[gretadolcetti]

This pull request adds a new blog post, "Tests Pass, Smells Stay: Adding a Quantum Linter to Qiskit IVR," which describes integrating the LintQ quantum linter into the Mellea Instruct-Validate-Repair (IVR) loop for Qiskit code generation. The post explains how quantum-specific code smells can persist even after passing all tests and demonstrates how combining functional testing with quantum-aware static analysis significantly increases the share of generated programs that are both correct and free of quantum code smells.

Key additions and themes:

New Content: Quantum Linting in IVR

• Introduces LintQ, a CodeQL-based static analyzer for Qiskit, and explains its integration into the Mellea IVR loop alongside functional tests.
• Details specific quantum code smells detected by LintQ, such as applying gates after measurement and unused qubits, with a table of example rules.

Validation Pipeline and Implementation

• Provides Python code snippets showing how correctness and LintQ validation are sequenced, ensuring only functionally correct programs are linted.
• Explains how validation results are fed back into the repair loop, enabling iterative improvement of generated code.

Results and Impact

• Presents experimental results on Qiskit HumanEval (hard) and Qiskit Quantum Katas

[gretadolcetti]

@ajbozarth

[planetf1]

Hi @gretadolcetti - thanks for blog post. great to have a quantum example. Just two quick things

a) I noticed the description stopped after 'Presents experimental results on Qiskit HumanEval (hard) and' - not a big deal but you might have wanted to put more

b) You need to 'signoff' on your commits. this is why DCO is failing. Since you only have one commit it's fairly easy to add it (for example read here

Will take a read

[gretadolcetti]

Hi @planetf1, thanks for the feedback! I should have fixed both aspects.

[planetf1]

Thanks @gretadolcetti! I tried running the code (macOS, mellea 0.8.0) and left a few comments inline.

[gretadolcetti]

Thanks @planetf1 for the very quick and detailed review!

I believe I’ve addressed all your concerns in the latest fixes. Most of them stemmed from the fact that I intended the blog post not as a directly runnable example but as a didactic one, with pointers to the real code (in generative-computing/mellea#1716), where users can run it themselves and experiment, building on @ajbozarth’s existing example.

That said, I think your comments have improved the blog post a lot!

[ajbozarth]

Review generated by Claude. I'll do a separate manual pass through the blog myself.

Thanks @gretadolcetti — this is a great use of IVR, and a fun one to see written up. The code lines up with the current mellea APIs and the Results numbers match the example's benchmarking, so no concerns on accuracy.

Before this merges: the mellea example this links to may not include the LintQ code (see generative-computing/mellea#1716), so let's make sure the post doesn't depend on it — details inline. A few small copy nits inline as well.

[gretadolcetti]

Hi @ajbozarth, thanks for the feedback and suggestions!
I should have clarified all the points that you mentioned.

[ajbozarth]

Review from Claude — the broken paragraph below came from a mis-anchored suggestion in Claude's earlier review, not your edits. Apologies for the churn.

The LintQ-paper citation and the CodeHelm note both look right. One leftover to fix, inline.

[ajbozarth]

Review from Claude.

One suggestion on the paper citation, inline.

[github-actions]

@gretadolcetti — this PR is scheduled to merge on 2026-10-15. Please enable auto-merge manually so it goes through the merge queue with CI.

[ajbozarth]

I do believe this looks good pending what happens with generative-computing/mellea#1716, either way we'll need to update the release date, I would suggest updating it to next Monday for now

[gretadolcetti]

Thanks @ajbozarth for the positive review, I updated the release date to Monday 19th of October.

[github-actions]

@gretadolcetti — this PR is scheduled to merge on 2026-10-15. Please enable auto-merge manually so it goes through the merge queue with CI.

[github-actions]

@gretadolcetti — this PR is scheduled to merge on 2026-10-15. Please enable auto-merge manually so it goes through the merge queue with CI.

[github-actions]

@gretadolcetti — this PR is scheduled to merge on 2026-10-15. Please enable auto-merge manually so it goes through the merge queue with CI.

[github-actions]

@gretadolcetti — this PR is scheduled to merge on 2026-10-15. Please enable auto-merge manually so it goes through the merge queue with CI.

[ajbozarth]

I've removed the remind from the description so the bot stop posting everyday. I am still actively reviewing both this and more so the example PR to actually test run them manually and validate the readme and blog steps (slow due to low bandwidth). Once the other PR is merged we can update the date on this and release it.

[ajbozarth]

Per my recent comment on the example PR generative-computing/mellea#1716 (comment) I think we can unblock this with some minor tweaks and post it tomorrow.

And sorry for the delays in reviewing the PRs

[ajbozarth]

not a huge issue given you define it in a code block below, but technically validate_lintq won't be in validation_helpers after addressing my review in the example PR. We can probably leave it as is, but if you have a better way of writing the import it may be clearer to a reader

[gretadolcetti]

Solved in gretadolcetti@d305c3f

[gretadolcetti]

Perfect, @ajbozarth. I should have addressed all of your last comments.

[AngeloDanducci]

Thanks for this, easy to follow, generally looks good! I'll let Alex give final approval as he's been more involved.