Want to ship faster? Build a faster horse 🧐

Wrong Answers Only: I wanted to analyze my tweets so, I started by using AI to write a python script.

How I Could Have Used AI to Rethink My Coding Process From the Start

X has an option to export your archive. This is a rich amount of data, as I’ve been on Twitter/X for over a decade with more than 80K posts. I downloaded my archive only to find that my posts and data were wrapped in javascript. My old programmer’s logic told me I needed to write a script to strip away the javascript to make use of the data. I started running down this path without thinking. After the fact, my friend Matt Wallace highlighted my failed logic.

Here’s a lesson of what I did vs. what I should have done in a world where we have AI to not only make horses faster but rethink how we ship faster.

---

What I Did: Sticking to Old Habits

Here’s the process I followed. It got the job done but I’d struggle to describe this process to a non-technical user:

1. Downloaded and installed VSCode
   It’s a great tool for developers, but for someone who hasn’t coded in a while—or for non-coders—it can be overwhelming.

2. Signed up for a code assistant
   I chose an AI-powered tool like GitHub Copilot, hoping it would do the heavy lifting. Not to be confused with Microsoft Copilot, GitHub Copilot is very much geared toward developers.

3. Enable and learn the assistant in VSCode
   I had to help a friend enable the code assistant. I’ve gotten plenty of value over the past few days. My non-technical friend is still finding their way around VSCode + Copilot

4. Created the script
   I played around with Copilot and had a few different applications running in minutes.

5. Debugged the logic of the code
   While the AI was great at writing code, I still had to manually troubleshoot logic errors—this is where frustration really set in. I approached the problem from the wrong direction. What I initially thought was a file size problem turned into a format problem. The code assistant made assumptions about the dataset that I had to manually troubleshoot.

After all that work, the final output was an ugly command-line interface—far from the polished, modern UX users are accustomed to.

---

What I Could Have Done Differently: Rethinking with AI from the Start

Instead of defaulting to old habits, I should have approached the process with fresh eyes and asked a different question:

"How can AI help reshape my workflow rather than just assist my existing process?"

If I had let AI guide me from the beginning, here’s how things might have been different:

1. Start with the Problem, Not the Tools

Rather than diving into VSCode, I could have started by defining the problem I was trying to solve. Using an LLM, I could have asked for advice on the simplest approach to achieving my goal. For example:

• "How do I build a simple user interface for this project?"

• "What tools are best for someone who hasn’t coded in years?"

An AI could have suggested solutions that didn’t involve setting up complex environments, saving me hours right from the start.

2. Explore No-Code and Low-Code Platforms

Instead of getting stuck in VSCode, I could have explored no-code or low-code platforms that integrate AI. These tools, like Bubble or Glide, allow non-developers to build applications with a user-friendly interface—without needing to write complex code. An LLM could have pointed me in this direction early on, helping me create something with a modern UX from the start.

3. Use AI for Brainstorming and Planning

One of my biggest mistakes was not involving AI in the brainstorming phase. Had I asked the AI for help earlier, it could have recommended:

• The best tools for my specific needs

• Step-by-step guides to make the process smoother

• Alternatives to traditional debugging methods, saving me from unnecessary frustration

LLMs could have guided me through the learning curve more efficiently, offering tailored advice rather than making me figure things out alone.

4. Leverage Open Source and Collaboration

Another missed opportunity was not considering the open-source community. Had I shared my process sooner, others could have built on it or even helped optimize it. Modern AI tools and platforms enable collaborative development, allowing both coders and non-coders to contribute. By open-sourcing my workflow, I could have made it easier for non-technical users to leverage AI, while benefiting from community feedback.

---

Moving Forward: Letting AI Guide the Process

Looking back, it’s clear that AI has the potential to do more than just assist with coding—it can advise on reshaping the entire workflow. Instead of building a faster horse, we can use AI to:

• Rethink how we approach problems before jumping into tools

• Utilize no-code and low-code platforms to simplify development

• Collaborate with AI from the start to create more efficient and intuitive solutions

Had I embraced this mindset from the beginning, I would have saved myself hours of effort and frustration. Instead of an ugly command-line output, I could have built a process I could share with other X.com users to analyze their post data.

---

Conclusion

AI is transforming the way we approach problems, but only if we let it. By allowing AI to advise us, we can rethink outdated workflows and create smarter, faster, and more accessible tools. Whether you're a seasoned developer or just getting started, there’s enormous potential in using AI not just as a tool, but as a collaborator. Next time, I’ll let AI help me re-think the process.