Switched from Copilot to Cursor for a React Native team - what we lost

Hi everyone. I'm still pretty new to the data engineering side of things, but my team recently switched our primary coding assistant from GitHub Copilot to Cursor. We're a React Native team, and while Cursor is amazing for some things, we've hit a specific, reproducible failure case when it comes to refactoring our data-fetching logic.

The issue appears when we ask Cursor to help us switch from a basic `fetch` to a more robust library like `axios` or `react-query`. It often hallucinates method signatures or misinterprets the structure of our state management. For example, I asked it to refactor this simple component:

```javascript
export default function DataScreen() {
const [data, setData] = useState(null);
const [loading, setLoading] = useState(true);

useEffect(() => {
fetch('https://api.example.com/items')
.then(response => response.json())
.then(json => {
setData(json);
setLoading(false);
});
}, []);

if (loading) return Loading...;
return {data?.title};
}
```

The prompt was: "Convert this component to use axios instead of fetch, with proper error handling."

Cursor produced code that used `axios.get` correctly but then introduced a non-existent `axios.transformResponse` call inside the component, which completely broke the render. It also placed the error state in a strange, invalid hook call.

The correct refactor shouldn't invent new methods. It should look something like this:

```javascript
import axios from 'axios';

export default function DataScreen() {
const [data, setData] = useState(null);
const [loading, setLoading] = useState(true);
const [error, setError] = useState(null);

useEffect(() => {
axios.get('https://api.example.com/items')
.then(response => {
setData(response.data);
setLoading(false);
})
.catch(err => {
setError(err);
setLoading(false);
});
}, []);

if (loading) return Loading...;
if (error) return Error fetching data.;
return {data?.title};
}
```

It's a small thing, but these little hallucinations make me nervous. I have to double-check every change it suggests, especially around APIs I'm less familiar with. Has anyone else found Cursor to be less reliable than Copilot for these kinds of straightforward, syntax-specific refactors? I'm worried about introducing subtle bugs into our data-fetching layer.