CLOUDY Podcast | #31 How AI Is Changing Web Development
- News
AI Is Shifting the Developer’s Role Toward Agentic Programming
Just a few years ago, AI was perceived mainly as an assistant that helped with searching for answers or auto-completing code. Today, however, development is moving toward so-called agentic programming, where AI can independently solve partial tasks, work with multiple tools, and understand the broader context of a project.
According to Branislav Ballon, this is an exponential evolution that fundamentally affects developers’ day-to-day work, while at the same time placing higher demands on their critical thinking.
The End of Classic Googling and the Decline of Stack Overflow?
One of the major turning points was the arrival of models such as GPT-3, which changed the way developers learn and solve problems. Instead of lengthy searches for answers on forums, developers today communicate directly with AI:
they explain the problem in their own words,
ask follow-up questions,
request alternative solutions.
Traditional resources like Stack Overflow are therefore losing importance, as large language models (LLMs) can synthesize answers from across the entire internet and present them in a more understandable way.
Does AI Increase Productivity? The Numbers Speak Clearly
According to a Stanford study (2025) based on a sample of more than 100,000 programmers, AI can increase a developer’s efficiency by 15 to 30 percent. AI delivers the greatest benefits in so-called greenfield projects (projects built from scratch) and when using commonly adopted programming languages.
However, Braňo points out an important “but”: with complex, older systems (legacy code), efficiency gains may drop close to zero or even turn negative, as AI can slow work down due to errors.
The conclusion is clear: AI is not a universal solution, but a powerful tool when used correctly.
Vibe Coding: Programming Without Writing Code?
One of the most interesting topics of the episode is Vibe Coding, a concept named by Andrej Karpathy. It refers to creating software using natural language, where a developer describes to the AI what the application should do, without writing traditional code.
This approach is particularly suitable for:
prototypes,
simple websites,
fast MVP solutions.
With more complex systems, however, it reaches its limits, because architectural decisions are made by AI without a deep understanding of their consequences.
Model Context Protocol (MCP): Another Major Step Forward
The Model Context Protocol enables AI agents to work with a broader context - including documents, designs, specifications, and external tools.
In practice, this means that AI:
understands the entire project,
can work with Figma, documentation, or APIs,
is able to make decisions based on real data.
This represents a fundamental shift in how AI is integrated into software development.
Risks and Challenges of AI
One of the key topics of the podcast is the impact of AI on critical thinking. There are concerns that excessive reliance on technology may weaken deep problem-solving skills, especially among junior developers.
“AI is an excellent servant when it comes to learning and brainstorming, but responsibility for architecture and critical decisions must always lie with a human,” emphasizes Braňo Ballon.
Are you interested in whether AI will replace full-time programmers, or how to properly educate the next generation of developers?
🎧 Listen to the full 31st episode of the CLOUDY podcast on 👉 Spotify, 👉 Apple podcasts , or watch it on YouTube👈.
