Programmers have embraced AI to aid their productivity. But how should they adjust to really benefit? What skills are required for a successful relationship with AI?

According to Anthropic CEO Dario Amodel in the future code will be entirely AI-generated and he has predicted that AI could generate up to 90% of all code within six months. This prompted the Replit founder and chief executive Amjad Masad to give the advice: "I no longer think you should learn to code,"

More important are the words that follow:

"Learn how think, how to break down problems ... to communicate clearly ... with humans and with machines."

Breaking down problems is a skill that many of us learned through learning to program a computer. Once you understand the fundamentals of writing a program - functions that tackle more and more specific tasks, then it is natural to start seeing problems in terms of a flow chart.

The need for clear communication has become apparent as we start to interact with LLMs such as ChatGPT, Gemini and Anthropic's Claude. In order to make the most of their assistance we need to to take care with framing appropriate questions. This requires a new skill - that of prompt engineering. It's name indicates that the idea is the the question itself provides information that leads logically towards the answer, the danger being that an overdone prompt can land you with having nonsense confirmed as fact. Increased use of chatbots should reduce their tendency to "hallucinate". Something we hear about less and less.

The rise of artificial intelligence had led CEO of GitHub Thomas Dohmke to the alternative conclusion, included in a recent podcast interview:

"I strongly believe that every kid, every child, should learn coding .. in school, in the same way that we teach them physics and geography and literacy and math and what-not."

In the same interview, Dohmke said that going forward:

"It's so much easier to get into software development. You can just write a prompt into Copilot or ChatGPT or similar tools, and it will likely write you a basic webpage, or a small application, a game in Python. And so, AI makes software development so much more accessible for anyone who wants to learn coding."

You might think that this "democratizing" effect of AI would be detrimental to established programmers and threaten their jobs. However, in an earlier blog post, Dohmke paints a more positive picture, writing:

"AI won’t eliminate developers – it will empower them to tackle new, more complex challenges. Through machine learning models that generate or suggest code, developers can accelerate mundane tasks and experiment faster. AI agents can sift through documentation, identify bugs, and even propose architectural improvements. This emerging paradigm lets developers think more strategically about the user experience, performance, and reliability, rather than obsessing over syntax and rote implementation.

He goes on the outline the developer's role in a future where code is co-written by AI agents: 

The human sets the overarching goal, determines constraints, ensures ethical considerations, and divides the work into small chunks that can be handled by the state-of-the-art model, while the AI agent takes on the grunt work of writing, testing, and refining large swaths of code.
The developer, in essence, becomes the conductor of an AI-empowered orchestra."

The concomitant change is that more can now be achieved by small teams. As already claimed by Garry Tan, President and CEO Silicon Valley incubator Y Combinator, using AI-assisted coding, aka "vibe coding":

"allows 10 or so engineers to build what would've once required the efforts of "50 or 100."

The term "vibe coding" was coined by Open AI co-founder Andrej Kaparthy in a post on X that went viral:

There's a new kind of coding I call "vibe coding" where you fully give in to the vibes, embrace exponentials and forget that the code even exists. 

In this Lightcone Podcast, with the title Vibe Coding Is The Future, Gary Tan and his Y-Combinator partners Jared Friedman, Harj Tagger and Diana Hu look at this new way of programming and the impact it is already having on startups. 

Reporting on the results of a survey of  the founders in the current YC batch the video includes quotes expressing their overwhelmingly positive take on Vibe coding and the "crazy" finding that a quarter of the founders claim that more than 95% of their code base was AI generated. 

Noting the limitation of current AI Assistants are that they are not good at debugging, the skills of being able to read code is one that is still required.

Signing off, Gary Tan tells us:

Vibe coding is not a fad - it's time to accelerate.

So with the help of AI we can expect to achieve more in a shorter time. However, you need to be just as good a programmer as you ever were, if not better. The key skills will be ability to breaking down problems into manageable chunks, which is a natural part of orchestrating a computer program, and the ability to communicate with both other humans and with machines in language that although characterized as "natural language" needs to be both precise and accurate. 

Comments

Make a Comment or View Existing Comments Using Disqus

or email your comment to: comments@i-programmer.info