[General] JOSS Paper review comments

[abhijit-c]

Hello,

I'm still working through my entire review of the (large!) codebase, but I've at least some feedback re. the JOSS paper that can be handled now. These are largely minor / nitpicks, I'm happy to say that overall it looks good to me!

All line numbers referenced aren't from the markdown file, but physical line numbers on the PDF.

1. Re. the AI Usage Disclosure on line 175: Per the AI policy in JOSS (https://joss.readthedocs.io/en/latest/policies.html#ai-usage-policy) I believe the section ought to revised to include more details about the tool specifics and versions.
2. Re. the GPU claim on line 21 and 87: Extremely minor nitpick, it feels slightly off to call GPU integration seamless when in reality it requires either ExaModels or MadNLP per http://control-toolbox.org/OptimalControl.jl/stable/manual-solve-gpu.html (and is NVIDIA only). I don't think these are unreasonable requirements, and I do think that OptimalControl.jl 's design that enables GPU solvers is a very worthy contribution worth calling out. However, I think at minimum, it should be clarified that it's NVIDIA only.
3. Re. Line 166: I think would be a good idea to archives of one such workshop. I believe the package was presented in juliacon2025 , no? I'm happy if you don't want to do this though.
4. Re. Line 51: "our plan is" should be present tense.
5. Re. Line 71: "Matlab" -> "MATLAB"

Otherwise, happy with the paper, excellent work! I felt somewhat conflicted as to whether the illustrative example subsection belongs in a JOSS paper, but since the other reviewer feels strongly about expanding it even I'll refrain from suggesting anything since I don't have hard opinions on the matter.

[ocots]

Hello @abhijit-c,

Thanks for the careful reading of the paper and the very positive overall assessment. All changes are gathered in PR #801 (joss-reviewer-abhijit-c → joss).

---

1. AI Usage Disclosure

We have updated the AI Usage Disclosure section to name the tools and versions explicitly, as required by the JOSS AI policy:

Generative AI tools—specifically Claude Sonnet (version 4.6) and Claude Opus (version 4.8) by Anthropic—have been used to assist with code refactoring and generation of unit and integration tests, always under human review.

2. GPU claim ("seamless" + NVIDIA-only)

We agree that "seamless" was misleading and that the NVIDIA constraint should be stated explicitly. We have updated the three relevant passages:

Summary (line 21 of PDF): replaced "enabling seamless execution on both CPU and GPU with minimal user intervention" with:

The architecture relies on a modeler-solver separation that provides a modular and extensible foundation, enabling GPU execution (currently limited to NVIDIA hardware) with minimal user intervention alongside standard CPU solvers.

State of the Field — Julia packages bullet (line 71 of PDF): replaced "works seamlessly on both CPU and GPU" with:

OptimalControl.jl supports both CPU and GPU execution (currently NVIDIA-only, via ExaModels.jl + MadNLP.jl), [...]

Key design trade-off §2 (line 87 of PDF): replaced "seamless CPU-to-GPU transition" with an accurate description that also documents the :gpu token as the user-facing entry point:

Users select the ExaModels modeler with MadNLP and the CUDSS linear solver for GPU execution, or simply append the :gpu token to the solve call. [...] GPU support is currently limited to NVIDIA hardware and to the ExaModels + MadNLP modeler-solver combination.

3. Workshop archive (JuliaCon 2023)

We have added a citation to the JuliaCon 2023 presentation in the Research Impact section. Note that the documentation itself served as the presentation support, so there are no separate slides; the available resources are:

• Talk abstract: https://pretalx.com/juliacon2023/talk/HDC8F7/
• Video recording: https://www.youtube.com/watch?v=RYUtVnzLj5k

The citation [@caillau2023juliacon] has been added to paper.bib and referenced in the text.

4. "our plan is" → present tense

You are right that "our plan is to provide" sounds prospective. We have rephrased line 51 to:

In contrast with InfiniteOpt.jl that is designed as an extension of JuMP.jl, our package provides a modeler-agnostic approach, accepting general Julia code and leveraging various optimization solvers.

5. "Matlab" → "MATLAB"

Fixed directly in the source. Thank you for spotting this.

---

We hope these changes address all the points raised. Thank you again for the thorough and constructive review.