- World of AI
- Posts
- Andrej Karpathy: Vibe Coding Era
Andrej Karpathy: Vibe Coding Era
Andrej Karpathy talks about the future of vibe coding. The workflow is shifting, the jobs are shifting, and iteration speed is the new edge in software.
World of AI
December 23, 2025
In partnership with
Andrej Karpathy Talks About The Future
As 2025 comes to a close, Andrej Karpathy published his year-in-review on X — and paused to reflect on the term he coined back in February: vibe coding.
And this time, he went further than defining it.
Prediction: “Vibe coding will terraform software and alter job descriptions.”
That’s the claim that lit up developer conversations earlier this year, and it’s the claim people are still unpacking this week.
What Karpathy meant by that
He believes developer work is shifting:
Describe the outcome first
AI generates the first version of code
Developers test early, break safely, refine fast
The job becomes about direction and iteration, not repetitive typing
His prediction is simple: the workflow changes the work, and the work changes the role.
Why the industry reacted so strongly
The term spread because it captured something real developers already felt:
Models are getting faster
Testing loops are tightening
Multi-file updates through AI are becoming normal
Writing every line manually is no longer the highest-value part of the job
He didn’t name a tool.
He named a transition in how teams build software.
AI that works like a teammate, not a chatbot
Most “AI tools” talk... a lot. Lindy actually does the work.
It builds AI agents that handle sales, marketing, support, and more.
Describe what you need, and Lindy builds it:
“Qualify sales leads”
“Summarize customer calls”
“Draft weekly reports”
The result: agents that do the busywork while your team focuses on growth.
What developers are doing with vibe coding now
This week, serious builders are treating AI code generation as:
a starting point
something to iterate on quickly
something to test in sandboxes or local servers
a way to handle large code changes without rewriting everything themselves
The devs benefiting most are the ones who can:
steer ideas clearly
validate them critically through tests
iterate confidently across files and failures
That’s the real advantage.
Why this matters heading into 2026
This trend points toward a new expectation forming in software teams:
You’re not judged by:
how much code you can type
You will be judged by:
how fast you can improve systems
how confidently you test and refine
how well you express direction and intent
how comfortable you are steering AI through failures and fixes
This isn’t the end of coding.
It’s the end of coding as the only starting point.
What to watch next (next 2–3 weeks)
Developers are tracking:
How AI-generated code improves through iteration
Whether teams adopt vibe coding for internal copilots
How job expectations evolve
How far this workflow shift pushes into real engineering teams
The conversation is still early.
But the direction is locked in.
🚀 Bottom Line
Karpathy’s real message:
Describe clearly → Test fast → Improve confidently.
The workflow is shifting, the jobs are shifting, and iteration speed is the new edge in software.
Reply