Dohyun Jung - Principal Consultant, ROBOCO

TL;DR

  • Even in the vibe coding era, what the best developers need at their core are fundamentals like curiosity, quality standards, and communication.
  • The more AI takes over implementation, the further a developer’s role shifts toward problem definition, evaluation, and accountable decision-making.
  • Agency matters more than intelligence, and an execution-first mindset like the Amazon Leadership Principles becomes the benchmark for collaborating with AI.

Introduction: The Traits of the Best Developers

Not long ago I came across a blog post on GeekNews by Matthias Endler, an open source maintainer and Rust consultant: “The Best Programmers I Know”. In it he lays out the traits the best developers he has worked with have in common.

The best developers are curious, humble, and know how to simplify complex problems. They communicate well, dig into the deep parts of the technology, and never stop learning steadily. And above all, they place enormous value on the quality and maintainability of software.

I broadly agree with him, but one thought stayed with me the whole time I was reading. In the era of vibe coding, where AI takes the lead in solving problems, how exactly do these traits of an excellent developer change?

Agency Over Intelligence

I found a hint toward an answer in a tweet from Andrej Karpathy. In a recent tweet, Karpathy made this interesting claim.

Agency > Intelligence. For a long time I believed intelligence was the single most important factor, but I have come to think that agency matters more than intelligence. Intelligence is possibility, nothing more than potential, while agency is the force that actually changes reality and produces results. In other words, no matter how smart you are or how much potential you have, it means nothing if you do not act.

This kind of agency will establish itself as the single most important virtue for developers in the era of working alongside AI, the era of vibe coding. That is because a developer’s role is shifting from solving problems to defining and evaluating them. But if we are talking about defining and evaluating problems, is that not already the job of a product manager (PM) or product owner (PO)? If a time comes when developer involvement is no longer needed, developers may genuinely disappear and leave only PMs and POs behind.

The Amazon Leadership Principles in the Vibe Coding Era

Over the past eight years I spent six as a technical trainer at AWS and two and a half as a developer. As the company grew rapidly and its headcount expanded exponentially, Amazon created the ‘Amazon Leadership Principles’ so that everyone could run the business with as close to the same mindset as possible. When I read Andrej Karpathy’s tweet about agency, my first thought was that agency connects very closely to Amazon’s leadership principles.

Let’s look at how the Amazon Leadership Principles apply in vibe coding specifically.

1️. Customer Obsession

  • You have to convey the customer’s problem to the AI accurately, and be able to define problems and set priorities with the customer at the center.
  • Provide clear, granular context so the AI does not miss the customer’s needs.

2️. Ownership

  • Since a human ultimately bears responsibility for what the AI produces, ownership over the output matters more than ever.
  • Treat the result as your own rather than simply the AI’s responsibility, and keep working to improve it to the end.

3️. Invent and Simplify

  • Use AI to boldly simplify existing complex procedures, and take on the role of inventing more effective ways to solve problems.
  • Enable continuous innovation by evaluating and improving the solutions the AI produces through a human lens.

4️. Learn and Be Curious

  • Continuously learn how AI technology is evolving and explore ways to use it with curiosity.
  • Pick up new frameworks and techniques quickly, growing yourself so you can collaborate with AI more effectively.

5️. Hire and Develop the Best

  • The work of finding and developing people with strong technical leadership remains extremely important.
  • Contribute by building up your teammates’ technical skills and teaching them how to use AI effectively.

6️. Insist on the Highest Standards

  • Do not trust AI output blindly; be able to evaluate and improve quality against a high bar.
  • Maintain uncompromising standards across every dimension, including quality, security, and ethics.

7️. Think Big

  • Use AI technology to hand off the technical implementation to AI, so that people can focus on setting and executing bolder, more innovative goals and visions.
  • Present the big picture so the AI can break past the constraints of existing thinking, and explore larger possibilities.

8️. Bias for Action

  • Use AI technology to experiment quickly, gather feedback, and iterate on improvements.
  • Do not just deliberate; pick up speed through action-oriented rapid iteration and validation.

9️. Frugality

  • Value cost and resource efficiency in your use of AI too, and design systems without waste.
  • Keep taking an efficient approach that gets the maximum result from minimal resources.
  1. Earn Trust
  • Share AI results transparently and communicate in a way that earns trust.
  • Present the AI’s decision-making in an interpretable form to increase confidence in the results.
  1. Dive Deep
  • Even when the AI’s output looks good on the surface, understand the underlying principles and analyze it closely before trusting it.
  • Hold an attitude of investigating all the way down to grasp the essence of the problem.
  1. Have Backbone; Disagree and Commit
  • Do not agree blindly with the AI’s output; state your own view clearly.
  • Even when opinions clash, once a decision is made, work responsibly with the AI’s results to produce the best outcome.
  1. Deliver Results
  • Ultimately, use AI to produce meaningful results in terms of actual business value or user experience.
  • Focus on delivering concrete, valuable results that go beyond technical excellence.

Among these, ‘Customer Obsession’ and ‘Ownership’ in particular are the heart of the drive Karpathy emphasized. AI can solve problems well, but defining those problems clearly and steering them in the right direction is still a human job. You have to understand the customer’s needs deeply and lead the problem-solving with a firm sense of responsibility for the solution. In the end, the excellent developer of the vibe coding era will be someone with clear agency and strong accountability, not just technical potential.

In the vibe coding era that has only just begun, developers will take on a role that goes beyond simply using AI: defining problems proactively alongside AI and taking responsibility for what gets produced. In other words, the developer who acts, who sees the problem accurately, takes it on as their own, and doggedly works toward a solution, will emerge as the most valuable talent of the vibe coding era.