Understand matter to become better engineers of reality.
Artificial intelligence can only learn what has been measured, and most of the physical world has not been. Wallace measures it, models it and puts the models to work.
What we measure, what we model and what we build.
Artificial intelligence has read the internet. It has never measured reality.
To use artificial intelligence to discover more about reality, we have to datamaxx the physical world.
Models learned from text
Language models trained on text. Text says little about how cells, molecules and materials actually behave.
Physical data is scarce
Biology, chemistry and materials are measured rarely, inconsistently or not at all. Failed experiments are thrown away.
Measurement is getting cheaper
Robolab automates the lab with a target of 1 cent per measurement, which makes measuring everything affordable.
Every test result trains a model that earns revenue from predictions, so each test is worth more than the result alone.
Measure the world finely enough to predict it, then engineer it.
Measure
Physical world data at fine granularity and high frequency, from molecules to ecosystems.
Predict
Models trained on that data forecast what will happen before it happens.
Engineer
The same models show how to use the resources we have far better, and how to change outcomes.
The work covers six areas.
Where data is missing today.
The body over time
Healthy people, measured repeatedly. Almost every record today is one snapshot.
Microbiomes
Gut, soil and water microbes shift daily. Almost nobody samples often enough to model them.
Food as eaten
Composition, contaminants and residues change by batch, farm and season.
Pathogens and resistance
New variants and antimicrobial resistance appear faster than public data updates.
Soils and crops
Nutrients, microbes and residues per field, per season.
Water and air
Contaminants, spores and allergens change hour to hour. Most testing is monthly.
Batteries and materials
Ageing data takes years, and every new chemistry resets it.
Failed reactions
Chemistry models learn from published successes. Failed reactions are rarely published.
These systems change constantly, so a model trained once goes out of date. Realm is updated with new measurements every day.
Where the same model applies.
Biological systems
Humans, animals, plants, crops and microbes. Predict bodies and metabolism, design better molecules.
Ecosystems
How species, soils and water interact. Predict change and engineer healthier landscapes.
Weather
The chemistry and physics of the atmosphere. Forecast further ahead and design interventions.
Resources
Soils, minerals, water and feedstocks. Predict yield and grade, recover more and waste less.
Materials and chemistry
Catalysts, batteries, alloys and polymers. Predict properties and invent new materials.
Built environment
Concrete, steel and structures over decades. Predict ageing and failure before it happens.
Oceans
Currents, chemistry and marine life. Predict fisheries, carbon uptake and coastal change.
Space and extreme environments
Radiation, vacuum, heat and pressure. Data almost nobody can generate.
Biology comes first, because health matters most and its data is the scarcest.
01 · Now
Biology
Bodies over time, food, microbiomes and pathogens.
02 · Next
Chemistry and materials
Batteries, reactions, construction and corrosion.
03
Resources
Soils, crops, water and recycled feedstocks.
04
Ecosystems
Species, air, water and oceans.
05
Weather
Atmospheric chemistry and physics.
06
Space
Extreme environments almost nobody can test.
Each step reuses the same platform: Robolab measuring, Realm modelling.