One of the emerging applied AI labs named “Cognition” has developed Devin – the world’s first fully autonomous AI software engineer.
Just like a human software engineer, it can plan an entire project, write, execute, and test complex code, take freelancing jobs from freelancer websites, and also deploy code on its own.
Table of Contents
What is Devin AI?
As per the official statement from Cognition–labs:
“Devin is a tireless, skilled teammate, equally ready to build alongside you or independently complete tasks for you to review.”
In simple terms, it is like a software engineer who can plan an entire project, write, execute, and test complex code, take freelancing jobs from freelancer websites, and also deploy code on its own.
It has the potential to execute software engineering tasks independently and make the end-to-end software development process faster and smarter.
How does Devin work?
It owns three parts – a command line, a code editor, and a web browser. With the help of these three resources, it analyzes the given task, makes an execution plan, writes the code, and executes code.
Scott from cognition-labs AI has shown an example of how it works. He asked Devin (through prompt) to Benchmark the performance of Llama and a couple of different API providers.
Following are the steps/tasks it performed following the above prompt:
- First, it analyzes the requirements and makes a step-by-step plan on how to tackle the problem.
- It builds the whole project using the required tools that a human software engineer would use.
- Next, it decides to use the browser to pull up API documentation so that it can read up and learn how to plug into each of these API’s.
- It runs the code but gets an unexpected error while running the code.
- It adds a debugging print statement to check for the error.
- It re-runs the code in debugging mode and then uses the error in the logs to figure out how to fix the bug.
- Finally, it decides to build and deploy a website with full styling as the visualization.
Following is the generated website image:
Capabilities/Features of Devin
It can analyze, plan, and execute complex engineering tasks and make several important decisions like a human software engineer performs.
Following are some of the capabilities of it as described on the Cognition official website:
1. End-to-end application development
It can develop the applications as per the requirements and incrementally add features requested by the user. It can write code in different languages like Java, Python, C# and Javascript.
Following is the demo video where Devin makes an interactive website that simulates the Game of Life!
2. Finding and fixing bugs in codebases
It can find and fix the bugs in the application codebase, debug the code, test the fixed issue, and re-ran the application after bug fixing.
3. Continuous learning
One of the key features of Devin is continuous learning. It has the capability to learn new skills, upgrade itself, learn from different resources, refine code writing skills, etc.
4. Train and fine-tune its own AI models
It can train and fine-tune its own AI models to perform specific tasks within a project. For example, it can set up fine-tuning for a large language model by providing a link to a research repository on GitHub.
5. Requirements gathering and environment set-up
It has the ability to set up an environment and gather the requirements. For example, by providing an open-source repository link, it can set up the environment and perform the required context gathering.
6. Take and perform freelancing jobs
It can take freelancing jobs from freelancer websites like Upwork, freelancer, etc., and write and execute code to meet the job requirements.
7. Code deployment
After completing the code for a specific requirement, it can deploy the code on its own and make sure the code works fine for the different users.
How to get access to Devin?
As per Cognition-labs, it is currently in its early stage and access and you need to join the waiting list by filling out a form here or get in touch at info@cognition-labs.com.
Devin’s Performance
Cognition-labs has evaluated its performance on SWE-bench, a challenging benchmark that asks agents to resolve real-world GitHub issues found in open-source projects like Django and scikit-learn.
Following was the result published by the cognition labs on their official website:
“Devin correctly resolves 13.86%* of the issues end-to-end, far exceeding the previous state-of-the-art of 1.96%. Even when given the exact files to edit, the best previous models can only resolve 4.80% of issues.“
*Devin was evaluated on a random 25% subset of the dataset. It was unassisted, whereas all other models were assisted (meaning the model was told exactly which files need to be edited).
Future Outcome
Currently, Devin is in the early stages and is still learning. But with the speed at which AI is evolving daily, and with the continuous learning feature, it will become more mature and could start doing the software engineer tasks that software engineers do.
Also, more companies will develop similar tools in the future that can perform the following software engineering tasks:
- Analyze project requirements and plan the tasks.
- Write and execute the code to achieve the desirable output.
- Test the application and look for the potential bugs.
- Debug the code and bug fixing.
- Re-run the application and verify the fixed bugs.
- Code deployment on its own.
Will Devin replace Software Engineers?
Looking at its capabilities, one can fear that it might replace software engineers or impact software engineering jobs in the future. But it is in an early stage and without testing and using it properly, we can’t conclude whether it will replace software engineers or not.
But like the other AI tools like ChatGPT, Gemini, etc., it will assist human software engineers to perform complex tasks faster and thus increase their productivity.
Conclusion
Devin AI is the world’s first fully autonomous AI software engineer and has different capabilities to perform complex software engineering tasks. It can plan, write, execute, test the code, and also deploy it on its own.
Looking at its capabilities, it feels like it will replace software engineers in the future, but most likely it will assist software engineers in carrying out complex tasks faster, increase their productivity, and let them focus on the more challenging tasks.
Frequently Asked Questions
Q1: What is Devin AI?
A: It is the world’s first fully autonomous AI software engineer. It is like a software engineer who can plan an entire project, write, execute, and test complex code, take freelancing jobs from freelancer websites, and also deploy code on its own.
Q2: Who developed Devin?
A: One of the emerging applied AI labs named “Cognition” has developed it.
Q3: Will Devin replace Software Engineers or programmers?
A: No, most likely it will assist software engineers in carrying out complex tasks faster, increase their productivity, and let them focus on the more challenging tasks.
Q4: What Devin can do?
A: It can analyze, plan, and execute complex engineering tasks and make several important decisions like a human software engineer performs.
It can peform various tasks like end-to-end application development, finding and fixing bugs in codebases, train and fine-tune its own AI models etc.