Know where to look.
Models that forecast outcomes and know how sure they are.
Scientific intelligence, starting with chemistry
Discovery.Compounded.
Models that predict the lab. Labs that correct the models. Every experiment makes the next one smarter.
Explore our thesis ↓FIG. 001 / CONTINUOUS FIELD
PREDICT · TEST · LEARN
PREDICT · TEST · LEARN
One surface, seen from many viewpointsScroll to discover ↓
01 / Our thesis
Chemistry is a vast map.
We have barely explored it.
Around 1060 drug-like molecules could exist. We have catalogued about 200 million, mostly by hand. We think science can learn from every single run.
FIG. 002 / Orders of magnitudeLog scale · 10⁰ → 10⁶⁰
108.3 · everything ever catalogued
1011.2 · largest virtual library
1060 · what could exist
10⁰10¹⁰10²⁰10³⁰10⁴⁰10⁵⁰10⁶⁰
1060possible drug-like molecules
87%of chemists have failed to reproduce a peer’s result
9.6%of drugs entering trials get approved
$2.6Bto bring one new drug to market
02 / Our approach
One continuous loop.
An expanding frontier.
Let reality answer.
Robotic labs run the experiments that matter most, around the clock.
Make knowledge cumulative.
Every result, including failures, makes the next prediction sharper.
03 / First frontier
Why chemistry first.
Almost everything is made of it.
A $5 trillion industry inside nearly every product. Fast, honest feedback: a reaction works or it does not, and we know within hours.
A
Catalysis
Same reactions, less energy.
EnergyB
Medicines
Faster routes to new treatments.
HealthC
Materials
Designed for a property, not found by luck.
ElectrificationD
Sustainable chemicals
Everyday chemicals, without oil.
Climate04 / The lab
A lab that
never forgets.
Every step defined in software and recorded in full, so any result can be rerun exactly, by any machine.
Design
Experiments written as code.
Run
Robots, around the clock.
Measure
Calibrated, comparable data.
Remember
Failures kept, nothing lost.
FIG. 003 / Autonomous workcell · 96 wellsSimulated
Reacting 0Measuring 0Model agreement 0.91
05 / What we believe
Principles for
machine-speed science.
i.
Every result counts.
Failure is data too.
ii.
Confidence must be earned.
Every prediction says how sure it is.
iii.
Reproducible by default.
If it can’t be rerun, it’s a rumour.
iv.
Wider access, better questions.
Discovery for anyone with a good question.
PREVIEW / Ask the systemIllustrative
A catalyst that makes ammonia at a fraction of today’s energy cost.
Explored5,240
Predicted viable118
Sent to the lab16
Confirmed3
For the questions still unansweredBring us a molecule
Bring us a molecule
you can’t make yet.
Partnering with chemicals, materials and pharma teams. Hiring researchers, chemists and robotics engineers.
hello@parallaxmachines.com