Move over, human coders – there’s a new kid on the block, and it’s an AI that can seemingly do it all. This was the vibe that we’re getting from the tweet/x released by cognition when they introduced their new product ‘Devin’
Devin is artificial intelligence model developed by the startup Cognition, and it has been turning heads with its ability to write complex code, debug errors, and even learn new technologies from the documentations. But is Devin really a game-changer, or just another coding assistant in a long line of AI tools?
What is Devin AI?
Devin is an AI model backed by heavy hitters like Peter Thiel’s Founders Fund, former Twitter exec Elad Gil, and Doordash co-founder Tony Xu. Cognition, the startup behind Devin, has secured an impressive $21 million in funding to bring this AI coding assistant to life. While AI coding tools like Microsoft’s Copilot have been around for a while, Devin claims to take things to the next level by handling end-to-end development responsibilities all on its own.
What Makes Devin Different?
The question on everyone’s mind: what makes Devin so special? To answer that, let’s first look at the most challenging parts of the assisted coding tools available now. It’s the back-and-forth communication a developer needs to engage in to get something to work. Additionally, there’s the lack of ability of the LLM to test their written code. Although there are some plugins available as workarounds, they are still not refined and are mediocre at best. Cognition bridges the gap by giving Devin a browser, a command prompt, and the environment to run the code all on its own in a sandbox environment. This enables it to run the code, get resources from the web, and then analyze the outputs and fix errors. It has the ability to recall relevant context at each step, self-learn over time, and rectify its own mistakes – just like a human programmer would.
On the SWE-Bench benchmark, which evaluates large language models on real-world software issues from GitHub, Devin correctly resolved a whopping 13.86% of the issues without any assistance, compared to the previous state-of-the-art model’s unassisted score of just 1.96%.
Interacting with Devin
You might be wondering: how does Devin actually work? Although its not available for the public yet. According to Cognition’s founder and CEO Scott Wu, Devin operates within a secure sandbox, planning and executing complex engineering tasks through common dev tools like code editors and web browsers.
All a human needs to do is provide instructions to Devin via a chat interface. From there, the AI dynamically maps out a solution, writes the actual code, fixes bugs along the way, tests its work, and keeps the user updated in real-time. If the programmer spots any issues, they can simply message Devin to course-correct. Wu demonstrated Devin’s impressive range in a blog post, showcasing its ability to deploy web apps and websites, as well as fix issues in GitHub repositories.
Is Devin a threat to Programmers? and The Future of Coding
While Devin’s capabilities are undoubtedly impressive, its true impact on the programming profession is yet to be determined. And to answer the question “is Devin and tools like it are a threat to programmers ?“ the answer is both Yes and No. The rise of AI in coding will likely lead to shifts in the job market, but it also presents opportunities for those willing to adapt and embrace the changes. For now we don’t think any skilled and talented developers should be worried. However, those who are merely getting by without enhancing their skills may need to reflect on their future.
We are saying this because, Devin’s brain still remains reliant on an LLM and lacks any recent advancements in AI technology. However, through meticulous fine-tuning, training, and providing it with a secure sandbox environment and offering freedom, Cognition have undeniably enhanced its capabilities, making it exceptionally impressive and currently the top performer in the SWE benchmark.
As with any disruptive technology, it is crucial to approach AI in coding with a balanced perspective, recognizing both its potential and its limitations. Devin may be a glimpse into the future of coding, but it’s a future that still involves humans.
Find more about Devin Here