmune research across laboratories globally.

If AI models trained on 22 million functionally mapped immune cells can predict therapeutic targets, how quickly could this compress discovery timelines?

#FunctionalGenomics #CancerImmunotherapy #AIinBiology #VirtualCell #Immunology #PrecisionMedicine #PerturbSeq #InnoDexis

[InnoDexis LLC @innodexis]

Waste heat drives cooling directly - no electricity or refrigerants

A solid-state cooling system just generated cold from waste heat - with no electrical input.

Researchers at KIT and the University of Tsukuba developed a heat-driven elastocaloric cooling system using two coupled nickel-titanium shape-memory films - one converting heat into mechanical work, the other into cold.

The prototype operates at 130°C and has been experimentally verified to generate cold without any electrical input.

🔬 Key signals:
• Achieved a refrigerant temperature change of nearly 13°C driven entirely by waste heat
• Operates at an external heat source temperature of 130°C with an actuator temperature of 86°C
• Eliminates the electric motor actuator - thermal energy drives mechanical actuation directly
• Solid-state design uses no gaseous or liquid refrigerants

Why this matters:
Conventional cooling relies on electrically driven compressors and chemical refrigerants - contributing to energy consumption and global warming.

Waste heat from processors, engines, or solar thermal sources drives cooling directly - no electrical intermediary.

The current unoptimised prototype already delivers a 4°C temperature difference at the component level.

What's changing:
From electrically driven compressor cooling → to heat-activated shape-memory films generating cold directly from waste thermal energy

The thermal energy already being wasted becomes the energy source - no additional input required.

Cooling is one of the fastest-growing sources of global electricity demand - this approach removes it from that demand curve.

Scaling this film architecture could open passive thermal management for datacenters, automotive powertrains, and off-grid solar.

If waste heat can drive cooling without electricity or refrigerants, what does that mean for the energy footprint of datacenters and electric vehicles?

#SolidStateCooling #WasteHeatRecovery #Elastocaloric #ShapeMemoryAlloy #Sustainability #ThermalManagement #CleanTech #InnoDexis

[InnoDexis LLC @innodexis]

In vivo CAR therapy and targeted LNP delivery combined in one platform

Current CAR-T therapies require complex, costly ex vivo cell manufacturing before a single dose reaches a patient.

CREATE Medicines and WestGene Biopharma entered a strategic R&D collaboration combining in vivo immune programming and RetroT RNA gene-writing with WestGene's targeted lipid nanoparticle delivery technology - spanning autoimmune diseases, solid tumours, and four clinical markets.

A platform architecture designed to programme immune cells inside the body - without manufacturing them outside it.

🔬 Key signals:
• More than 60 patients already dosed across CREATE's in vivo CAR clinical programs
• RetroT site-specific RNA gene integration combined with WestGene's tLNP delivery in one platform
• Dual-target CD19 x BCMA immune programming included in the pipeline
• Clinical trial execution in China accelerates translational data generation with global relevance

Why this matters:
In vivo CAR therapy programmes immune cells inside the body - eliminating the ex vivo manufacturing that makes current CAR-T logistically complex.

Combining targeted LNP delivery with RNA gene-writing addresses durable expression and precise targeting in one architecture.

Multi-geography clinical execution generates regulatory-relevant data faster than single-market programmes.

What's changing:
From ex vivo cell manufacturing → to in vivo immune programming via targeted LNPs enabling off-the-shelf cell therapy

This removes the logistical complexity that has limited CAR-T access at scale.

RNA gene-writing combined with targeted nanoparticle delivery is among the more technically integrated approaches in in vivo cell therapy.

Multi-geography clinical data generation could accelerate regulatory and commercial validation for this therapeutic class.

If in vivo CAR therapy removes ex vivo manufacturing, how quickly could this expand access to populations current CAR-T infrastructure cannot reach?

#InVivoCAR #CellTherapy #LipidNanoparticles #mRNATherapeutics #Oncology #AutoimmuneDisease #Biotechnology #InnoDexis

[InnoDexis LLC @innodexis]

The most tracked domain in this dataset is the least disruptive:

Quantum delivers 37.2% high disruption on 43 corporate records; AI delivers 14.6% on 3,079 - volume and signal concentration point in opposite directions.

The 248 cross-stream matches are the most actionable layer - an organisation in both streams is the clearest signal that research has reached the market.

Volume is not the signal - concentration is.

[InnoDexis LLC @innodexis]

The highest-volume domain carries the lowest disruption signal

19,378 innovation signals. Twenty-one domains. One pattern separates signal from volume.

