Azizi O.’s Post

Devin, seems incredible! #CognitionLabs #Devin #PuttingProgrammersOutOfWork "It can autonomously plan and execute thousand-step tasks. It can build and deploy entire software projects all by itself. It can research and fix bugs 7x better than OpenAI's GPT-4, and it trains and deploys its own custom AIs to solve problems. Cognition Labs has announced Devin, the world's "first AI software engineer." And while it's true that previous LLMs like GPT-4 and Anthropic's Claude have been able to write and execute code for some time now, Devin seems like a significant step change. In essence, this new AI is designed to act like an entire software team – tell it what you want, and it'll put its project management and business analysis hats on to devise a plan and build requirements. It'll then create little AI minions to go and execute certain steps, flipping between their own sandboxed terminals, code editors and browsers. It'll then test, debug and iterate until it assesses the entire application complete, and deploy it for you. If you want, it'll do this whole process – which could involve thousands of decision points – completely autonomously, simply giving you a final product to look at and request changes to. Or experienced programmers can treat it more as a collaborator, staying more involved in decision making and design, or simply use it as a team of coding or testing minions, or a documentation specialist. In some sense, then, it looks somewhat like what AutoGPT promised, but couldn't immediately deliver on: an AI executive in charge of its own team, that manages an entire project from go to whoa. It does seem to have some wild new capabilities though; Cognition Labs says it's capable of boning up on new technologies it might need to get a job done. In the below example, it reads a blog post to figure out how to use ControlNet on Modal, then within a couple of minutes, it's used this previously unfamiliar tech and techniques to achieve the desired outcome: in this case, generating AI images with words embedded in them. Possibly more freaky is Devin's ability to create and train its own slave AIs. In the video below, the Devin system clones a version of Meta's 7 billion-parameter open-source Llama language model, checks out the readme file to learn how to set it up, and then does so – even deleting and reinstalling packages that aren't working. It then starts a training run, and within a couple of hours, it has cloned and trained a new AI model specifically for a task. AIs spawning and training their own home-brewed AI agents; it's a remarkably powerful idea and absolutely the kind of thing a next-gen autonomous programmer probably needs to be able to do, since so many tasks now can and should be handled by increasingly capable custom AIs. On the other hand, good lord; anyone on the "AIs will seek power and kill us all" side of the fence is unlikely to be delighted by this idea. ..." Via Timothy Horton

At Barnacle Labs we're full into AI code generation tools - things like Github Copilot, Cursor and the ChatGPT, Claude and Gemini assistants. We've seen these tools have a dramatic impact on our day-to-day activities. Those impacts include: 🏎️ Increased productivity - if you get AI to generate code, you can get more done in the same amount of time. However, the impact varies greatly on the task at hand, from "I did what would have taken me a whole day in 1/2 hour" to "AI had a marginal impact". I don't have formal metrics, but productivity improvements are easily in the multiple 10's of percent. That's a big deal. 🛠️ Multi-Skilling - it's dramatically easier for an engineer to work in adjacent spaces where they might have limited experience - a back-end engineer now able to do a decent job at front-end, etc. This results in every engineer becoming significantly more multi-skilled, meaning there's less need for more specialist skills on a project. One person doing multiple things is frequently a lot more efficient because we avoid the hand-offs and mis-communications that come with lots of niche players. 😊 Feel-good-factor - it's simply more fun if you can get AI to do the mundane and complex tasks, freeing your time to focus on the more creative aspects of our roles. A lot of the "grunt" work can now be automated and the truth is that few of us enjoyed the mundanity of those activities. After all, who doesn't want a co-worker who never argues and does whatever you ask them to? ✨ Quality - I've personally been amazed to see AI rewrite my sometimes quite ropey code, simplifying and structuring everything for me. It seems clear to me that we're going to see increased code quality and better maintainable systems. But there are considerations: 🤔 Knowing how to use AI, when not to use it, how to prompt it, etc, becomes a core skill that needs to be perfected. 🚫 AI can't solve every problem, but rarely admits such. Instead, it'll sometimes take you down a blind alley - so recognising this early becomes essential. 🔐 Security of API keys/passwords and the risk of sensitive source code being leaked is a key concern that needs to be thought through. As a practitioner, I can state with confidence that the AI-ification of software engineering is proceeding rapidly. It's fair to say that we've reached the point in Barnacle Labs where it's inconceivable to imagine coding without AI. But despite the concerns for jobs, the truth is the world is desperate for more software - the result is likely to be more software, not fewer engineers. A recent report examining the impact of Github Copilot has just been published and it supports many of my experiences: https://lnkd.in/eV4QFNe4 My final thought... why should we believe these impacts will be restricted to software engineering? It seems clear to me that it won't - similar impacts will ripple through many disciplines in the coming years 💥

