A major figure acclaimed by developers around the world, Amazon’s technology director delivered his latest advice to enable developers to adapt to AI, during his latest keynote on AWS re:Invent.
This is one of the most anticipated keynotes of this 2025 edition of re:Invent, the one which traditionally illuminates the months to come with a clear technological vision. In Las Vegas, where the JDN is present, the highly regarded Amazon CTO Werner Vogels engaged in a real transmission exercise: offering developers his latest advice for evolving, and prospering, in the age of AI, as he prepares to bid farewell to the re:Invent panel.
Will AI replace developers?
The end of one era, the beginning of another. For his last keynote, Werner Vogels chose to clearly answer the question that crosses the entire profession in the age of code assistants and augmented IDEs: “Will AI take my job?” His response is immediate: “Absolutely not. If you evolve.” The tone is set.
Werner Vogels recalls that developers have always had to adapt their skills. In the 2000s, we deployed on-premise. With the cloud, everything changed: we had to learn new architectural models, new tools, new responsibilities. Today, the arrival of environments like Cursor or Kiro is still pushing us to reinvent the way we code. “In ten years, everything will have changed again,” he says.
For the Amazon CTO, the transition is very concrete: with AI we spend less time writing code, and more time understanding, verifying and structuring it. Methods are evolving, we are no longer content with “vibe coding”, we are learning to write specs, to review the generated code and to maintain responsibility for what goes into production. His thesis is simple: the profession does not disappear, it changes. We are entering a “Renaissance” of development where AI serves as an accelerator. But the key point remains: these tools do not replace the work of the developer. They change the career, not the mission. “It has never been so interesting to be a developer,” assures the CTO.
Continuously learn and communicate
Continuous learning is becoming, according to Vogels, the founding skill of the modern developer: “Curiosity is not a personality trait, it is a professional discipline. Read, experiment, fail, start again. The developers who progress will be those who agree to expose themselves to change and test new tools before they become the norm.”
Communicating is also the other pillar. Werner Vogels reminds us that systems have become too complex to be understood individually: an engineer must know how to clearly describe what he wants, both to a machine and to a colleague. Specifications reduce ambiguity, improve designs, and limit errors. Likewise, code reviews, often perceived as a constraint, are once again becoming critical. They reintroduce human judgment into the loop. For the CTO, knowing how to explain an architecture, asking the right questions to a client, or clarifying what an AI agent must produce, becomes as strategic as knowing how to code.
Think in a system, take responsibility for your productions
Thinking in a system is now becoming an imperative. For Vogels, with AI, a developer can no longer think “function by function.” Each change, even small, modifies the overall behavior of an architecture. Adding a cache, adjusting a policy, moving a team responsibility: all create new flows. The CTO hammers home that the real skill today is to understand the ecosystem in which his code fits. Know how to anticipate side effects, read weak signals and reason in feedback loops.
Likewise, taking responsibility for one’s productions is the other non-negotiable point in resisting AI. AI can accelerate development, but it does not remove responsibility or requirement. “You build it, you own it,” he insists. The developer remains accountable for what goes into production, including when the code is 90% generated by an assistant. The CTO warns of two pitfalls: verification debt, created by a volume of code that is faster to produce than to understand, and hallucinations. The answer, he says, lies in concrete mechanisms: systematic human reviews and reinforced testing.
Becoming a polymath, an obligation
At this stage, the message leaves no doubt: AI does not replace developers, it raises the level of requirements. But Vogels doesn’t stop there. He insists on what he considers to be the foundations of the profession in the era of generative models: accepting experimentation as a method, and above all radically broadening one’s field of expertise.
But the most strategic point is still elsewhere. For Vogels, the defining skill of the coming years will be neither a language nor an AI tool, but the ability to think beyond code. The developer of tomorrow must become a polymath. Understand systems, user psychology, economic mechanisms, related sciences. In a world saturated with generative models, “transversal thinking” will become the ultimate competitive advantage in the market.
Finally, to close this ultimate masterclass, the legendary CTO takes care of his release. One last look at the crowd, a pithy “Werner out”, and he drops his remote on stage. An American-style drop mic that will go down in the annals of re:Invent.
