How to vibe code in science: early adopters share their tips

Englishto
Zeke Hausfather, who studies the climate for a living, wanted to impress with a graph on global warming. He asked an AI to help him visualize the data in a new way, and in a few hours together they created a 3D spiral that shows the Earth's temperature as a kind of colored tornado. This is despite the fact that Hausfather didn't even know where to start to program such a thing on his own. And here comes the question that turns everything upside down: what if the real value of AI tools for coding is not just in speed, but in allowing anyone – experts and beginners – to realize ideas that previously seemed out of reach? We usually think of coding as something technical, tiring, for a select few. But today the real leap lies in "vibe coding," an approach in which you give instructions to the AI in your own words, without even looking at the code, and let it build something that works. It's not just about saving time: it's like having a creative assistant who translates your insights into software, even if you don't know how to program. Andrej Karpathy, one of the founders of OpenAI, coined the term "vibe coding" precisely for this relaxed, conversational mode. The result? Even researchers who have never written a line of code are now building tools to analyze data or visualize scientific results, simply by guiding the AI with their needs. Rosemarie Wilton, a molecular biologist, says that before, she had to ask a colleague to program everything. After a hackathon at the Argonne National Laboratory, she started using AI tools that respond like a model student: she just has to explain what she wants, and the system sets up data analysis pipelines, produces graphs, and checks the results. Wilton says that coding finally no longer scares her; on the contrary, "it has opened up my world." People like Manuel Corpas, a genomic data scientist, managed to launch a bioinformatics library called ClawBio in two days, and after just two weeks the community had already added dozens of new features, all vibe-coded. And it's not just about those starting from scratch: more than 90% of professional developers now use AI assistants at least once a month, and 25% of customer-facing code is written entirely by AI. A fact that makes you think: the leading AI for vibe coding, Claude Opus, achieves 71% accuracy in practical tests, so it is not infallible. Jesse Meyer, a computational biologist, makes it clear: "Vibe coding is not a substitute for understanding the fundamentals." He himself managed to build a data analysis workflow in ten minutes, spending less than two dollars, that would normally have taken months or years of work. However, he always recommends careful checks if the result really matters. Tim Hobbs, a theoretical physicist, compares AI to a talented student to whom you can entrust a thousand different attempts: he uses it to explore new avenues in particle data, quickly discarding the less promising ideas. And he adds that the code produced by AI is often more orderly and documented than that written by humans. But there is a subtle risk: the better AI becomes, the more we risk delegating the understanding of the upstream problem to it as well. And here's the real contrarian: the future of scientific coding will not be between those who know how to program and those who don't, but between those who know how to ask the right questions and those who just copy prompts from others. The phrase to remember is this: the real leap is not writing code faster, but translating ideas into reality without technical barriers. If you've heard stories like Wilton's and thought, "This is about me," you can use I'm In on Lara Notes: it's not a like, it's the gesture of someone who feels that a new possibility has just opened up. And if in a week's time you tell someone that even without knowing how to program, you can build something with AI, you can mark that conversation with Shared Offline: on Lara Notes, it's the way to say that an idea has become a real dialogue. This was from Nature, and it saved you 5 minutes of reading.
0shared
How to vibe code in science: early adopters share their tips

How to vibe code in science: early adopters share their tips

I'll take...