Hey everyone! 👋 I'm diving into Kling for the first time this week and I want to make my evaluation structured and useful. I know it's easy to get lost just asking it random questions, so I'm trying to focus on what matters for actually integrating it into a dev workflow.
From my experience with other AI coding assistants, I think these three areas give you the best signal on whether a tool will stick:
1. **Context Handling & "Project Awareness"**
This is huge. Can it understand your project's structure beyond a single file? Test by:
* Asking it to explain relationships between modules in your codebase.
* Having it refactor a function, but tell it to follow the existing project's naming conventions and patterns (e.g., "Make this async like the other data fetchers in `utils/`").
* See if it can correctly reference imports from your own internal packages.
2. **Code Review & Improvement Suggestions**
Don't just ask for a review on perfect code. Give it some subtly flawed code and see if it catches the important stuff. I like to test with a Python snippet like this:
```python
def process_items(items):
"""Process a list of items."""
result = []
for i in range(len(items)):
item = items[i]
# Simulate some processing
if item.status: # Potential None check?
result.append(item.data * 2) # What if data isn't a number?
return result
```
Does it catch the potential `TypeError`, the possible `None` status, the inefficient indexing, and suggest better patterns (like a list comprehension or more specific type handling)? The depth here is key.
3. **Testing & Debugging Workflow**
Can it help you go from bug to fix? Try:
* **Test Generation:** Give it a function and ask for comprehensive pytest cases, including edge cases.
* **Debugging:** Provide an error traceback (real or fabricated) and a chunk of relevant code. Does it pinpoint the likely cause and suggest a concrete fix, or just give generic advice?
* **Explanation:** Ask it to explain a complex piece of code from your domain. Does it just paraphrase, or does it connect concepts meaningfully?
What did you all test when you first tried Kling? Any other "must-try" scenarios I'm missing? I'll report back with my findings!
Happy coding!
Clean code, happy life