InnoDexis tracked 19,378 validated innovation signals across two streams - 10,133 corporate and 9,245 research records - classified across 21 domains.

Quantum outscores AI on disruption by more than 2.5× - on a fraction of the volume.

🔬 Key signals:
•AI leads both streams by volume - 37% of corporate, 16% of research - but high-disruption scoring sits below corpus average at 14.6%
• Quantum: 39.5% high momentum, 37.2% high disruption on just 43 corporate records - sharpest signal concentration in the dataset
• Climate and sustainability: 9× research-to-corporate skew - widest gap of any major domain
• Research forward-looking language: platform adaptability into adjacent applications - more common than breakthrough framing

Why this matters:
Quantum and Nuclear/Physical Sciences account for under 4% of volume - yet carry the highest signal concentration in the dataset.

Most near-term research impact flags extension of proven methods into adjacent problems - not dramatic breakthroughs.

The 9× research-to-corporate skew in climate shows how wide the translation gap can reach.

What's changing:
From volume and headline summary as the primary filter → to signal concentration and cross-stream convergence across 21 domains

The framework that reads volume reads noise - the framework that reads concentration reads direction.

248 cross-stream matches concentrate in AI (79) and biotechnology (32) - NVIDIA leads with more mentions than the next two organisations combined.

Personalised and precision medicine leads trend naming across the research stream - ahead of any AI label.

If volume leadership and signal concentration point to different domains - which one is your intelligence framework actually tracking?

#InnovationIntelligence #GlobalInnovation #DeepTech #QuantumTechnology #InvestmentIntelligence #ResearchToMarket #TechStrategy #InnoDexis

[InnoDexis LLC @innodexis]

Physical AI builds from perception up - not from humanoid down

The least mature Physical AI cluster commands the most public attention.

InnoDexis analysed 1,524 signals across two streams - 1,446 research disclosures from 537 institutions and 78 corporate announcements - mapping Physical AI from January through August 2026.

The technology receiving the least attention carries the clearest deployment evidence.

🔬 Key signals:
• Perception and sensing leads the technology mix at 16.2% - ahead of general robotics (14.1%) and autonomous vehicles (13.2%)
• Embodied Foundation Models and Agentic AI: highest average corporate score at 7.50 on just 8.9% of signals
• Hardware-plus-software bundling defines 54% of corporate product announcements - pure software at 6%
• Humanoid and legged robotics: almost entirely functional-prototype stage - the least mature cluster in the dataset

Why this matters:
Scaling Physical AI is blocked by integration into existing infrastructure - not by a missing scientific breakthrough.

The infrastructure layer attracts strategic investment and contract awards - not conventional venture rounds.

49% of corporate announcements score in the top tier - concentrated where research signal and capital are already moving together.

What's changing:
From humanoid spectacle and application-layer launches → to infrastructure and platform positioning as the primary value-capture layer

The value-capture layer has already shifted - most public frameworks have not caught up.

Thirteen organisations - NVIDIA foremost - appear on both sides of the research and corporate ledger months before any deal is announced.

Measurement maturity varies five-fold - autonomous driving quantifies performance 85% of the time; construction just 20%.

If public coverage concentrates on the application layer and capital on infrastructure, which half is your framework reading?

#PhysicalAI #EmbodiedIntelligence #InnovationIntelligence #DeepTech #RoboticsIntelligence #InnoDexis #AutonomousSystems #AIInfrastructure

[InnoDexis LLC @innodexis]

Physical AI builds from perception up - not from humanoid down

The least mature Physical AI cluster commands the most public attention.

InnoDexis analysed 1,524 signals across two streams - 1,446 research disclosures from 537 institutions and 78 corporate announcements - mapping Physical AI from January through August 2026.

The technology receiving the least attention carries the clearest deployment evidence.

🔬 Key signals:
• Perception and sensing leads the technology mix at 16.2% - ahead of general robotics (14.1%) and autonomous vehicles (13.2%)
• Embodied Foundation Models and Agentic AI: highest average corporate score at 7.50 on just 8.9% of signals
• Hardware-plus-software bundling defines 54% of corporate product announcements - pure software at 6%
• Humanoid and legged robotics: almost entirely functional-prototype stage - the least mature cluster in the dataset

Why this matters:
Scaling Physical AI is blocked by integration into existing infrastructure - not by a missing scientific breakthrough.

The infrastructure layer attracts strategic investment and contract awards - not conventional venture rounds.

