As our compute capacity grows, science simulations are not only becoming bigger, but more complex. Simulations are carried out at multiple scales and using multiple kinds of physics at once. Boundaries are irregular, grids are irregular, computational domains can be dynamic and complex. In such scenarios, the ideal way to parallelize often cannot be statically determined. At the same time, hardware is becoming more heterogeneous and difficult to program. Increasingly, scientists are turning to asynchronous, dynamic parallelism in order to make the best use of increasingly challenging hardware. As a result, numerous frameworks, platforms, and specialized languages have sprung up to answer this need.

The objectives of this workshop are to bring together experts in asynchronous many-task frameworks, developers of science codes, performance experts, and hardware vendors to discuss the state-of-the-art techniques needed to program, analyze, benchmark, and profile these codes to achieve maximum performance possible from modern machines. This workshop will promote a dialogue between these communities, and help identify challenges and opportunities for advancement in all the disciplines they represent.

The topics of interest include, but are by no means limited to:

- Novel task-based runtime environments
- Experiences of using task-based runtime environments for large applications
- Experiences comparing task-based runtime environments
- Experiences gathered from porting one large-scale parallel solution to another, e.g., MPI to Charm++, etc.
- Profiling and performance monitoring of task-based environments
- Benchmarks for task-based runtimes
- Tools for debugging programs using task-based runtimes
- Challenges to task-based runtimes in scaling to large clusters
- Hardware challenges and solutions in using task-based environments
- AI-Assisted and Agentic Code Development
- Usage of AMTs in AI workflows