Luke Parker’s Post

Devin, the first AI software engineer can finish entire software projects on its own in minutes, is capable of complex multi-step reasoning, and never derails even after performing thousands of tasks. Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser. When evaluated on the SWE-Bench benchmark, which asks an AI to resolve GitHub issues found in real-world open-source projects, Devin correctly resolves 13.86% of the issues unassisted, far exceeding the previous state-of-the-art model performance of 1.96% unassisted and 4.80% assisted. 👉 https://lnkd.in/dx7kTWCw #machinelearning #swdevelopment

Meet Devin by Cognition AI the first AI software engineer! It explains how it debugs and develops via its own IDE, command prompt, and web browser in real time! Thoughts on Devin as your next web developer? #conversationalai #innovation #softwareengineering

Designation to An Agent Quick Dive : Cogintion-labs dropped a bomb called Devin, claimed to be first AI software engineer. In the quick demo provided, the agent takes requirement inputs from the user, develops a step to step documentation, writes the components plus testcases and generate the deployment along with code pushed to repository. The multi-modal agent is an extension to code assistant agents that helps with just code snippet. It also performs debugging in a human reasoning fashion to comment on the issue, logs the same for reference. The creators claim it to have solved 13.86% of the Software Engineering Benchmark problems second best to Claude 2. Opinion : It's been quite sometime, I have been using and evaluating code assistants like Github Copilot, DeepSeekCode and LlamaCode 70B along with conservational search engines like Phind and Perplexity. The reasoning has always been lacking in terms of the arithmetic sense of the problem. Copilot badly fails in writing test cases and always says it is trained till 2021 data and cannot help with libraries post that. Devin can be called a multimodal agent with improved reasoning capabilities for programming but calling it AGI is not justified. A narrowed AGI would still be an ambitious call but just to satisfy the GenAI endorser's ego. Quite impressive in the video demo. It's only when the agent is going to get out in the wild that we will hear more real-world opinions of the same. #Genai #generativeai #llamaindex #llama2 #codeassistant #copilot #ml #machinelearning #machinelearningengineer #deeplearning #ai #artificialintelligence

This is evolving too fast ✨ Exciting yet concerning. Companies must hit the accelerator to keep up and stay ahead of competitors. Devin is an autonomous agent that solves engineering tasks through the use of its own shell, code editor, and web browser. The creators claim that when evaluated on the SWE-Bench benchmark, which asks an AI to resolve GitHub issues found in real-world open-source projects, Devin correctly resolves 13.86% of the issues unassisted, far exceeding the previous state-of-the-art model performance of 1.96% unassisted and 4.80% assisted.

Is Devin the death blow for software developers, or should I say "human software developers" or is it just a hype tool that will just be thrown away once the dust settles? I honestly didn't expect such technology to be available this soon, I thought it might take years to get to where we are now, but damn, we're already here and I just might be replaced by this robot once it becomes available for companies. but I don't think it will take anyone's job, because everyone knows that these AI tools are intended to solve small and less complex problems, but when you really get down to the real deal where the system is really complex, these tools might harm your code more than repair it. he. What do you think of that ? for more information watch the official video

What does this mean for the future of Software Engineering...? Yesterday, Cognition AI introduced Devin, the "world's first AI software engineer". According to reports, Devin has the capability to take a single command and turn it into a functioning website or software programme. It can also identify and rectify bugs, reducing the time and effort required for troubleshooting. Instead of just offering coding suggestions and auto-completing some tasks, Devin can take on and finish an entire software project on its own. It is equipped with its own command line, code editor, and browser which helps it to operate independently https://lnkd.in/ePNrTR6V

This demo blew up recently. Devin is being marketed software engineer replacement. You ask it questions, and it creates full projects. Of course, without being able to access Devin myself, it's impossible for me to confirm the general performance of this system. I recall that GPT-4 was touted as a replacement to developers a few years back. So far, it seems like it's only replaced stack overflow (kind of). It is an exciting product, though. In my opinion it's likely fueled more by advances in prompting strategies rather than advances in LLMs. The step by step plan, followed by iterative processes which tightly couple with specific tools feels very similar to the ReACT framework published in 2023. Here's a link to that paper: https://lnkd.in/djW3VKMa I also happen to have an article on ReAct, and agents in general: https://lnkd.in/ge7shmCj

This could remove the need for software developers totally... But I do not see that happening in a hurry. Devin looks some amazing piece of technology but I have all sorts of questions. How do you give it access to private API's or API's that don't have great documentation (Crazy that this is still a thing in 2024)? How do you ensure that it follows the SSDLC that your companies follow? How well does it pick up legacy code bases? Are there languages that it will just say no to (like F#)? There's a ton of question that I need answers for but when I finally get access I will give updates. Side note: If AI is sooooo great, why does the website look so bad?!

Introducing Devin, the first AI software engineer Devin is a tireless, skilled teammate, equally ready to build alongside you or independently complete tasks for you to review. With Devin, engineers can focus on more interesting problems and engineering teams can strive for more ambitious goals. Devin’s Capabilities With the advances in long-term reasoning and planning, Devin can plan and execute complex engineering tasks requiring thousands of decisions. Devin can recall relevant context at every step, learn over time, and fix mistakes. We’ve also equipped Devin with common developer tools including the shell, code editor, and browser within a sandboxed compute environment—everything a human would need to do their work. Finally, we’ve given Devin the ability to actively collaborate with the user. Devin reports on its progress in real time, accepts feedback, and works together with you through design choices as needed. So, what's your thought on the Devin? What impact it can make to the job market of the software engineers? Looking our your valuable thoughts and insights. https://lnkd.in/eBj4zj9b

You may have heard about Devin, the Artificial Intelligence Software Engineer focused on duplicating what a developer that do. Should developers worry? The interesting thing about this tool is that it's basically doing what Copilot announced it was working on during the last GitHub Universe, but hasn't released. If you look at the demo you'll see that it receives a task and breaks it down into step. GitHub wants to convert that into a trackable Github project.   One thing that's different is that it's using an IDE interface with access to a terminal that it can run, it can check websites to read up on docs and accomplish tasks. So in a way, it behaves more like a regular developer in order to take care of things.   It all looks futuristic and dangerous, although the demo was a simple completion of a chart. Something you can do with ease on just about any chatbot.   I think this is a space to watch, specially because what we will get from GitHub might be more impressive than this, but I do like the rabbit-like ability of this tool to use the IDE to more literally behave as a developer. Honestly, this and the Rabbit tool. I don't see them working perfectly enough to trust to do a developer job, but I've been wrong before. To me when we perfect something like shipping, what we end up with is more of that thing. So these tools have the potential to explode development in the world. And that's still going to need people who can use these tools to their max.    https://lnkd.in/e4jWssbd