Still seeing the hype train pull into the station. Everyone's jumping from Cursor to Windsurf like it's going to solve their actual problems.
For me, the hardest part was unlearning the muscle memory for the agent prompts. Cursor's `@` workspaces felt more contained. Windsurf just throws everything into the main chat context by default. Now my terminal is clogged with AI-generated YAML that's more verbose than my Ansible playbooks.
```bash
# My typical Windsurf detritus after 10 minutes
# [AI] I've generated a draft Kubernetes manifest for your nginx deployment...
# [AI] Considering your request, here is an optimized Dockerfile...
# [AI] The monitoring setup you asked for...
```
Half of this "assistance" is for tooling my team doesn't even use. We're not at Google scale. We don't need a five-service mesh defined every time I tweak a config. The hardest habit to break is expecting the tool to understand what's actually necessary versus what's just the current trend.
Keep it simple
The "detritus" you mentioned is so real. I started trying to use it to configure a simple SSO flow and got a full OIDC implementation plan with three different auth providers, none of which we license. 😅
Do you find yourself fighting the "helpfulness" more than actually getting work done? I'm spending more time pruning suggestions than writing code.
I'm not even touching the Kubernetes side of things yet. That sounds stressful.
Yeah, the context spill is rough. I miss Cursor's compartmentalization too.
Do you have to manually clear the chat history often, or is there a config trick to stop it from aggregating everything? I'm new to this and already overwhelmed by the log noise.
It sounds like the tool guesses your stack based on open tabs, but that's clearly backfiring. Makes me hesitant to even open a config file.