When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
Devin AI, ‘the world’s first fully autonomous software engineer,’ likely marks the next disruption that artificial intelligence (AI) leaves on the world: Who needs a skilled coder anymore?
That’s the question we pose to various experts across the AI and software engineering fields, as we are not convinced that human coding is going anywhere soon.
When Cognition, the team behind Devin AI, unveiled videos of its early-access master coder last week — writing code from prompts, bug-fixing on the fly, even handling paid-for Upwork tasks for a cheeky ROI — there have been breathless exclamations that this is the end of coding as we know it.
Is Devin-AI going to impact the software arena with the same force that ChatGPT unleashed on the wider landscape less than 18 months ago? Are we mere months away from seeing a Devin AI-sized asteroid wipe out the field in which software engineers graze?
Let’s find out.
Key Takeaways
- Devin AI, developed by Cognition, aims to revolutionize coding by managing the entire development lifecycle, deploying apps autonomously, and detecting/fixing bugs.
- Devin AI’s performance metrics show promising results, but it’s still in its early stages and can’t handle all real-world programming tasks.
- Despite the excitement around Devin AI, skepticism remains about its actual impact and the hype surrounding it.
- Our verdict? While AI tools like Devin AI may change the role of software engineers, they’re unlikely to replace them entirely in the near future.
What Devin AI Brings to the AI Coding Landscape
Let’s be clear upfront: Devin AI, also written as “DevinAI,” is a remarkable tool — your virtual software engineer that can handle coding from a prompt, test the code, and deploy it autonomously.
Not sure which codebase is the best for you? It doesn’t matter: Devin AI can handle multiple languages, skipping easily from Python to Javascript, depending on what is the best fit for the purpose.
From smaller tasks like building websites to creating and deploying apps and complex software, Devin AI can converse with you as it goes, explaining what it’s doing and incorporating your dev team’s feedback as it goes. Possibly giving the developers more time to dust off their CVs and re-skill in other areas.
While we explore the pros and cons of AI coding assistants elsewhere, users of X, formerly known as Twitter, are already blowing up about the “death of software engineers.”
AI just killed Software Engineers officially?
Cognition introduced Devin, the first AI software engineer.
Here're 7 insanely powerful things Devin can do:
— Poonam Soni (@CodeByPoonam) March 13, 2024
software devs: ai WiLL NEVer bE goOD EnoUGh to RePLAce SoFtWARe DEVElOPeRs, we are SpeCIAl!!!111
devin: hallo 🫡 https://t.co/otMzLZHvqc
— @levelsio (@levelsio) March 12, 2024
Meanwhile, you can watch on YouTube as Devin AI complex a coding request on the freelancer platform Upwork — effectively offering ‘money for nothing’ for a prompt engineer with early access to the new virtual software engineer.
But, in our opinion, along with our expert panel drawn from companies including Figma and DataGPT, the quote attributed to Mark Twain seems most appropriate when it comes to software engineering: “The reports of my death are greatly exaggerated.”
How Devin AI Promises to Revolutionize Coding
Giving the floor first to Cognition, the company behind Devin AI, here are the claims to the platform’s strengths:
“It can manage the entire development lifecycle, from initial design to final implementation, handling various aspects like coding, integrating functionalities, and testing.
“It can autonomously deploy apps to platforms like Netlify, ensuring a seamless transition from development to live environments, thus streamlining the delivery process.
“One of Devin AI’s key strengths is its ability to autonomously detect and fix bugs in code. It can scrutinize codebases, identify errors, and implement fixes, thus enhancing code quality and reliability.”
Cognition is very direct in not calling Devin AI a replacement for coders — it “not only assists developers, but also collaborates with them, bringing a new era of human-AI partnership in software development.”
But a line on the Devin AI website is also telling:
“While highly advanced, it is not yet a complete substitute for human expertise, especially in complex, nuanced scenarios requiring deep contextual understanding.”
Not yet.
Devin AI Performance Metrics
While many large language models (LLMs) can code to some degree, Devin AI is specifically designed to be a virtual coder, working inside an integrated code editor and shell, familiar to anyone who has designed software before.
According to SWE benchmarks, it can solve nearly 14 of every 100 coding challenges set for it — a number that analysts expect to grow as the platform learns over time from previous work.
Will Devin AI Make Software Engineers Redundant?
