Johan Bujari

[Johan Bujari @JohanBujari]

@thsottiaux 5h usage always resets to 99% for me, have seen some other people say the same

[Johan Bujari @JohanBujari]

@Im_IrushiK

[Johan Bujari @JohanBujari]

@thsottiaux Please look at the usage limits. They are draining way too fast.

[Johan Bujari @JohanBujari]

@rezoundous Yes, I am noticing that too.

[Johan Bujari @JohanBujari]

I used Codex to build a small Codex skill: record-browser-proof.

When Codex uses Browser to verify a UI fix or feature, it captures visual proof and can turn it into an MP4.

Useful when “fixed” or "done" needs evidence, not just a summary.

Repo:
https://t.co/oXbwxbxOrt

[Johan Bujari @JohanBujari]

@thsottiaux Performance

[Johan Bujari @JohanBujari]

@moshhamedani Codex feels like it just goes deeper to get work done even when you sometimes may think it won’t make it.

[Johan Bujari @JohanBujari]

@sama 1. Frontend
2. It should not always agree with everything I say. Additionally, when it makes a mistake and I correct it, it shouldn't just reflexively say, 'You are right.'

[Johan Bujari @JohanBujari]

@thsottiaux 1. Can’t open terminal while on full screen mode.
2. There should be a way to change the model or reasoning when implementing a plan.

[Johan Bujari @JohanBujari]

This view in Codex, using Debug Mode to fix a bug and the Browser Use to reproduce and verify the fix, is magical and too addictive. What they have created!

@OpenAIDevs @thsottiaux 👏👏

[Johan Bujari @JohanBujari]

@thsottiaux It is so good at running commands

[Johan Bujari @JohanBujari]

@ericzakariasson It would be great if you added Azure too

[Johan Bujari @JohanBujari]

More platforms will be supported soon. Open source, feedback welcome.

Open source, feedback welcome.
📷 https://t.co/tXIZE13Xzz
#opensource #npm #AI #DevTools #CodeReview #AzureDevOps

[Johan Bujari @JohanBujari]

After the launch of the Claude PR reviewer I was excited to try it, then I realized it only supports GitHub.
A lot of my mates work on Azure DevOps, so that did not really help our workflow.
So I decided to build something simple and open source to help with that.

[Johan Bujari @JohanBujari]

Introducing axiom-pr: an AI PR reviewer you can run directly from npm.
Lightweight, developer friendly, and flexible enough to work in environments where existing AI reviewers do not.
If you review PRs on Azure DevOps and want AI in the loop, this might help.

[Johan Bujari @JohanBujari]

@vasuman In my experience, Codex tends to generate sloppy code while Claude produces more human-readable code

[Johan Bujari @JohanBujari]

@OpenAIDevs @romainhuet Do you plan to expand its usage on other version control tools such as Azure Repos?

[Johan Bujari @JohanBujari]

@cursor_ai Will you be adding the codex model?

[Johan Bujari @JohanBujari]

@StackOverflow How come your surveys always show data not in favour of AI usage and trust while in fact more and more developers are using and adopting AI as a productivity tool. I think you are worried about stack overflow no longer being the first tool of soving programming issues

[Johan Bujari @JohanBujari]

@cursor_ai I gave it a try and still not better than claude. I think there's something off, I dont know but it didn't feel as smart as expected