Hi everyone, Brilian here. Today, I'm writing differently than usual. Instead of a deep dive into technology, I want to share how I've been using AI—and I'm also curious how you’ve been using it.
I feel that over the past few months, AI has progressed very quickly. Not necessarily in terms of accuracy, but definitely in terms of usability. There are a lot of products that help me develop things now, especially around agentic coding.
Building Products Within 15 Minutes - Vibe Coding
Have you ever built and shipped an app in 15 minutes—without touching a single line of code? I did.
I've recently explored agentic coding tools like Cursor, Windsurf, and the OpenAI codex, and I’m honestly really impressed. For those who don't know, agentic coding refers to AI systems that can independently generate, test, and even debug code based on prompts—almost like having a junior engineer on call. I’ve built some services that turn articles into tweets and schedule them to post on Twitter, almost without writing any code—even the UI! (even though I have little experience with UI design). I used bolt.new to create the prototype, downloaded it, and asked Windsurf to add the backend calls by giving it the API spec (also generated by Windsurf).
What really impresses me about agentic coding is that it not only generates the code—it can also debug it if there's a syntax error. It can even analyze logs when something unexpected happens while running the app. This feature really allows people to ship products without knowing much about software engineering.
For example, I created an article summarizer API using Python in just 15 minutes by prompting Windsurf:
Create an article summarizer project using Python and FastAPI. There should be a single API endpoint that accepts a URL and returns the summarized content of the article from that URL. Use the OpenAI API for the summarization. Ensure the project has a clean and well-organized structure.
It generated a full Python project for me—complete with a README!
When I asked it to run the project, it noticed I hadn’t provided my OpenAI API key and asked me to enter it.
After I did and asked it to try summarizing one of my blog articles, it tried a few things
...and in the end, it worked perfectly.
How awesome is that? I can now ship simple products within 15 minutes—without actually coding!
How AI Changed the Way I Research
Another area where AI has changed my workflow is in search and research.
Previously, I'd go to search engines like Google or Kagi and visit multiple website to understand a topic. Nowadays, I mostly use Perplexity and ChatGPT to research and summarize the result for me.
This saves me a ton of time. What used to take hours now only takes minutes. These LLM-based search tools read multiple sources, scrape the content, and generate well-organized summaries. If I have a follow-up question, I can just ask—and they’ll find the answer for me. I’ve also noticed that hallucinations (incorrect or made-up info) are far less common these days, especially with Perplexity.
LLM-powered search has seriously changed how I work. I now use it more often than traditional search engines. I use it to compare products (like bags I’m considering), and even to help outline my blog posts.
Is AI Replacing Software Engineer?
So, after fiddling with agentic coding, do I think software engineers will be replaced by AI anytime soon? Honestly, I don’t think so—but I also wouldn’t be too sure. AI is progressing really fast, and many companies like Google, OpenAI, Cursor, and Windsurf are investing heavily in agentic coding capabilities.
That said, even though I can now ship products with minimal coding, I don’t think AI is there yet. When building more complex apps, I still have to guide it a lot. Sometimes, it fails to fix bugs even after multiple prompts. Other times, it breaks existing features when introducing new ones. These issues become more frequent as the codebase gets more complex.
Still, I believe the rise of agentic coding and LLMs like ChatGPT help us on becoming more productive as a software engineer. They’re great for prototyping product ideas quickly which we can build on. They can also help with writing unit tests, generating documentation, explaining code, and even implementing or adjusting simple features.
In fact, many companies already expect their engineers to learn how to use AI effectively—whether by building AI-powered tools or by using AI directly to speed up feature development.
Another thing I'm thinking is that software engineers will be expected to understand system design more deeply. AI is getting better at implementation, but someone still needs to tie everything together, review the results, and ensure quality.
All these changes got me thinking—if AI is already transforming how we code and research, what does this mean for the future of the developer ecosystem?
People Aren’t Visiting Websites as Much
This hits close to home since you're reading this on my blog. People don’t visit individual websites as much anymore when researching. They just use LLM search tools to get the answers directly. Even major platforms like Stack Overflow have seen declining traffic.
People Are Depending More on Code Generation
Previously, when we needed to write code, we either did it manually or copy-pasted snippets from websites like Stack Overflow. But even then, we usually tried to understand what the code did and how to adapt it to best practices.
Today, agentic coding and LLMs can generate entire code blocks that “just work”—and people often use them without fully understanding what’s going on. This shift brings speed, but also raises questions:
What are we losing in the process?
Less Original Content Being Written
Because so many people use LLMs for coding and research, fewer are creating original content—whether that’s blog posts or open-source code. But if fewer developers share new knowledge, and fewer websites publish fresh articles, where will LLMs get their training data in the future?
It’s a bit of a paradox: AI depends on human-generated content, yet it’s also changing the incentives to create that content in the first place.
Conclusion
AI has progressed a lot, and it's clearly something every software engineer—and really, everyone—should start learning how to use. We don’t yet know exactly how it will reshape the future, but it’s already improving productivity in a big way.
If you haven’t tried integrating AI into your workflow yet, now’s the time.
What About You?
How’s your journey with AI going so far? What tools do you use the most? Are there any AI-powered services you’re subscribed to?
Drop a comment—I’d love to hear your thoughts and learn from your experience!
