Causal World Models for Physical World
Causal World Models for Physical World
SYKY is building causal world models that discover the hidden variables, mechanisms, and causal structure that govern the physical world.
SYKY is building causal world models that discover the hidden variables, mechanisms, and causal structure that govern the physical world.
WORK & EDUCATION BACKGROUNDS
TEAM
Our team brings experience from leading technology companies and research institutions.
Our team brings experience from leading technology companies and research institutions.
+ BEYOND PATTERN MATCHING
Prediction ≠ understanding
"
“The goal isn’t to memorize the world. It’s to discover how the world works.”
— syky.
"
“The goal isn’t to memorize the world. It’s to discover how the world works.”
— syky.
+ THE HIDDEN WORLD
What we observe is only the surface.
Mass. Friction. Compliance. Contact. Geometry. These hidden properties determine what happens next, but they are often absent from observational data.
Mass. Friction. Compliance. Contact. Geometry. These hidden properties determine what happens next, but they are often absent from observational data.
— Discover hidden variables..
— Identify reusable mechanisms..
— Reason beyond observed situations.
SYSTEM FOUNDATION
The discovery loop
A world model that doesn’t only observe the world. It experiments with it.
A world model that doesn’t only observe the world. It experiments with it.
Observe
Build representations from interaction and identify where the current model fails or becomes uncertain.
Observe
Build representations from interaction and identify where the current model fails or becomes uncertain.
Intervene
Choose actions that distinguish between competing explanations of what is happening.
Intervene
Choose actions that distinguish between competing explanations of what is happening.
Discover
Identify variables and mechanisms that explain the outcome — then integrate them back into the world model.
Discover
Identify variables and mechanisms that explain the outcome — then integrate them back into the world model.
THE SYSTEM
THE SYSTEM
An engine for discovering
how the physical world works.
SYKY combines an AI scientist, recursive self-improvement, and a distributed network of physical experiments.
SYKY combines an AI scientist, recursive self-improvement, and a distributed network of physical experiments.
01 / AI Scientist
Decides what to learn next. Identifies gaps in the current world model, proposes hypotheses about hidden variables and mechanisms, and designs interventions that can distinguish between competing explanations.
01 / AI Scientist
Decides what to learn next. Identifies gaps in the current world model, proposes hypotheses about hidden variables and mechanisms, and designs interventions that can distinguish between competing explanations.
01 / AI Scientist
Decides what to learn next. Identifies gaps in the current world model, proposes hypotheses about hidden variables and mechanisms, and designs interventions that can distinguish between competing explanations.
02 / Recursive Self-Improvement
Improves how the world is represented. Uses experimental evidence to refine representations, mechanisms, and experimental strategies — so previous discoveries make future discovery more efficient.
02 / RSI
Improves how the world is represented. Uses experimental evidence to refine representations, mechanisms, and experimental strategies — so previous discoveries make future discovery more efficient.
02 / Recursive Self-Improvement
Improves how the world is represented. Uses experimental evidence to refine representations, mechanisms, and experimental strategies — so previous discoveries make future discovery more efficient.
03 / DARE Network
Turns hypotheses into physical evidence. A distributed network of robots across labs and real-world environments executes targeted interventions and returns evidence the model cannot obtain from observation alone.
03 / DARE Network
A distributed network of robots across labs and real-world environments executes targeted interventions and returns evidence the model cannot obtain from observation alone.
03 / DARE Network
Turns hypotheses into physical evidence. A distributed network of robots across labs and real-world environments executes targeted interventions and returns evidence the model cannot obtain from observation alone.
SYKY'S TEAM
The people behind the next super intelligence.
SYKY is built by a multidisciplinary team bringing together artificial intelligence, Robotics, product strategy, and autonomous systems.
SYKY is built by a multidisciplinary team bringing together artificial intelligence, Robotics, product strategy, and autonomous systems.
Metaseeker
Previously built Vlge Worldbuilder, generating 4M+ hours of gameplay data.
Metaseeker
Previously built Vlge Worldbuilder, generating 4M+ hours of gameplay data.
Nakimo
Previously at Intel Robotics and RWTH Aachen, Germany.
Nakimo
Previously at Intel Robotics and RWTH Aachen, Germany.
Macron
Previously in operations at Amazon, Australia.
Macron
Previously in operations at Amazon, Australia.
Tobor
Previously built causal models at Meta; former founder with a successful startup exit.
Tobor
Previously built causal models at Meta; former founder with a successful startup exit.
BEGIN WITH SYKY
Put intelligence into motion.
Build autonomous systems that can understand, decide, and act—without becoming impossible to control.
BEGIN WITH SYKY
Put intelligence into motion.
Build autonomous systems that can understand, decide, and act—without becoming impossible to control.
BEGIN WITH SYKY
Put intelligence into motion.
Build autonomous systems that can understand, decide, and act—without becoming impossible to control.