It’s great to see the Stackoverflow survey not only show that Codeium is the most satisfying AI tool to use but is also the most productive! https://lnkd.in/g58S6Xxt

Meet Devin, the amazing new AI tool from Cognition. It's super smart – it can write code, make websites, and create software, all with just one instruction. Devin is like having your own super-skilled AI engineer. But here's the cool part: Devin isn't here to replace human engineers. Instead, it's here to team up with them, making their jobs easier. Cognition says Devin is all about helping human engineers, not taking their place. What is Devin? Devin is a super-smart computer program created by a company called Cognition. It's like having a really clever assistant for software engineering tasks. With just a simple instruction, Devin can write code, build websites, and make software all on its own. But here's the cool part: Devin isn't trying to replace human engineers. Instead, it's meant to work together with them to make their jobs easier. What makes Devin special is its ability to think ahead and solve tricky problems. It can learn from its mistakes and keep getting better over time. Plus, it has all the tools that a human engineer needs, like a way to write code and browse the internet. Devin has been tested against other AI programs, and it did way better, solving almost 14 out of 100 problems compared to just under 2 for others. And it's not just a test, Devin has already done real jobs on platforms like Upwork, fixing issues and making reports. So, in simple terms, Devin is like a super-smart assistant that helps engineers do their work faster and better, without taking their jobs away. How does Devin work? Devin works by using advanced artificial intelligence (AI) algorithms to understand and execute tasks related to software engineering. When given a prompt or instruction, Devin analyzes the request and uses its vast database of knowledge and problem-solving techniques to generate code, design websites, or develop software. One of Devin's key features is its ability to think ahead and plan complex tasks. It can make thousands of decisions based on the given task and learn from its mistakes to improve its performance over time. Devin also has access to essential tools like a code editor and web browser, enabling it to complete tasks from start to finish. What sets Devin apart is its adaptability and versatility. It can learn new technologies, tackle a wide range of engineering challenges, and even train its own AI models. Additionally, Devin can collaborate with human engineers in real-time, providing updates, accepting feedback, and contributing to design choices. Overall, Devin works by harnessing the power of AI to automate routine tasks, streamline workflows, and empower engineers to focus on more complex problems. By combining human expertise with machine intelligence, Devin represents a significant advancement in software engineering technology.

GenAI applications are revolutionizing the ways that we live and work, and they are forever changing how we think, gather information, and relate with others. This is not only true in the programming industry, it’s also a harbinger of greater change to come. The automation of software development and maintenance continues to rapidly advance and that’s opening up new opportunities for developers in more complex and higher-level work. This trend is expected to grow and spread across a variety of industries which hints at a broader movement rather than just a singular event.   These developments are also reshaping how humans and machines interact. AI-driven apps can now understand codebase nuances and patterns which enables them to generate complete modules of code rather than mere snippets. And this has significantly altered the software development cycle. The process of detecting and correcting programming errors, and then running tests, is also changing things. AI apps can now perform many tasks faster and with better results than humans can, and these are not just one-time patterns. They signal major changes that are set to ripple through other sectors of our economy as NLP software continues to interact with people - in real-time. Also, technologies like speech-to-text, automated text generation, and text analytics are making are making inroads. These tools are revolutionizing human-machine interaction and transforming work processes into highly collaborative, efficient, and accurate work sessions. Code translation and refactoring are just some examples of the capabilities of GenAI software. These apps can translate code from one programming language to another thereby bridging the gap between legacy systems and new technologies. And they not only suggest improvements to existing code, they also implement the changes, assess the outcomes, and optimize the entire process. These tools can also be adapted to individual preferences which makes for happier people. This technology simplifies and refines the development of programs and other products. It also promotes human involvement and paves the way for more significant and fulfilling work. It also helps workers to focus more on business strategy and play a bigger role in organizations. As GenAI continues to remold software development and other industries, expect these trends to gain momentum. This is all a part of our Brave New World of technology as the transformative effects of the Fourth Industrial Revolution continue to unfold. “There is a lot of discussion about the potentially positive or negative impact of generative AI on job roles and creativity. But the good news for software developers is that generative AI is assuming many mundane tasks and elevating their roles to business consulting and customer experience.”

Performing mundane tasks and reading through uncommented, messy code can be quite frustrating for developers 😖. With the advent of AI, why not utilise it to enhance efficiency? Explore the pros and cons of incorporating AI in development; click the link to learn how to protect your data🛡️ while using AI tools🚀.

