Review: testing devin.ai in 2025

I've been using Devin for a couple of days now and am dumbstruck. I had read a few reviews about this new AI developer online but nothing could prepare me for this.

In two days, I've been able to develop and deploy a complete web app with a server written in Python and a frontend written in React. In this post, I will describe my experience working with Devin and my intuitions as to what it means for the future of software development.

As a side note, the Devin team must have done an incredible amount of work that will be useless in just a few months as AI get's better. Still thank you Devin team for your work!

Working with Devin is incredible

I've worked as a product manager for years and working with Devin is exactly like working with a developer. It is no god though: if I describe something poorly, it won't work but as long as I'm clear enough (as any good PM needs to be), it will complete the task, test it and once it's confident it has completed the task, submit the PR.

This prompt for example is enough to build a complete logout feature for you:

The team who developed Devin must have spent an incredible amount of time on pipeline/sewage code as Devin is able to access everything from the IDE, Github, terminal, UI and more.

You can ask it to build a zip for your Chrome extension, you can ask it to test every screen of your app and it will login, click around and test features, you can provide visuals or designs and it's able to look at them and implement them (this is miles away from what the Replit Agent is able to do for example).

Test: Building a simple knowledge base software

I've been looking for a knowledge base / help center kind of tool for one of my products. My requirements where:

• Should be simple to use
• Should be multilingual
• Should cost around 30/mo
  .. and I was surprised to discover that it doesn't exist. That most help center software is super complex to use (for example, you'll have to create a separate article for each language instead of just clicking a translate button (or copy pasting the text into DeepL) and rather expensive (70-100/mo is the average price I found).

So I decided to build it over Christmas using Devin. It is worth noting that I have some backend development experience and was able to write the app.py Flask server myself (or rather let ChatGPT PRO do it under my supervision).

I then passed the api documentation to Devin and an empty repo and asked it to build the app for me:

And 20 minutes later, I received a notification with a link to test the newly deployed app. it was working. With a few more prompts, I added support for multiple locales and improved the UI a little:

The only difficulty

The only difficulty I encountered (aka. where Devin got stuck) was when there was a CSS clash between the menu bar and the main content of the page.

But once again, I solved it using AI by simply copy pasting the main files containing the ArticlesList & Menu component into ChatGPT and asking the AI if it saw any issues and explaining that the page was not displaying correctly.

ChatGPT o1-pro refactored the files, I copy pasted them back into my IDE and the problem was solved.

Do we still need developers?

For the time being yes. As of January 2025, you need some basic coding knowledge to be able to work effectively with Devin & ChatGPT as you still need to know what to ask for and still need some intuition as to what might be going on in the case Devin get's stuck.

Yet, it seems very doubtful to me that this will be the case by the end of 2025. And for any product manager, it is now possible to build any app without ever coding yourself or hiring a developer. Startup time!

What does it mean for the future of the software industry?

There is no way we'll need more than 1% of all the current developers by the end of 2025. At my own company, I'm redirecting developer efforts to the very complex and wholistic tasks that Devin can't do yet (like refactoring an mp3 encoder), but all the low-story-point tasks will be done by Devin from now on.

If anyone can build their complete Saas for their specific use case in a couple of days (by the way, if you're looking for a simple knowledge base software, try out simplestdocs.com), will this lead to many more people building Saas and selling them, or will it instead lead to everyone just building their own things and no one buying from external vendors anymore?

For sure, the cost of software will go down drastically over the coming years. My experience with Devin has seriously set me into some sort of tech existential crisis that I yet have to come out of. I hope this post will inspire you to try out Devin for yourself and realise what's going on.