We are currently in the AI bubble. AI is the new buzzword being thrown into every conversation lately.
We now have Deepseek, which seems to be better than ChatGPT. Multiple AI startups and evangelists have emerged, all claiming that AI is the future and will replace many existing jobs, including software developers. Being a software engineer myself, I was curious to know if AI is really ready to replace me.
First, I've been an avid user of ChatGPT, and I can attest that it has indeed improved my productivity — from brainstorming ideas to solving bugs. But it's not perfect. It still requires a lot of correction and refinement. This is where another popular buzzword comes in Prompt Engineering. Prompt engineering is the process of refining prompts to get the desired output. This is a tedious process and still requires human intervention.
I stumbled upon a startup called Lovable.dev, which claims to be a superhuman full-stack engineer. The process of using it is quite simple—just prompt it to create anything, and it will generate a Git repository with the code, which you can then add to your own GitHub account. It also has the ability to create a database on Supabase and connect it to your app. I decided to test it out by creating a simple chatbot app that:
- Allows users to log in via email providers and regular email/password authentication
- Uses the ChatGPT Assistant API for message responses
- Saves the message history for each user
Sounds simple, right?
I was impressed by how fast it created the app, but this didn't come without bugs and errors—though, surprisingly, it did attempt to self-correct. It took multiple prompts to get everything right, but still, the results were impressive. Even though the app was very lean and bare-bones, it was a good starting point for a POC or MVP.
Verdict
Who is this for?
This is for non-technical founders or PMs who want to quickly prototype an idea before committing an entire dev team or sprint to it.
Given the simple app I built, if I wanted to add more complex features, it would be faster to do it myself rather than continuously prompting the AI. Secondly, context is a huge factor in software development. AI does not have context regarding business requirements or users’ needs that are not explicitly documented—especially those that are implicit or inferred. It can only do what it's told, and this is where the human touch comes in. For example, designing the data layer isn’t always straightforward.
Thirdly, most developers work on existing codebases, which requires a deep understanding of the architecture before making changes. Even though some AI agents can explain code, you still need context to understand how everything fits together to meet user needs.
Conclusion
AI is not ready to replace software developers, but it can significantly speed up the development process.
Top comments (0)