A few people have remarked to me recently that they think software developers are going to be replaced by AI. You may have seen Devin, an AI developer, that tends to play into this idea. However, I think we are a very, very long way from developers being automated out of a job. I just watched a very detailed debunking of Devin which prompted me to write some thoughts. Firstly, the hard part of software development is figuring out what is actually needed. Often customers and managers ask for things that make no sense. A back-and-forward game, figuring out what it is they actually need, is the most valuable part of the development process. Just building what someone asks for is almost always a mistake. They may have badly formulated their request, they may be unaware of some critical details or they may just not realise there are different or better ways to solve their problem. As someone who uses AI in software development I can state that it’s a huge productivity benefit. But it also has big limitations. 50% of the time it’ll take me down the wrong path and it’s my job to realise that and ignore the AI when it happens. It is a useful tool, but it’s nowhere near ready to build entire systems on its own. Things like Devin suggest that AI is further ahead than I personally think it is. The detailed debunking video I just watched shows that Devin is fixing bugs (impressive) but also that the bugs it’s fixing were ones it introduced into a very simple task (not so impressive). This doesn’t surprise me. I see this pattern a lot - the AI does something stupid, I tell it I got an error and it says “oh, you are right - the code I gave you is wrong”. It’s smart enough to suggest a solution to a problem but dumb enough to, half the time, introduce more problems as it fixes the original bug. A human can recognise this death spiral and ignore the AI. But an AI on its own will just keep fixing and introducing more problems. What AI does today is help lessen the need to remember syntax, to automate some things like writing simple utility functions, help auto suggest variable names. It removes the drudgery elements of writing code and it improves productivity. But the idea that someone will be able to just say to the AI “build me a product catalog website” is farcical. For a long time yet it is going to be a tool to help developers. It will get better and help them more over time, but it’s unlikely to replace the developer any time soon IMHO. And if it does replace developers, it’ll probably also replace those making the requests as well! But I think that’s a long way off. So let’s roll back the hype. AI is a useful tool that can be a big productivity benefit. But it’s not about to automate developers out of existance any time soon. And let’s not forget that “yes I know that’s what you asked for, but have you considered…” is often the most valuable conversation a developer ever has.

Is AI over-hyped? I think there are two different ways to answer that question: What *can’t* the AI do for us? or What *can* the AI do for us? Let’s look at an example to see how these different questions can yield very different perceptions of how valuable large language models (LLMs) actually are. Software development is one area where AI appears to have a lot of traction right now. If we ask what AI *cannot do* in software development, we might see that AI cannot: - Invent algorithms that were not in the AI’s training data - Design innovative/creative user experiences without lots of human help - Solve problems the AI hasn’t been trained on - And so on… You’re basically left with the conclusion that AI is over-hyped and not having a big impact on software development. If, however, we ask what AI *can do* when it comes to development, we see that AI can: - Suggest routine code completions as developers type - Act like a tutor to help us understand new technologies / languages / frameworks - Write detailed comments and documentation - And so on… When we ask the question that way (what can AI do for us?), there’s really only one conclusion: that AI is actually having a positive impact on how we write code. That shows up in developer productivity data as well: GitHub claims that 88% of developers who used their Copilot product felt they were more productive and completed tasks 55% faster than those that didn’t. 60–75% of users reported they feel more fulfilled with their job, feel less frustrated when coding, and are able to focus on more satisfying work when using Copilot. Software development is just one example. I’m sure we can find others in areas like customer service or sales. So, it really comes down to how we ask the question: can or cannot? I’d encourage us all to ask “can” questions, rather than “cannot” questions.

New research from Stack Overflow reveals that AI tool usage among developers increased to 76% in 2024, up from 70% in 2023. 81% of developers cite increased productivity as the top benefit of AI tools. No doubt AI is an incredible tool but the article continues on to talk about the risks associated. Only 43% of respondents trust the accuracy of AI tools, and 79% report misinformation. I believe it is critical to first understand AI and then how to effectively use AI to mitigate such concerns before you dive in trying to leverage AI. https://lnkd.in/gfQfpTKb #keepupwiththenews #keeupwithyourskills #pluralsight #acloudguru

New research from Stack Overflow reveals that AI tool usage among developers increased to 76% in 2024, up from 70% in 2023. 81% of developers cite increased productivity as the top benefit of AI tools and 70% don’t see AI as a threat to their jobs. In fact, individuals believe AI is democratising coding, meaning the number of developers is increasing rather than decreasing. Yes, yes, I think everyone understands AI = helpful. But what's the next step for the everyday developer? From what I hear, is using sandbox playground environments to run such codes. What do you think? https://bit.ly/3WmvjXY #keepupwiththenews #keepupwithyourskills #pluralsight #acloudguru