Stealth Startup, Cognition, backed by Peter Thiel’s Founders Fund and former-Twitter executive Elad Gil and Doordash co-founder Tony Xu, announced a fully autonomous AI software engineer called Devin. Devin can handle entire development projects end-to-end. It can write the code, fix bugs associated with it until final execution. This is the first offering of this kind and even capable of handling projects on Upwork.
They had a $21 million Series A led by Founders Fund. They have support of industry leaders including Patrick and John Collison, Elad Gil, Sarah Guo, Chris Re, Eric Glyman, Karim Atiyeh, Erik Bernhardsson, Tony Xu, Fred Ehrsam and so many more.
Devin already did real jobs on Upwork.
Seven Times Better Than Previous Best
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. Devin was unassisted, whereas all other models were assisted (meaning the model was told exactly which files need to be edited).
Devin can train and fine tune its own AI models. Devin sets up fine tuning for a large language model given only a link to a research repository on GitHub.
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.
So, Cognition AI just introduced Devin, the most capable autonomous coding agent I've seen.
Instead of just autocompleting tasks, it can write entire apps on its own in minutes.
It also correctly resolves 13.86% of issues, autonomously. pic.twitter.com/7EWncjRQmd
— Rowan Cheung (@rowancheung) March 12, 2024
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Saw a nine year old video that sounds exactly like today’s arguments: driverless cars are “already better than humans”. Machines are writing code. Etc.
One point the vid made that stands is that of the top most common jobs (I forget the number) only one was less than a hundred years old and not immediately in danger of automation; computer programmers (yes there was probably some woman programming mechanical machines more than a hundred years ago but just talking about common careers)
The vid mentioned that horse populations peaked in 1915. They didn’t become unemployed because they became lazy but because they became unemployable. They didn’t choose to stop breeding like humans did but not being economically viable parts of a productive economy had something to do with it.
Humans won’t be replaced anywhere near as soon as people think but it will happen.
Do keep in mind its only going to get better. As its said just wait for two more papers down the line…
That said we will always need some form of software engineer even if the job changes. They have said it’s the end of the sw engineer since the first compiler was written. We don’t need anyone to write machine code anymore. 😜
I am a software engineer and have been for 30+ years. I’m going to go with what I read in Infoworld just the other day, as it speaks to my own experience in such matters:
“The prospect of AI-driven programming has led to some fairly grandiose predictions about the future of the software industry, particularly from C-suite execs, consultants, and the pundits who follow them. But what about the programmers and managers who work with AI tools on a day-to-day basis? We asked a handful of people programming with generative AI how it’s working out for them so far. What we learned is that AI really is changing the way people work—but machines aren’t about to replace human coders anytime soon.” — Josh Fruhlinger
Is this another “the machine spits out Python code, it must be good” product? Or does it just do javascript and CSS?
I super hope some naive MBA starts a software company with nothing but Devin. That will be fun to watch.
AI coding software Version(n) writes its own next Version(n + 1)…
AI coding software Version(n + 1) writes its own next Version(n + 2)…
AI coding software Version(n + Y) writes its own next version(n + Y + 1)…
How does this end and what is the outcome?
OOF……..
wow, I’m so happy I’m not a coder, their days are truly numbered. I’d bet by the end of the year, half will be jobless. Then the following year will be 90% of the remaining.
That is a statement that can only be made by someone who has never done SWE work.
I think he was being sarcastic. What AI really seems to be doing is giving birth to a new software engineering discipline that some are calling “prompt engineering.” The name is largely self-explanatory; it’s about providing prompts while a coder codes. By collaborating with the human programmer, what emerges should be a faster and more efficient way of producing solutions than either a machine (at least at this point in history) or a human, regardless of skill level, might create.
<14% successful error correction is not only unimpressive, it would get a human coder fired.
When I was still coding we had code generators too, that could supposedly write entire programs. But they produced bloated code – as in 3-6X more code – that was slow, inflexible, hard to decode by a human later on. I always found keeping an extensive library of subroutines I could tweak for individual programs to be much more efficient and able to produce a similar look and feel so that users didn't have to relearn entire ways of doing things.
OK, AI can do a lot of things I could never do, but getting better than 14% error-correction is going to get exponentially harder from here. The last 25% or so will take years, just as FSD has slowed down as it's gotten better and better. The fringe cases take more and more time to resolve.
Just wait till AI has to deal with spending 20 minutes compiling every change and is unable to quickly try random stuff to fix a bug.
FWIW this is good, less coders in the future because kids are scared away from it because of AI means guaranteed long term employment with nice raises for me.
Just wait until the AI decides the compiler is too slow and rewrites the compiler.
Thanks, that was worth a good laugh.