MurphyLabs
MurphyLabs
Mayan Wasu was a lieutenant in the Singapore Armed Forces, has co-founded a prediction-market trading fund and published CV and HCI research. Now he's accelerating the path to a future where robots live amongst humans
Mayan Wasu was a lieutenant in the Singapore Armed Forces, has co-founded a prediction-market trading fund and published CV and HCI research. Now he's accelerating the path to a future where robots live amongst humans
Meet
Meet
Mayan
Mayan
Robots are coming into the world. Something has to decide they're ready. MurphyLabs is building the alignment layer for physical AI — the proving ground every machine passes through before it moves among us.
Robots are coming into the world. Something has to decide they're ready. MurphyLabs is building the alignment layer for physical AI — the proving ground every machine passes through before it moves among us.
In the physical world, a model failing isn't a bad answer. It's a person.
We stopped programming robots and started training them, which means we traded predictability for capability. A trained policy is brilliant at what it was shown and unrecognizable in front of what it wasn't — and no one is showing them the hard part, because failures live on a long tail nobody has ever collected.
MurphyLabs is building the place those failures get found on purpose. We search a machine for its worst cases before reality does, amplify each one into everything around it, and hand back the data that closes the gap.
Every failure found is a failure that doesn't happen. Multiply that across every robot from every maker, and you get the largest map of how physical AI breaks ever assembled — the missing input to training it, certifying it, and insuring it.
In the physical world, a model failing isn't a bad answer. It's a person.
We stopped programming robots and started training them, which means we traded predictability for capability. A trained policy is brilliant at what it was shown and unrecognizable in front of what it wasn't — and no one is showing them the hard part, because failures live on a long tail nobody has ever collected.
MurphyLabs is building the place those failures get found on purpose. We search a machine for its worst cases before reality does, amplify each one into everything around it, and hand back the data that closes the gap.
Every failure found is a failure that doesn't happen. Multiply that across every robot from every maker, and you get the largest map of how physical AI breaks ever assembled — the missing input to training it, certifying it, and insuring it.