49% of corporate announcements score in the top tier - concentrated where research signal and capital are already moving together.

What's changing:
From humanoid spectacle and application-layer launches → to infrastructure and platform positioning as the primary value-capture layer

The value-capture layer has already shifted - most public frameworks have not caught up.

Thirteen organisations - NVIDIA foremost - appear on both sides of the research and corporate ledger months before any deal is announced.

Measurement maturity varies five-fold - autonomous driving quantifies performance 85% of the time; construction just 20%.

If public coverage concentrates on the application layer and capital on infrastructure, which half is your framework reading?

#PhysicalAI #EmbodiedIntelligence #InnovationIntelligence #DeepTech #RoboticsIntelligence #InnoDexis #AutonomousSystems #AIInfrastructure

[InnoDexis LLC @innodexis]

The drug does not target MYC - MYC's own partner does.

GT19630 exploits the feedforward relationship between MYC and GSPT1 - routing both through the cell's natural disposal system simultaneously.

That bypasses the binding constraints that made MYC resistant to conventional inhibitors for decades.

The feedforward loop is what single-target approaches consistently missed - and what makes this mechanism potentially generalisable.

[InnoDexis LLC @innodexis]

A dual protein degrader reaches MYC - previously undruggable

MYC drives 70% of human tumours - and has resisted every drug design attempt for decades.

Researchers at MD Anderson Cancer Center developed GT19630 - a dual protein degrader that simultaneously eliminates MYC and GSPT1 using the cell's own natural disposal system.

GT19630 showed selective toxicity toward resistant leukemia stem cells - sparing normal ones.

🔬 Key signals:
• GT19630 prolonged survival by more than 300% in one preclinical blood cancer model
• Simultaneously degrades both MYC and GSPT1 via the cell's natural protein disposal system
• Restored sensitivity to venetoclax in resistant AML models
• Preclinical results demonstrated therapeutic safety potential distinct from single-target approaches

Why this matters:
Targeting MYC has defined one of oncology's most persistent unresolved challenges.

GT19630 exploits the feedforward relationship between MYC and GSPT1 - addressing a vulnerability single-target approaches could not reach.

The selectivity for malignant over normal stem cells is a critical finding for therapeutic safety potential.

What's changing:
From MYC classified as undruggable → to dual protein degradation through a feedforward loop making MYC actionable

This reframes how previously inaccessible cancer drivers might be approached in drug discovery.

If this mechanism translates clinically, it could offer a blueprint for targeting other proteins previously beyond reach.

The dual-degradation strategy opens a new question: which other cancer drivers share a comparable feedforward vulnerability.

If a feedforward loop can make MYC actionable, which other undruggable cancer drivers share a comparable vulnerability?

#Oncology #ProteinDegradation #BloodCancer #AML #MYC #PrecisionMedicine #DrugDiscovery #InnoDexis

[InnoDexis LLC @innodexis]

5,632 barriers - mapped by what research says it cannot yet do

Nine research fields. 5,632 barrier statements. One question: what cannot yet be done?

InnoDexis analysed 5,632 barrier statements and 2,049 stated actions across 1,323 records in nine research fields in August 2026.

This is not a map of what exists - it is a map of where each field is actually stuck.

🔬 Key signals:
• Long-term validation leads all barriers at 11.4% - distributed 9.4–14.6% across all five domains
• Healthcare's defining limit: 19.7% of barriers are that evidence is not yet in humans
• Physical sciences' defining limit: scale-up and manufacturing at 16.1% - more than 4× the healthcare rate
• Commercialisation intent: 7.6% of action items name partnering or licensing - vs 0.6–2.2% in formal vehicle fields

Why this matters:
A barrier resolved by elapsed time is more forecastable than one resolved by discovery - that distinction is only visible here.

Action items surface translation intent at five times the rate of formal vehicle fields - the earliest commercialisation signal in the schema.

The gap between intent and formal record is where early positioning has the most leverage.

What's changing:
From research assessed on publications and builds → to a self-described barrier map showing what each domain cannot yet do

The resolution horizon is now visible - domain by domain, barrier by barrier.
43% of barrier statements resist any category - the frontier is particular, not systemic.

Standards and protocol development leads action items at 14.8% - methodological infrastructure before the science moves forward.

If time-bound barriers are more forecastable than discovery-bound ones, why are most frameworks treating both as the same risk?

#ResearchIntelligence #InnovationIntelligence #DeepTech #ResearchToMarket #EarlyStage #InnovationScouting #LifeSciences #TechStrategy #ScienceToCommerce #InnoDexis