Augment Code

Context-aware suggestions are generally very accurate, which is important given our large existing code base. This was a challenge using a competitor product. Slack chatbot: this has been a huge added benefit for the entire company. Excellent support: the team is very nimble and responsive and it helps having direct access to the team through a shared slack channel to escalate any issues, ask questions, or share learnings.

It's incredibly helpful to have large context blocks about our code instead of smaller amounts so when we ask questions or iterate, or set rules for how the agents work we can give them more helpful context that they can run with more consistently than other products.

It's very fast, very accurate, easy to use and if you know even basic terminologies of coding you can make apps of your own. No coding knowledge is required.

Like any AI assisted coding it requires vigilance and you can't put on autopilot. The tool could also benfit from absorbing more context through connections with ticketing, documentat, and slack tools. This is possible through Agents but might not be possible with the slack bot yet. Ideally we have a slackbot to rule them all that connects with these other sources.

When the agent feature starts changing how my code works and is structured because it missed how to fix an issue it frustrates the crap out of me, because it just abandons doing it the right way. An example was that when I asked it to troubleshoot my testing, the result it came up with was to just disable all my tests to get the pipeline to pass. The critiera that it ends up using for success don't align with real world success criteria.

Only thing is that I have to do QA every time once it does the coding. It sometimes gives wrong output but it completely depends on your prompts.