Designed around discovery.
What would a language look like if the whole research workflow mattered? These are the questions guiding Aner.
Make the assumptions behind an experiment as understandable as its code.
Reproducibility as part of the workflow
Can explicit partitions, fitted preprocessing, and source identity reduce avoidable research mistakes?
Data contracts before model code
Make schema, missingness, provenance, and resource decisions explicit without a long cleaning script.
Learning that can be inspected
Connect source, data, network structure, and training history in a coherent research workspace.
One language, deliberate execution choices
Evaluate explicit CPU and accelerator placement, numerical agreement, and synchronized operation costs before extending to larger systems.
An ecosystem others can build on
Design stable library boundaries, package identity, and documentation that grows with the language.
Have a question worth exploring?
Help turn a real research challenge into a better language design.