Cognition Lab’s Devin AI: A New Dawn for Software Development?

Wouldn’t it be amazing to have a coding buddy who could work tirelessly, pump out perfect code, and maybe even help you focus on the creative aspects while you grab a real coffee break? Well, buckle up, because Cognition Labs might have just invented your new best friend: Devin AI.

In their recent announcement on X, Cognition Labs demoed their futuristic AI Software Engineer, Devin that can handle the entire software creation process, from that first spark of an idea to actually launching the finished product. The demo came as a shock for many as people are now scared that this could mean the end of the millions of jobs that only humans could do earlier – well until now.

Is Devin here to replace programmers, or is it more like a super-powered teammate who can help us code like rockstars? We’ll crack open Devin’s skillset in the next sections and see how it might change the way we build software in the future.

A Peek Inside Devin’s Workspace

At the outset, the UI looks pretty clean and straight forward. Theres a section to chat with the AI agent and an integrated Workspace section on the right where all the magic happens.

The Workspace is further divided into four section viz. a Shell, a Browser, an Editor and a Planner. 

The user initiates the conversation by asking Devin to do a real life programming task in natural language. The AI agent then prepares a step-by-step plan of action before starting the actual implementation of the task.

Once the plan is made, it builds a whole project using the same tools a human software engineer would use.

Devin even can use its own integrated Browser to pull out API documentations or other resources as required for the task at hand. It then implements the code and finally outputs the end result.

While the specifics remain under wraps, Devin AI’s interface seems geared towards a user-friendly experience that minimizes traditional coding and allows developers to focus on the bigger picture.

Decoding Devin’s Skills: A Sneak Peek into AI-Powered Coding

While we might not all have personal access to Devin AI yet, other developers have been putting it through its paces. Let’s eavesdrop on their experiences to see what Devin can actually do in the real world.

Devin can pickup jobs from Upwork and deliver

In one of the videos they shared on X, a team member from Cognition Labs demonstrated how Devin handles a task picked directly from Upwork, an online freelancer job portal where software jobs are posted on a daily basis. Here’s what the client from Upwork wanted to achieve:

In laymen’s term, the job is to build a computer vision model that can look at photos and tell if the road is broken.

Once the executive posted the task in Devin, it immediately kicked off and fetched the git repo, setup and tried to run it locally (here locally means, wherever in the cloud Devin’s AI agent is running that piece of code). Occasionally it encountered errors which it debugged using simple print statements and moved past the issue by narrowing down the problem like a human counterpart would do.

For example, in the demo, Devin got stuck at some point and using its debugging skills could identify that it needs to update some of its dependencies to make the code work; so it updated the requirements.txt file at runtime. Although a simple trick to perform, the autonomous nature of its execution is quite fascinating to see, to be honest.

One more interesting feature that I found was the Chronological Time Slider (or whatever they call the blue button at the bottom) of the Workspace which you could slide from left to right to see all the historic steps taken by Devin to finally reach the end solution. It’s a nice feature to have in an end-to-end AI Engineer like Devin.

In the Demo, the AI agent continues the process of running code and fixing bugs over and over again until it fixes all the problem. It ran across all the images given by the client through the vision model it built and finally dumped a report at the end.

It was one of the coolest demo of an autonomous ai in the software development field so far. Artificial Intelligence has really come very far!

Devin can train other AI Models

In another demo from Cognition Labs, a developer demonstrates how Devin can also train another AI Model. They use QLoRA which is used to fine tune LLMs. The workflow was pretty smooth and all the developer had to do is provide Devin with the repository link to initiate processing. The prompt was pretty simple too – “Can you fine-tune a 7B Llama model using https://github.com/artidoro/qlora?“

Devin can setup development environments

In yet another demo, an independent developer posted on X his workflow of setting up a chatbot UI using Devin. He just let Devin pull his repo from GitHub and do its job; took the agent 27 minutes to run but it successfully did what was expected of it.

We’re seeing more and more demos of Devin AI in action every day, helping programmers with their daily tasks. I’ll share some cool ones later, but for now, it’s clear – Devin-like agents are here to stay.

On X, AI researcher Andrej Karpathy commented on Devin AI, saying : “Devin is an impressive demo of what perhaps follows next: coordinating a number of tools that a developer needs to string together to write code: a Terminal, a Browser, a Code editor, etc., and human oversight that moves to increasingly higher level of abstraction. There is a lot of work not just on the AI part but also the UI/UX part. How does a human provide oversight? What are they looking at? How do they nudge the AI down a different path? How do they debug what went wrong? It is very likely that we will have to change up the code editor, substantially. In any case, software engineering is on track to change substantially. And it will look a lot more like supervising the automation, while pitching in high-level commands, ideas or progression strategies, in English. Good luck to the team!“

How to get access to Devin AI?

At the moment Devin is in closed Beta where access is restricted to a limited group of testers or potential early adopters. If you are a Developer and would like to try Devin you could reach out to Cognition Labs via their website or directly on X.

The Future of Coding: AI Partner or Terminator?

Devin AI’s arrival throws a wrench into the software development world. The question burns: Will AI like Devin eventually render human programmers jobless?

Let’s face it, AI agents like Devin can automate many tasks currently handled by entry-level or mid-level coders. Think writing boilerplate code or squashing minor bugs – those might become increasingly automated, potentially leading to job losses in those areas. But before you dust off your resume for a career in underwater basket weaving, consider this: a complete AI takeover seems far-fetched. Here’s why:

• The Creativity Factor: Software development isn’t just about churning out lines of code – it’s about tackling complex problems and crafting innovative solutions. These are areas where human creativity and problem-solving skills remain irreplaceable.
• New Skills Emerge, Old Ones Adapt: While some skills might become less in demand, entirely new skillsets will emerge. Developers will need to learn how to collaborate effectively with AI tools and translate their ideas into a format AI understands.
• The Human Touch Endures: Understanding user needs, designing intuitive interfaces, and navigating the complexities of human interaction will remain crucial for successful software development. In these areas, AI can be a powerful tool, but it can’t replace human judgment.

The Future is Collaboration, Not Competition:

The future of coding is likely a collaborative one, where humans and AI work together to create even more incredible software. Devin AI, and the inevitable wave of similar tools on the horizon, might not be stealing your job – they might be transforming it.

Here’s how developers can prepare for this collaborative future:

• Lifelong Learning is Key: The ability to learn and adapt to new technologies will be critical. Developers who stay up-to-date with the latest advancements in AI and coding will have a significant advantage.
• Collaboration is King: Effective communication and collaboration with AI teammates will be crucial.
• Embrace Your Humanity: With AI handling the routine tasks, developers can dedicate their time to the unique human strengths of creativity and problem-solving.

The Bottom Line:

While entry-level coding positions might face challenges due to automation, the software development field as a whole isn’t going extinct. The key will be to embrace lifelong learning, develop new skillsets, and leverage AI as a powerful tool to become a more well-rounded developer in this evolving landscape. After all, Devin’s success is likely just the first ripple in a wave of innovation. Get ready to see a surge of similar AI assistants, each with its own strengths, further empowering developers and accelerating the pace of software development. The future of coding is bright, and it’s a future built on collaboration, not competition.

You may also be interested in Phind AI. In another article, I take a closer look at this intelligent answer engine for developers. It’s a great resource for anyone interested in streamlining their coding workflow.