Dropbox former CTO Aditya Agarwal, after spending a weekend coding with AI, suddenly realized, “Humans will never write code by hand again.” Even more brutally, in the age of AI, resumes and seniority mean almost nothing; what truly matters is whether you’re willing to embrace change. This article is based on Aditya Agarwal’s post on X titled “When Your Life’s Work Becomes Free and Abundant,” translated and reorganized by FanShu哥.
(Background: Even high-salary senior engineers are being beaten by AI: submitting 800 resumes into the abyss, delivering food for a living, living in trailers)
(Additional context: A global trend called “Vibe Coding” is emerging, with AI helping everyone become engineers)
Recently, Aditya Agarwal spent a weekend using Anthropic’s AI assistant Claude to write code.
This isn’t an amateur programmer; he’s been coding for over 20 years, one of Facebook’s earliest engineers, personally building Facebook’s initial search engine, later becoming Dropbox’s CTO, expanding the engineering team from 25 to 1,000.
Coding has been his life.
And after that weekend, he reached a conclusion: We will never write code by hand again.
Here is the main body of what he wrote:
Not long ago, I spent a weekend coding with Anthropic’s AI assistant Claude.
I’ve been coding for over 20 years. I was one of Facebook’s earliest engineers, responsible for building the initial search engine. Later, I became Dropbox’s CTO, growing the engineering team from 25 to 1,000.
Coding has always been the foundation of my career, a skill I’ve dedicated my entire adult life to refining.
But after that weekend, one thing became very clear to me.
We will never write code line by line by hand again.
What was once a skill I was very good at has now become a free, abundant ability.
When I use AI to build software, I also notice another thing: AI agents elsewhere are building their own community platforms, which is exactly the kind of product I helped create at Facebook.
Some small coding agents are building fully functional community platforms for themselves.
This whole thing is a bit absurd, but the results they produce are almost indistinguishable from what humans have built on large networks.
In terms of form and function, everything I did early in my career can now be generated by machines.
I sat there thinking for a long time.
What I felt was a mixture of wonder and deep sadness.
Watching the pillars of my professional identity—what I built, how I built it—being replicated in a weekend by a tool that doesn’t eat, sleep, or take breaks, was truly disorienting.
But this sense of disorientation has a characteristic: it will pass.
And what replaces the sadness is something I never expected.
A wild, even reckless energy.
In the five days after that weekend, I wrote more code than I had in the past five years.
This is not an exaggeration.
The software I produced was better and more ambitious than what I had written myself before.
Things I never would have tried before because of high development costs, now can be done in an afternoon.
I’m not watching myself get replaced.
I’m witnessing the disappearance of limitations I’ve silently accepted throughout my career.
This shift, from sadness to control, actually reveals an important truth.
The current debate about AI and work has become polarized into two camps:
One side is the doomers, who believe we will all be replaced.
The other side is the boosters, who think everything will be fine.
But both miss the real feeling.
The reality is much messier.
You can feel both awe and sadness at the same time, mourning your past self while rushing toward a new version of yourself.
And what I find more interesting than my own experience is the change I see in the people around me.
I now run South Park Commons, a community and venture fund gathering creators and builders pondering “what’s next.”
Through SPC, I see hundreds of engineers, entrepreneurs, and technologists facing this transformation in real time.
And I notice a recurring pattern:
The old methods of evaluating talent are failing.
One of our members recently conducted about 20 trial-period interviews for engineering roles—basically, a week-long practical test.
He found that:
Work experience and ability to adapt to AI tools are completely unrelated.
Another member told me he discovered that the true predictor of success is a “creator’s temperament,” such as:
Conversely, factors like:
-
Having FAANG (Facebook, Apple, Amazon, Netflix, Google) on the resume
-
Attending prestigious universities
have almost no predictive power.
A third member shared an even more astonishing story.
His company started designing interview tasks deliberately impossible to complete by hand.
This became a very clean screening mechanism.
You can quickly see who is genuinely using AI tools in their work and who has only read about them.
The difference in code output between the two is not 10%.
It’s nearly 10 times.
This might seem like a phenomenon specific to the software industry, but I believe it’s much bigger.
We are in the midst of perhaps the largest transformation in knowledge work in history.
And the most important trait isn’t:
But: how a person faces change.
It’s not whether they’ve experienced change before, but whether they actively embrace it.
Many assume that younger people adapt more easily, and older people resist.
But the real dividing line isn’t age.
It’s personality.
The willingness to change seems to be an independent variable, crossing different ages and backgrounds, difficult to categorize simply.
I’ve seen engineers with over 15 years in the industry quickly master these tools and perform astonishingly.
At the same time, I’ve seen recent graduates treat AI as a concept to discuss, not an immediate tool to use.
As an investor, this realization has also changed how I select entrepreneurs.
Now, I’m most excited about those who can’t stop building.
Those who get restless when things stay the same too long.
Those who see new tools and immediately want to solve the puzzle.
I’ve started to understand this as the difference between a résumé and restlessness.
And I always bet on the latter.
Silicon Valley has long been considered a place that values ability highly.
But that doesn’t mean education and experience aren’t important here.
It just means they’re less important.
And now, they will become even less so.
Paul Ford recently wrote a brilliant article in The New York Times about how vibe coding could democratize software development, allowing more people to create.
I share this optimistic view.
But I want to add:
This democratization isn’t just about access to tools.
It’s about reordering how we value people.
We’ve spent decades building a culture that worships education and experience.
These things are valuable, but they’re no longer enough.
The new currency is:
Adaptability.
And unlike a Stanford degree,
everyone can have it.
If this shift has taught me anything,
it’s that I’ve come to once again understand what it feels like to be human.
Not in the romantic sense of “AI can never replace humans,”
but in a more uncomfortable way:
You must let go of your current self to become the possible future self.
This has always been the hardest part.
Even before AI appeared.
But now, technology makes it impossible to ignore.
This article was first published in The Information.
Aditya Agarwal is a partner at South Park Commons.
He was formerly Dropbox’s CTO and one of Facebook’s early engineers.
