Archive · 18 Sep 2026
AWS Advocate Says Vibe Coding Will Continue Alongside Specs
AI tools can now help people write computer programs.
Darko Mesaroš of AWS says this does not mean developers will stop being important.
Developers still need to understand customers, system design, scaling and operations.
Vibe coding means working back and forth with an AI to explore an idea or build a quick prototype.
Another method, called spec-driven development, asks developers to describe the requirements and design before the AI writes code.
Mesaroš says both methods can be useful for different jobs.
He also says developers must check AI-generated changes because humans remain responsible for the software.
AI tools still have problems remembering past conversations and managing costs.
Mesaroš advises developers to learn both broad software skills and deeper technical knowledge.
AWS’s Darko Mesaroš says AI is changing coding, but software engineering involves more than writing code.
He describes vibe coding as useful for exploration, prototyping and experimentation rather than a practice that will disappear.
Mesaroš says spec-driven development creates a clearer contract between developers and AI agents about intended results.
He identifies persistent context, agent limitations and cost as continuing concerns for AI-assisted development.
AWS says it is expanding developer support in India through training, infrastructure investment and a planned Builder Loft.
- Who
- Darko Mesaroš, AWS Distinguished Developer Advocate, discussed AI-assisted software development and AWS’s developer initiatives.
- What
- Mesaroš said vibe coding will continue, while spec-driven development can provide a more structured way to work with AI coding agents.
- When
- During Mesaroš’s recent visit to India; specific dates were not provided.
- Where
- India, during an interview with Indianexpress.com.
- Why
- AI is changing how software is built, creating a need to balance experimentation with requirements, design, quality control and human responsibility.
Exploratory AI Coding
Structured AI Development
Best use of AI coding
Exploratory AI Coding
Vibe coding is useful for exploring ideas, building prototypes and experimenting through an unstructured back-and-forth with an AI agent.
Structured AI Development
Spec-driven development is preferable when developers need to ensure the implementation matches their intended requirements and design.
Relationship between the methods
Exploratory AI Coding
Vibe coding remains a legitimate part of software development and is not going away.
Structured AI Development
Spec-driven development adds a clearer contract between the developer and the AI agent and can help reduce low-quality AI-generated applications.
Responsibility for results
Exploratory AI Coding
AI agents can take on more of the coding workload and, with features such as memory, can learn from previous interactions.
Structured AI Development
Developers remain ultimately responsible for software changes and must approve AI-generated work.
- Speaker
- Darko Mesaroš, Distinguished Developer Advocate at AWS
- Structured approach
- Spec-driven development defines requirements and design decisions before code is generated.
- Vibe coding
- An iterative exchange with an AI agent used for exploration, prototyping and experimentation.
- AI tool
- AWS’s Kiro is an AI-powered integrated development environment focused on spec-driven development.
- Agent memory
- Mesaroš said Kiro Crew had learned 273 lessons from his interactions with it.
- Human oversight
- Mesaroš said every Kiro Crew pull request and software change had been approved by a human.
- India initiatives
- AWS cited more than $500 million for student cloud and AI training globally, more than 10 million people trained in India, and a planned AWS Builder Loft in India.