Anyone actually using an agentic coding IDE for real projects?

[Egglex]

I've been seeing a lot of talk about agentic coding IDEs lately, where the AI doesn't just autocomplete code but actually plans tasks, edits files, runs tests, fixes issues, and keeps iterating on its own. The demos look impressive, but I'm curious how well this works on real projects once you get past the landing page videos. Are any of you using one day-to-day, and if so, which tools have been the most reliable?

 

[Lovol]

Most of the agentic stuff still falls apart once the codebase gets bigger than a toy demo. They handle small isolated tasks fine, but multi-file changes with real dependencies tend to confuse them, and you end up babysitting the diffs anyway. The reliability swings a lot depending on how clean your test suite is too, since these tools lean hard on tests to verify their own work.

 

[Thirteenth]

True enough, the bigger the project gets, the more those cracks start showing during longer autonomous runs. Context tends to drift after a string of edits and it starts losing track of why it made earlier changes. There's an agentic coding IDE worth checking out here https://tetrees.ai/?utm_source=upv . It’s built around reusable code blocks and an AI matchmaker that helps assemble a setup from existing components, plus a critic agent that verifies changes against actual test output before they get applied. The idea is to reduce blind generation and rely more on verified building blocks and feedback from the codebase itself.