“The speed you can get compared to programming yourself is just crazy.” 3 vibe coding tips from top software engineers

Business Insider spoke with software engineers who say that vibe coding will not harm the Software Engineering profession, but rather the opposite — it can become a useful technique.

Catch the vibe

According to them, a good start is to learn and master vibe coding tools such as ChatGPT, Cursor, Replit, and Windsurf.

Marc Tuscher, CTO of Sereact, an AI robotics company, said that vibe coding tools make his work «a lot faster,» which is why he encourages his team of 25 software engineers to use them.

While Tasher acknowledges that these tools can make mistakes that need to be fixed, he says a software engineer who works with AI tools will have an advantage over someone who doesn’t. «The speed you can get compared to programming on your own is just crazy», — he says.

Tanay Kothari, CEO of Wispr Flow, a voice agent that can be used to input commands into vibe coding tools like Cursor, notes that there can be some reluctance among experienced developers to use such tools.

However, in his experience, engineers who think they are «just really good» will start considering vibe-coding tools when they realize they improve efficiency. «People who use AI tools get about twice as much done,» he said.

Improved prompt handling

By learning to give text prompts to vibe coding tools in a more intelligent way, engineers can also gain an advantage.

«The most popular new programming language is English,» joked Tesla’s director of artificial intelligence and computer vision for autopilot, Andrey Karpaty, in January 2023, just a few months after ChatGPT’s first public release.

The large language models that underlie modern generative AI tools can be, as Karpaty puts it, «programmed» inside a prompt. That is, a carefully constructed prompt can radically change the quality of the results of an AI system.

Ash Edwards, a former Palantir engineer who became CEO of Fern Labs, an AI agent development company, believes that software engineers get more out of vibrational coding tools if they are more «directive.»

«The real failure is almost letting the code choose its own direction. I think you can usually get much better results if you know what you want to do and roughly know how you want to build it», — he said.

Clear thinking

With artificial intelligence taking over much of the routine work of software development, what should engineers focus their extra time on? Software development experts say programmers should focus on a clear vision.

When software engineers create programs and applications, they spend a lot of time carefully thinking about what they are creating and how it might relate to the larger system. All of this requires careful thought before any lines of code are written.

According to Wispr Flow’s Kothari, one type of thinking that helps this process is reasoning from «first principles,» which is breaking down a complex problem into smaller parts to find a suitable solution.

«This is a very complex, high-level issue that you need to be able to solve no matter how good the AI ​​gets,» he says.

Kothari notes that this is a key attribute of what defines a high-performing engineer who is 10 times more productive than their peers. «The 10x comes from clarity of thought,» he added.

Thinking about systems as a whole is another useful way to approach software engineering problems. Kothari notes that people can waste a lot of time if they don’t do this, as it can lead to building «the wrong thing that no one wants.»

Edwards of Fern Labs agrees that engineers «will always need to be able to think about systems» — despite the trend toward using vibe-coding tools — because at the core, they will still be the ones who have to control the program.

Recall that the vibe coding method, based on large language models (LLMs) from companies such as OpenAI and Anthropic, is attracting attention because it can potentially lower the barrier to entry into the field of software development.

But questions remain about whether this approach can reliably produce code suitable for real-world applications, even as tools like Cursor Composer, GitHub Copilot, and Replit Agent make the process increasingly accessible to people with minimal programming skills.

Read the country’s main IT news in our Telegram

On the topic

Read the country’s main IT news in our Telegram

«Programmers are becoming process managers, not routine task performers.» AIT people discuss AI-powered vibe coding in Threads — most are skeptical

On the topic

«Programmers are becoming process managers, not routine task performers.» AIT people discuss AI-powered vibe coding in Threads — most are skeptical

Dima Maleev on the impact of AI on IT: «In a few weeks we will see more and more „vibe coders“ launching startups and receiving millions in funding»

On the topic

Dima Maleev on the impact of AI on IT: «In a few weeks we will see more and more „vibe coders“ launching startups and receiving millions in funding»

Vibe coding is becoming increasingly popular in IT — programmers simply ask AI to perform certain tasks and do not delve into the actual code that the chatbot offers.

On the topic

Vibe coding is becoming increasingly popular in IT — programmers simply ask AI to perform certain tasks and do not delve into the actual code that the chatbot offers.