In T194934, we had a great overview on the definition of technical depth. Among many other aspects, we discussed to reduce unintentional technical risks by better onboarding for greenhorns. I think this could be supplemented by AI that spots weak points and bugs in the code. We have a lot of training data and with ORES a lot of expertise in this field. Especially volunteers and employees who are new to the language or API make mistakes that should be easy to spot with AI methods.
For instance, in [like this](https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/extensions/Math/+/38d3ebf4e9ec99b4876a680541ae8ad783cfa7a7%5E%21/)
```
- $host = array_rand( $this->hosts );
+ $host = $this->hosts[array_rand( $this->hosts )];
```
the unawareness of the details of the php function array_rand might be a typical example of code developed by volunteers that do not use that particlular language on dayli basis and this should be easy to learn.