The end of coding as we know it
Nitin Gaur
AI~Digital Engineering & Solutions Lead | GenAI Consulting, Architecture, Thought Leadership
Recently, Anthropic CEO Dario Amodei shared his perspective on AI’s rapid advancements in coding. He suggested that in the next 3 to 6 months, AI could be writing 90% of all code—and within a year, it may generate almost all of it.
O'Reilly CEO also mentioned something similar a month ago in his blog post.
NVIDIA CEO has already said an year ago 'No need to learn to how to code'.
If AI reaches this level of automation, what does that mean for the role of human developers?
Will programming shift from writing code to reviewing AI-generated solutions?
How the programming has changed...
Before predicting the future, let's recap how did we get here in short span of time.
Most of the people in IT profession and software development must have heard coding assistant like GitHub Copilot by now. GitHub Copilot was made public in mid-2022 (not even completed 3 years!). Its adoption gained momentum as usage cost decreased and more features were added (newest one is the 'agent' mode).
Well...if you're thinking GitHub Copilot is the only option, here are few more...
? VS Code IDE extensions like Cline, Roo Code. These are extremely popular among developers as these are frequently updated with new features much ahead of GitHub Copilot. I used Cline and it is just amazing :-)
? IDEs like Cursor, Windsurf. These AI-native IDEs have taken coding experience to a next level. Cursor has just shattered every SaaS growth record without marketing (and a team of less than 20 people) by hitting $100M in recurring revenue at the end of 2024. Windsurf also claims to be first-ever agentic IDE.
? Terminal-based AI pair-programmers like Aider, OpenHands. These tools are an alternative for those who want to keep their IDE stable (rather traditional) and yet take advantage of AI-powered coding capabilities. These tools can make code change, create git commit which you can review in your IDE and continue.
? Full application generators like V0, Bolt. These are specially useful for product managers, business architects to quickly create a prototype application and demonstrate their ideas.
All these new ways of AI-powered coding, may overwhelm developers who are still unfamiliar with them. Eventually they must embrace these to survive and grow ??
How the programming will look like...
New terms are being coined to new phenomena people are experiencing...some of them are not what they actually mean (remember microservice or NoSQL), for example: Vibe coding, CHOP (chat-oriented programming), PDD (Prompt-driven development) etc. However, the fundamental truth of AI-assisted development will stay as new practices like?"conversational" programming?have emerged.
Addy Osmani has done great analysis (long posts, almost like a field guide on the topic!). He covers why we need to rethink our expectations, and what software engineering really is. He breaks it down into a 70:30 problem and provide a pragmatic guidance for maximizing the AI-assisted development in the realm of software engineering. Links in the reference section!
Conclusion
Everyone thought AI coding tools would make software engineering easier and that is still true. Future will belong to those software professionals who not only master these AI-assisted development tools but put greater emphasis on non-technical skills like critical thinking and problem solving. Writing good prompts in English language for their AI-pair programmer needs those skills.
Marco Argenti (Goldman Sachs CIO) went to the extent saying that engineers should study philosophy to code successfully in the age of AI. Whether developers need to take philosophy classes or not, the reasoning is sound. New ways to write code demand well-rounded developers in a post-GenAI landscape.
It's a defining moment in software development profession! Are you (and your team) ready?
References (made in this article)
Disclaimer: The views presented are mine