it has become the most effective way for me to solve problems. Most of the problems I encounter at work can be successfully solved by using agents. This is in contrast to solving tasks manually or writing the solution yourself.
In this article, I will give a high-level overview of how I approach problems and solve them using Claude's Code. As an engineer, you are essentially a problem solver. Your work should be like this:
- Identify and identify the most important problems to solve
- Come up with a solution to the problem
- Do it
And this doesn't just apply to editorial work. It also applies to jobs in marketing, sales, and customer service. I know this because this is what I do every day at work as a Series-A funded Data Science Lead.
Why you should troubleshoot with Claude Code
I always think it's important to know why you do something. If you take any engineering major in university, this is the mindset they will instill in you.
Always understand why for what you do
You should troubleshoot with Claude Code because it is usually the most effective way to solve the problem. Now, of course, you can use other methods of Claude Code, such as Cursor, although I will refer to Claude Code throughout this article, because it is the tool that I use.
However, the Claude Code does not only apply to step 3 of the problem-solving process (implementing the solution). It is also effective in finding and identifying problems, and how important they are. It is also very effective in coming up with solutions to the problems you have encountered.
You can, for example, find important problems by:
- Have Claude analyze the production logs and notify you of any issues
- Give Claude access to your CRM system so he can clean it up and let you know the most important areas to focus on
- Give Claude access to all your social media posts and performance, so he can analyze what works and what doesn't
After finding problems like the ones listed below, it's important to prioritize them. One of the simplest tools for this is the effort-value graph.
You simply plot all the problems you will solve on a 2D graph, with the axes being the amount you get from solving the problem and the effort required to solve the problem. After doing this, you simply select the problems with the highest value, relative to the effort required to solve them.
How to use Code Claude to find and solve problems
In this section, I will go through the three steps of the problem-solving process, which I highlighted earlier: identifying and prioritizing problems, coming up with a solution to the problem, and then acting on the solution.
To keep the article organized, I will have one subsection for each step of the troubleshooting process.
Step 1: Identify and prioritize problems
Earlier in the article, I highlighted some ways to find important problems with Claude Code.
I think the most important thing in finding important problems to solve are:
Give Claude's Code access to all the information you have access to
If you don't give your coding agents this access, you just can't expect them to work properly. Imagine if you had to solve a production bug without being allowed to view the logs. Of course, you wouldn't be able to solve it, because you can't understand where the problem is, without looking at the logs.
Therefore, make sure that your coding agent has access to all the necessary resources:
- Project management tools like Linear
- Notes from Notion
- Log in to groups on AWS
- GitHub to view the commit
- Browser access to reproduce problems
And probably many more platforms. The point is: if you use the platform to solve a problem, you have to give your coding agents access too.
Step 2: Come up with solutions to the problems
Okay, at this point, you've done the hardest part: You've identified an important problem that you need to solve. In software engineering, we often say:
Finding the bug, why it happened, and reproducing it is the hardest part. Solving it there is easy
So if you've come this far, you should be happy that you only have the easy part for now. Of course, how you find the solution depends a lot on the problem you're trying to solve.
If it's a production bug, you can usually tell Claude Code that:
Problem X occurs at location Y. I believe it is because of Z. View CloudWatch log groups †o understand why the problem occurred, and come up with a strategy to solve it
This is sufficient for more than 50% of the problems I work with, and works well for solving production bugs and implementing simple features. If the problem is more complicated, you usually need to iterate a few times through the program Claude Code creates for it, read it carefully, and make sure everything looks right. All of this ties in with the general practices and techniques I've discussed before to get the most out of Claude Code.
For some problems, you may need to tell the Claude Code differently, or ensure that it accesses the correct tools. If, for example, you are analyzing your CRM tool, you will need to provide API access to the tool and provide Claude Code with all the access it needs.
In general, I urge you to be liberal with the access you give your coding agent, because without proper access, they can't solve problems for you. This is exactly the same as asking people to solve a problem. If they don't have access to the right platforms and tools, you can't expect them to be good problem solvers.
However, you must ensure that the agent cannot perform any harmful actions. If you give it access to AWS, for example, you should sign in with View access only, instead of an admin user. In general, you can simply follow good security practices.
Step 3: Making solutions
The last part of the problem solving process is to create a solution. If you come up with a great idea with Claude Code, you can just tell it to run it in its program.
Again, this works for most of the problems I run Claude Code on. It was so good that it shoots one solution.
However, I still run into complex problems where it doesn't work. Sometimes you need to repeat several times with Claude Code. For example, if I create marketing materials like this:
- Transcripts of webinars
- PDF carousels for LinkedIn
- Posters
I often have the agent create an initial design and start revising and iterating on it. In general, this is a good method when working with complex problems. Come up with a decent first solution, test it, review what works and what doesn't, then repeat. By repeating this a few times, you will usually end up with a good solution.
The conclusion
In this article, I discussed how I use the Claude Code to solve problems. In general, my mentality is that whenever I encounter a new problem, I think: “How can I use Claude's Code to solve this problem for me. Many times, you will see that you can solve the problem with a simple acceleration. Sometimes, it takes a lot of repetition, or finding the right problems, prioritizing problems, creating a plan to solve the problem, or doing it, to motivate you to have a problem in general. a place, where you start with a decent solution, and repeat it until you are happy with it.
👉 My Free eBook and Webinar:
🚀 10x Your Engineering with LLMs (Free 3-Day Email Course)
📚 Get my free ebook Vision Language Models
💻 My webinar on Vision Language Models
👉 Find me on social media:
💌 Stack
🐦 X / Twitter
