Editorial illustration for Decart’s world model simulates hours of photorealistic driving
Decart’s world model simulates hours of photorealistic...
Decart is selling a machine that hallucinates highways. The two-year-old company just banked $300 million in new funding. Its product, the Oasis 3 world model, doesn't just render a street.
It simulates hours. Photorealistic driving, generated from pure code: lane markings, weather, a sudden jaywalker. That funding round is a stark market signal.
Investors now believe generating entire synthetic worlds is a viable business.
Oasis 3’s release comes a few weeks after two-year-old Decart raised $300 million, which Leitersdorf says followed “huge demand increases for the models we built” in e-commerce, live streaming and physical AI.
The immediate pitch is for autonomous vehicle training—simulate a million miles of chaotic storms, risk nothing. But that $300 million reveals broader ambition. It’s a bet on a persistent digital canvas for rent.
E-commerce sites could generate endless product videos. Live streamers might broadcast from fantastical, consistent environments. Every industry touching the physical world, from robotics to retail, becomes a potential client.
The caveats are serious. Oasis 3 fails at true novelty. It struggles with edge cases a child would grasp, producing a world that's a statistical echo—brilliant but predictable.
These aren't minor bugs. They're the fundamental limits of building from data patterns, not physical laws. Yet the gap narrows fast.
The technical feat is undeniable: Decart is engineering a new raw material, coherent and photorealistic time. The real uncertainty isn't the engineering. It's what we'll build once that material is cheap.
We are getting alarmingly good at teaching machines to dream. We just haven't decided what those dreams are for.
Related Reading
Industry Applications Westinghouse teams with Google Cloud to build AI platform for nuclear power
Industry Applications NVIDIA NeMo powers telco reasoning model for autonomous network workflows
Industry Applications Month-1 Agent Adds Holistic Observability with Trace IDs and Token Tracking
From Archives AI‑enhanced lessons in Sierra Leone: teachers lead impact study
From Archives MedicalRec releases MedicalRec-Bench: 5,000+ entries for medical image classification
Common Questions Answered
What is Decart's Oasis 3 world model and what can it simulate?
Oasis 3 is a world model that generates photorealistic driving simulations for hours at a time, complete with realistic details like lane markings, weather conditions, and pedestrians. The system creates these complex synthetic environments entirely from code, enabling realistic autonomous vehicle training scenarios without physical risk or resource constraints.
Why did Decart's $300 million funding round signal a major market shift?
The substantial investment demonstrates that investors now view synthetic world generation as a viable and valuable business model. This funding round indicates broader market confidence that AI-generated digital environments have significant commercial potential beyond just autonomous vehicle applications.
What are the potential applications for Oasis 3 beyond autonomous vehicle training?
Beyond autonomous driving, Oasis 3 could be used by e-commerce sites to generate unlimited product demonstration videos, enable live streamers to broadcast from consistent fantastical environments, and serve robotics and retail industries that need to simulate physical-world interactions. The technology represents a persistent digital canvas that multiple industries could license and customize for their specific needs.
How does Decart's world model approach differ from traditional simulation methods?
Rather than manually building or filming driving scenarios, Oasis 3 generates photorealistic simulations purely through code, allowing for infinite variations and hours of continuous simulation. This approach eliminates the need for expensive physical testing while maintaining high visual fidelity and realism that's critical for training autonomous systems.
Further Reading
- The Waymo World Model: A New Frontier For Autonomous Driving Simulation — Waymo Blog
- Decart AI Lab | Real-Time World Models — Decart AI Lab
- Researching Foundational Real-Time World Models - Decart AI Lab — Decart AI Lab
