Remove 'Code review required' setting and require it always.

I disagree, especially on the AI part. 

I tried to create a data model with chat GPT and the result was disastrous. That makes that checking code with AI can only give a feeling that things are done right, while the result can be totally wrong. 
 

The only solution is to have developers who can read code, see bugs, and take their responsibility from the minor features, to the data model, the user experience, and the long term end result.

If it doesn’t work, you should get rid of all the developers, who don’t understand the whole, find others, or cancel the project.

I wrote more on the previous comment but it was not saved, and this has been a known Thinkwise bug for 3 weeks.

If the developers did not stay awake 24/7, and did not do their best to fix this in 1 or 2 days, code review is not going to help,

Code review and AI don’t solve this, they only make it worse.
 

@Daan Heemskerk I understand your gripes with AI. The main problem with AI is that it has problems with longer and more complex statements. The software factory deals with this problem by chaining AI commands. the 2024.2 version of the platform was just released, this also includes AI data model creation. You would still need to do some fine tuning afterwards but it gives a pretty decent base to work from. I can tell once I got used to working with the AI data model creation, going back is not an option anymore. Sure there could be some more fine tuning options in the future. But this currently saves me loads of time for creating data models for reproduction scenario's. 

I invite you, if you have not done so yet, to try the data model creation in 2024.2. Did you use the 3.5 or 4.0 model of chatgpt? I can tell you from experience that the 4.0 is leagues better compared to the 3.5 model. 

I did not see any data model creation in version 2024.2 so please direct me to it.

The last days I have been working on 3 tables and a huge view over them,  of 107 columns, which should all be prefixed correctly, and have a perfect group ordening to keep things easy to read and maintain in the future. Each change required a lot of corrections to both the views and the tables.

It would be great if AI could do this for me …  

But, unfortunately AI does not understand the fact that columns first thought of as “owned_by_*” should actually be corrected to “home_*” at the moment that ownership is not the driving factor, and could be different.

Those decisions, which will become long term, when more functionality is build, are still outside the scope of AI.

@Daan Heemskerk 

The create data model task is available in data model tab. If you have any questions or remarks regarding this then please create a new topic in questions and conversations as we want to keep this topic focused on the code review idea.: 

https://docs.thinkwisesoftware.com/docs/sf/data_model#create-ai-generated-data-model

 

This is not serious, as the task mostly is to modify an existing model and to transfer the data.

I tried ChatGPT 12 months ago, and it was impressive that he/she knew a lot about the topic, but the data model became a mess at the moment that I asked to include some details.

One thing ChatGPT did for me was writing a perfect MS-Dos boot sector virus (After which ChatGPT was not able to admit that he/she just wrote a virus ...).

But writing for, and adjusting for specific requirements stays out of the range of a language based model … it can give answers to average questions as it is trained on large numbers of models. It can not solve a problem which goes out of that range ….

For me it only becomes interesting when I can ask to refactor ….

Let's get back on topic of the idea. Please vote if you agree.