Sunsetting Siml.ai
We shut down the Siml.ai platform on 31 March 2025. Here is what Siml.ai was, what happened to user data, and where the work continues.
On 31 March 2025 we turned off the Siml.ai platform. We announced the closure to users on 21 March, and on the last day every simulator, environment, project, dataset and plugin stored on the platform was permanently deleted, as the notice said it would be. This post is the longer version of that announcement: what Siml.ai was, why we closed it, and where the work goes from here.
What Siml.ai was
Siml.ai started in 2021 as a browser-based platform for building AI-based physics simulators. You could describe a problem with equations and geometry, train a physics-informed neural network or a neural operator on GPUs we provided, and explore the results in an interactive 3D viewer. Over the following releases we added custom equations and plugins, datasets, expressions in constraints, compute credits, and a lot of work on the user experience.
It brought physics-informed machine learning to people who would never have set up a training cluster themselves, and showed many teams for the first time what neural surrogates could do for their products. Thank you to everyone who built something on it, filed a bug or argued with us on Discord.
Why we closed it
Over the life of Siml.ai, the conversations that mattered most kept pointing in the same direction. The engineering teams with the hardest problems did not want a hosted sandbox for training models from equations. They already had what is far more valuable: years of high-fidelity simulations, run with their own solvers, on their own geometry, stored in their own data centres. What they needed was a way to turn those archives into models, evaluate the models properly, and deploy them inside their own infrastructure, without their data leaving the building.
That is a different product, with different priorities: connectors to the solvers companies actually use, datasets that keep every simulation's setup and provenance attached, evaluation on held-out designs, and deployment on workstations, clusters and air-gapped sites. It could not be built as a feature of Siml.ai, and keeping both running would have meant doing neither well. So we chose to close Siml.ai and put the whole team behind what came next.
What happened to your data
As announced, all data stored on the platform was deleted when it shut down on 31 March 2025. The product page at dimensionlab.org/products/simlai stays online as an archive, and the documentation remains on GitHub for reference. If you have questions about an account or data you held on Siml.ai, please get in touch.
Where the work continues
The ideas behind Siml.ai, making physics AI practical for the people who engineer real products, continue in DimensionLab Frontier: a platform to prepare simulation data, train and evaluate physics models, and deploy them on your own compute. You can read about it on the Frontier page, and if it sounds like something your team needs, join the waitlist.
Thank you for building with us.