We asked a panel of CEOs, chief product officers, and software engineers for their early takes on Devin AI.
Despite the rising sensationalist headlines, Ethan Gustav, Group President of North America, Infobip, is much more hopeful about the future.
“We understand that AI can be a source of concern, particularly for those in the technology field who may worry about job security. These positions are often filled by highly skilled and educated individuals with substantial salaries.
“However, we believe that AI, like Devin, holds promise for the future and can be harnessed to create new opportunities and advancements.”
Jenny Lea, Software Engineer at Figma, told Techopedia the end of software engineers is not in sight yet:
“I don’t come into work every day and set up a new app with all the integrations and deploy it. Can you take Devin to a stakeholder meeting? Can it give feedback on the feasibility of designs? Or mentor a junior developer?
“I do see the risk for freelance developers doing these sort of contained projects, but at the same time it’s a great opportunity for small/new businesses to get something up and running really quickly.”
Jithin Bhasker, general manager and vice president of the App Engine business at ServiceNow, also shared a more optimistic approach.
“Developers want to write less code but build more apps and automation fast with high-scale, low-code platforms. The more code you write or generate, the more developers you need in the loop to review, test, and publish. How about you shortcut to 80 to 90% finished app in minutes with simple prompts in English and you are adjusting just the UI for best user experience.”
Arshan Dabirsiaghi, Chief Technology Officer at Pixee.ai pointed out how embryonic AI coding is at the moment:
“Even by the metrics Devin measures themselves against, they can only handle a tiny portion of real-world programming tasks.
“The truth is that nobody knows yet when these things will take software jobs. Right now, the tools make developers more productive, but they don’t replace developers. I’ve never heard of a company that is cutting headcount to account for the improved productivity of tools like Copilot.
“In the very short term, the only thing that will replace developers is other developers who master using generative AI tools.
“Even if long-term tools turn out to effectively churn out senior engineer quality code, we’ll still need guarantees about the performance, security, and quality — and I think we’ll need human throats to choke for that.
Olga Beregovaya, VP, AI & Machine Translation at Smartling asks the question, “Who’s going to make sure the codebase doesn’t look like a total mishmash of code that humans can’t read if something goes wrong? Who’s going to secure the code, given study after study shows how insecure code is (even when written by senior engineers)?”
“Devin AI is most definitely a breakthrough in the way the world is implementing AI. We need to remember that, generally speaking, AI models perform better on solving math problems and can better “relate” to numbers and algorithms-based reasoning than dealing with purely linguistic tasks, where there is a lot more room for double meaning, vagueness, and errors.
“This is probably what will make Devin a successful project – “knowing” a huge library of code and being able to perform coding tasks independently should hypothetically be more suitable for AGI.
“Whether it will take jobs away from people – I believe it is the same question across a variety of “white collar” jobs, with the same answer – we need to view it as a co-pilot and see how the future unfolds.”
Evolution of the Sofware Engineer Role
And while the token-predictive skillsets of tools like Devin AI may be able to churn out code from prompts, there are areas of the human skillset they can not emulate yet.
Arina Curtis, CEO and co-founder of DataGPT, told Techopedia:
“It’s important to remember that AI can automate processes, but it cannot replace the nuanced skills and creativity that humans bring to their roles.
“[But…] embracing AI as a collaborator, not a threat, unlocks new levels of productivity and innovation.
“The real competition for work doesn’t come from AI or LLMs; it comes from an up-and-coming workforce that understands how to harness its capabilities to enhance their work and drive meaningful results.”
Kirimgeray Kirimli, President, Flatiron Software Co. believes that “AI software has not evolved fast enough to replace an engineer who’s been working for years.”
“While Devin might not necessarily have all the features ready, eventually some software will. It’s realistic to expect that software engineering jobs will transform as AI advances. It is highly likely that in the next 5-10 years, we will have some solution that lies between today’s prompt engineering and a full-fledged AI software engineer.”
Ryan Johnson, Chief Product Officer at CallRail, agreed that it may be the role of a software engineer that will change — evolution, not extinction:
“Devin AI and other similar companies will transform software engineering as we know it.
“Do I think they will replace every single engineer? No, but maybe they will change the role of software engineers in the future.
“Software engineers could be more focused on the ‘prompt’ of the code they would like to have written, which I would assume is way more detailed and sophisticated than what a non-engineer would ask.
