According to the latest reports on code quality by the DORA lab as well as by GitClear, the stability of the code pushed to production as well as the overall throughput is actually declining as developers are increasing their use of AI for code generation. This is the opposite of the result that we expected.

We thought that because AI would not make typos and basic formatting/styling mistakes, it would deliver more stable software. We also thought that it would increase the throughput by 2-3 fold; and we kind of experience that as well when we actually use them.

In this session, I want to deep dive into what is going wrong and what are the remedies to this situation. Here are some of the checks and balances that we can discussion:

1. Code reviews – quality and speed | expected impact
2. QA – quality, investment and speed | expected impact
3. Monitoring and observability – quality and investment | expected impact

For interested attendees, I would recommend that they go through these before the session:

• DORA report 2024
• GitClear Code Quality Report 2025
• OR watch this Youtube video discussing the GitClear report

• Type of talk: Deep Dive