tl; drag: The cloud code is noteworthy but it needs to be learned deliberately. I have used it for Azar’s development, 3D design editing, documents and even this article – alternate alternatives to the board. Top 8 points:
- Think of it as a very fast internal with a fine memory, but not much experience.
- When it makes mistakes, ask him to update the cloud. MD don’t make them again.
- Ask him to do the work that can improve the quality of the code that you will not have for the first time.
- Trust it with just reading commands.
- Use it for the entire software life cycle.
- Use the file system for your benefit.
- Keep an eye on the price, but still worth it.
- Try archive mode for more complex tasks.
In the past few days, I have been on a journey to learn how to work with Claude, put it in different scenarios at my own pace: traditional codes, editing 3D designs, writing documents and even writing this article. In all cases, it was better than anything I had tried earlier. Basically not better – basically better.
Remember the feeling of being empowered when you feel relieved from a new framework or device first? Is it possible for a sudden extension, such as when I first learned Open GL or Java? Provides cloud code. The feeling tucks you: “Wait, I can They Now? “I probably realized this 3 or 4 times in my career. I still had the fifth.
The following will look hyperbolic, but those who know me know that I have not been given to Hyperbal. One of the favorite references is that “the future is already here, not divided at all.” Cloud code is the future of programming. I have no doubt about it, because I never want to go back to the old ways of working, and already, the editing of the code feels like a step backwards. I have canceled my cursor and wind surf membership.
However, there was definitely a learning curve. The first few efforts were a lot, but over time, the cloud code and I learned how to work together. Another favorite reference to me is “Experience is a great school. Unfortunately, the fees are high. Let me share the benefits of my experience so you don’t need to pay a high fee.
Here is a mental model that has helped me: Think of the cloud code as a very fast internal with perfect memory. They are anxious to help, incredibly worth it, and remember everything – but still needed clear direction and occasional supervision. In view of the right instructions, it can operate at the internal transparent pace while you focus on creative parts that really require your skills. But like any new member of the team, their work needs to be reviewed. Despite its intelligence, always check every line. The cloud is impressively accurate, but validation is necessary.
Like the interns, they often solve problems in an overwhelming way. And if you let these changes go, your code will be more complicated.
For example, when I was working on a 3D design with the open scade, I asked to score the edges of the cylinder. It has come up with this highly criminal approach to making up and down. I pointed out that there was a parameter for round cylinders that he could use from the library, and could only be used in it.
Claud code a cloud. MD creates a background knowledge that it uses for the overall project. If the cloud does something that you don’t like, don’t just correct it once – ask it to update the Claude Dot MD file so that it can be missed next time. This seemingly small step changes your experience. This is the difference between explaining the same preference and having a device that actually adapts to you.
Example: After the above mentioned round cylinders, I asked him, “How can you edit Claude Dot MD to stop this problem in the future?” He proposed an amendment stating that he should first examine the functions in the library. He did not do it again.
As a developer, there are always toddoes that we never really do, either because we are slow or we don’t have time. The cloud is actually very good in them.
Ask him:
- Update your documents (seriously, when did you really want to do this?)
- Write a comprehensive test (even edge cases that you usually leave)
- Add clear comments on complex function calls
- Improve code quality and read qualification usually
The cloud is not just to write the code – it is very good to explain clear PR and produce meaningful comedy messages. In fact, its covenants are better than me because of the “perfect memory” thing. I might forget to add something, but that’s not. Promoting the quality of these frequently unconscious communication pushes mutual support for everyone involved. It has PR specifications Really Nice
Even I used it for aerogram in the design doctor. While working on the documents, I asked to prepare a tear using the graphways (dot). Then I had a shell script that converts .Dot files into .pngs and puts them in a separate directory. Finally, he added these sketches to the design doctor. The results were easily accurate. I have a “how? … what?… Really?” The reaction was very good because he read the whole design doctor and understood the relationship.
Another example, I made my gut rapo a little mess with integration and little ones. Instead of fixing it myself, I asked the cloud to handle it. His understanding of the gut was better than myself, and what would I have to know about it, he did it in minutes.
Every time the cloud code processes a Bash Command, it asks you whether you should ask again about this order. My Principle: Only on the read -type instructions, grant only the “auto executive”. “Yes” to “Gut status”. “No” to “Got Commitment”. “Yes” to “LS”. “No” to “RM”.
I have a caution story: I allowed the cloud to gut, and he enthusiastically added all the files in a folder with “Gut Aid”. – including some files with sensitive information. I had to pull the plug immediately, but I was very slow at the escape key.
Of course, I then asked to help eliminate sensitive information from the date of the gut (since it’s not enough to delete files from Ripo), which I might have been wrong myself.
Think about the file system as a way to work for you and the Claude. When you do, anything inside the working directory is useful. Here are some examples of how I used the file system:
- When working on 3D modeling, I pulled the source into the directory for one of the big open scade library (BOSL2). The Claude could then enter through these files, find them and learn from the existing code.
- For the aforementioned sketches, I not only saw the Claude dot files, but also examined the PNG files created by them to improve the concept.
- In the preparation of this article, I wanted to better understand my style, so I created an example/ directory and saved some of my writing there. Then I asked to read these examples and summarize my style in the Clened Dot MD file. When I asked to rewrite this article in my style, it came a voice Lot More like me can I really predict?
The cloud cost model is different from the cursor or wind surf: you pay through the token (including thinking token, etc.). This is not expensive, but it is possible to use the To 10 tokens a day, or $ 200 seriously in a month, when the cursor and wind surf are probably a tenth of it. Nevertheless, it is good and really worth it in promoting performance. For example, all the work on this article was less than a dollar (here is a /cost command you can use to work on how much it costs you).
There is a command line option for any major reflecting work that “enable the Factor.” This includes a new tool that enhances the ability to cope with the larger codecs of the cloud code. It’s not well documentary, but I found out that it definitely helps.
To me, the cloud code agent system and the LLM together have the promise to fulfill the promise. I think this will be a new way of programming. Whenever I go back to using the cloud code, I feel like I lost a key tool in my toll chain.
The benefits of production capacity are significant, but the real value comes from the mental bandout, which is free for the creative aspects of coding. The learning curve is capable of climbing because on the other hand there is a real extension you can perform in one day.