“Utilizing a software engineer to write the prompts should result in better code due to their background and experience.
“Additionally, the complexity of the project will make a big difference now and in the future. The example Devin AI gives of creating a website makes total sense. Basically, it can make a more customized website than the current website builders such as Wix and GoDaddy.
“I think this is great for businesses to have more options and customization, and allow them to generate these websites without knowing how to code.
“I don’t see Devin AI replacing software developers working on enterprise banking software due to the risks involved and the large scale.”
“Overall, I’m frankly excited about these recent developments with Devin AI and hope to be able to try it out in the near future.”
Benn Stancil, CTO and Founder of business intelligence company Mode, gave his thoughts on a blog post well worth reading, from which, with permission, we use some excerpts:
“Within a few hours of the [Devin AI] announcement, people were already suggesting that it might be the next $100 billion company, which is 25 percent more than even OpenAI is currently worth.
“There are tons of companies building AI agents like Devin, and you can never fully trust a pre-recorded demo. Instead, my belief in the potential of Cognition comes entirely from other people’s belief in the potential of Cognition—which was, at least to some degree, astroturfed by Cognition’s investors and PR team.
“The point here isn’t that the emperor has no clothes or even that they don’t have $100 billion in clothes. The point is that nobody has seen the emperor. Investors have, and we’ve seen curated demos. But the clothes are still invite-only. And yet, all at once, the entire internet began telling the same stories of their wonder.
“Imagine that you didn’t see the Devin demo because a Silicon Valley celebrity tweeted about it. Imagine that the video only had a few thousand views. Imagine that you found it on the second page of Product Hunt on a slow Thursday afternoon.
“Imagine it had fifteen upvotes and a lonely comment from the “maker,” saying they were very excited to share their project with the world. Imagine that the thing itself was exactly the same, but there was no party and no posting around it.
“Would it be impressive? Would you think anything of it? Would you want to invest? Would you roll your eyes at what was probably the fourth AI bot that you saw on your bored scroll through Product Hunt?
“I really don’t know. I doubt I would’ve thought that it was any different than the other belts I’ve seen. Which may well be wrong — Devin really might be different, and in ten years, Devin may be building all of our websites, and our phone apps, and even future Devins.
“But without the hype, I don’t think Devin and its five-minute sizzle reel would register to me as anything that was particularly noteworthy.”
Steve Hegenderfer, Vice President, Product and Platforms, P97 Network, told us:
“In general, these generative AI tools are interesting, to say the least. They can take a lot of the grunt work and project start-up costs out of software development.
“…But at what cost? Will the code be more bloated in the long run? How much optimization and bug fixing will they entail? Will I have to spend just as much time customizing for my organization?
“I think that the possibility for these types of AI tools is great, as the number of patterns and practices for software development keeps growing and giving tools like this even more things to learn from.
“But I still think that we are a ways away from an AI that creates really great, secure, bug-free code. But I do love the initiative of these folks!”
When we interviewed Dr. Nicola Hodson, IBM UK and Ireland’s Chief Executive, this week, around AI collaborating in the workplace, she suggested an example that we find apt for the promises of Devin AI:
“I listened a while back to a podcast on BBC Radio 4, and they were talking about the advent of the washing machine, which was expected to change housework forever and give us all more free time. Of course, what happens is that you find more tasks to do in that time that’s been freed up.
The same thing is happening with AI — if a software developer becomes 30% more productive, that doesn’t mean they do 30% less work. This means they have more time to do work of higher value.”
The Bottom Line
Devin AI is a remarkable advancement in the AI field and not an unexpected one. Any futurologist has been predicting AI taking over software engineering. As Nvidia’s CEO Jensen Huang states, ‘the programming language of the future is human’.
The question is which team will develop the answer first, and Cognition, with Devin AI, is the first company to take a real leap towards a practical solution. Perhaps they will impact the industry in the same way ChatGPT took the world by storm less than 18 months ago.
However we think there are many miles ahead before we see a prompt engineer create the next multi-billion company from asking Devin AI a few questions.
But tools like Devin might assist along the way, and be a key enabler for companies and their development teams.
We give the last line to Arshan Dabirsiaghi of Pixee.ai, on whether tools like Devin AI will replace the software engineer:
“On a long enough timeline, I think yes — AI will take over this job. This will happen. I am a technologist and an optimist. But I absolutely can’t see it happening in the next 10 years.”
