VS Code: Vanilla vs Kilo Code

I've recently lost access to my main PC (the SSD is currently back with the manufacturer) - in the meantime, I've been mostly without a machine, until now.

I was given an old laptop running Windows Vista, so as you can imagine, it needed a minor overhaul - mostly because all the browsers were outdated and not upgradable. However, as part of this overhaul, I've installed Linux Mint (Cinnamon edition).

While installing Linux and considering the alternatives (and avoiding my research project proposal), I thought about what would be available for Linux, and after having a quick exploration online I came across Kilo Code.

The individual edition is free and includes access to over 500 AI models as well as a number of features like code reviews, parallel agents and the bonus that they don't train using your code. Something that immediately struck me as being different is that there are multiple modes, the other options typically have an agent mode and a chat mode - from their documentation, Kilo Code has five key options:

• Code Mode: For general-purpose coding tasks

• Architect Mode: For planning and technical leadership

• Ask Mode: For answering questions and providing information

• Debug Mode: For systematic problem diagnosis

• Custom Modes: Create unlimited specialisedoptimisation personas for security auditing, performance optimization, documentation, or any other task

One of the main features I was dying to try out is the code reviews - one of the features is automated code reviews with AI-powered analysis, this feature is installed directly into your github repository so is slightly different to the IDE extension, but being able to AI sense check your code prior to merge is a fantastic step when you're developing something on your own.

After installation of the Kilo Code (KiloConnect) integration, you can enable AI Code Review and fully customise the review style as well as focus areas. Personally, I've set my review style to strict to highlight all potential issues, but also highlighted security, performance and bug detection as being the areas for special attention. There is an element of configuration for custom instructions, but that will develop over time and as I find more areas I wish to pinpoint and improve.

The actual agent element of the IDE extension isn't too different to the other IDEs I've tried so far - but it does have a much more descriptive output of changes both during the changes and after.

Watch this space! I'm going to continue to trial it and will update everyone with anything else, either positive or negative!