Some high priority ideas for GSoC are those that will bring LPython closer to production:

• Refactor cpythoncall into an ASR pass #2470
• fastGPT implementation in LPython #2409
• Implement all NumPy intrinsics via ASR #2474
• Improve CPython interoperability #2471
• Improve dictionaries, sets, tuples, lists for all backends #2473
• Get ASR->Fortran finished
• Get ASR->Python finished
• Improving the symbolic backend
• Performance: take some simple benchmarks and ensure they are as fast or faster than Clang/C++.
• Any other issue from LPython Roadmap #1600

In general, improving the interoperability with CPython and the end user experience of actually using LPython to speedup existing Python codes, as well as writing new codes from scratch, is what we need to focus on. Together with performance.

Similar issue for LFortran: lfortran/lfortran#3280