Key insights of today’s newsletter: Cognition Labs, emerging from stealth last week, announced Devin, the world’s first fully autonomous AI software engineer. Devin is equipped with common developer tools including a shell, code editor, and browser within a sandboxed compute environment and can engage in end-to-end coding tasks.
I’m a professional software engineer. I’ve been waiting for this since LLMs first appeared on scene. I got into software because I love building things and exploring about what can be invented with code and data.
A virtual programmer, for me, doesn’t need to be perfect. Just good enough to enable me to close the gap between idea/question and implantation a bit faster.
My only hope is it gets even better even faster so I can quit my job and build the things I’m already working on faster.
Junior devs who actually love understanding how digital systems are constructed will be fine. But yes, devs who mostly copy paste are over.
Supposing Devin and his mates DID replace junior engineers. We’d still need senior engineers to deal with the other 87%.
Senior engineers don’t arise spontaneously, they spend years being junior engineers to learn those skills. I’m struggling to visualise the development path in this scenario. May be a failure of imagination on my part tho, i am old :)
That’s assuming Devin and his mates remain at junior engineer level of course. And that they’re not smoke and mirrors :)
However, any new technology and its applications tends to follow the Gartner Hype Cycle and generative AI is now on top of it, before going through the slope of disillusionment until we figure out the exact scope in which these new technologies will thrive.
As more companies are using SaaS applications and fewer custom made applications, I think software development will happen more and more in specialised software companies, who are selling software to the "more traditional companies".
I foresee that such LLM powered tools will be more efficient in traditional companies who are developing "standard applications" as opposed to organisations, which are developing very custom made applications, which might be too far "from the norm" for such a tool.
Such tool might change very much data pipeline development, which tend to follow "more standard operations".
I am curious to see where we will be in one year from now about it.
'Devins running wild on the Web" maybe even with capacity for arbitrary "thought"... Yep. Scary. 😬
I’m a professional software engineer. I’ve been waiting for this since LLMs first appeared on scene. I got into software because I love building things and exploring about what can be invented with code and data.
A virtual programmer, for me, doesn’t need to be perfect. Just good enough to enable me to close the gap between idea/question and implantation a bit faster.
My only hope is it gets even better even faster so I can quit my job and build the things I’m already working on faster.
Junior devs who actually love understanding how digital systems are constructed will be fine. But yes, devs who mostly copy paste are over.
Supposing Devin and his mates DID replace junior engineers. We’d still need senior engineers to deal with the other 87%.
Senior engineers don’t arise spontaneously, they spend years being junior engineers to learn those skills. I’m struggling to visualise the development path in this scenario. May be a failure of imagination on my part tho, i am old :)
That’s assuming Devin and his mates remain at junior engineer level of course. And that they’re not smoke and mirrors :)
I believe the future is bright.
However, any new technology and its applications tends to follow the Gartner Hype Cycle and generative AI is now on top of it, before going through the slope of disillusionment until we figure out the exact scope in which these new technologies will thrive.
As more companies are using SaaS applications and fewer custom made applications, I think software development will happen more and more in specialised software companies, who are selling software to the "more traditional companies".
I foresee that such LLM powered tools will be more efficient in traditional companies who are developing "standard applications" as opposed to organisations, which are developing very custom made applications, which might be too far "from the norm" for such a tool.
Such tool might change very much data pipeline development, which tend to follow "more standard operations".
I am curious to see where we will be in one year from now about it.