A quarter of startups in the current YC group has a symbol rules that are almost completely created

With the release of new artificial intelligence models that are better in coding, developers are increasingly used artificial intelligence to create code. One of the latest examples is the current batch of Y Combinator, Silicon Valley Start. “A quarter of the W25 start starting batch, A conversation posted on YouTube.

Friedman said that this number 95 % does not include things like the code written to libraries, but taking into account the code written by humans compared to the prosecution.

“It is not as if we were financed by a group of non -technician founders. Each of these people is very technical, and they are fully able to build their own products from scratch. A year ago, they had built their products from scratch – but now 95 % of them were built by artificial intelligence.”

In a video entitled “VIBE CODING is the Future”, Ferdeman, along with the CEO of YC Garry Tan, Harj Taggar, and the general partner Diana Hu, discussed the direction of using language and natural instincts to create software instructions.

Last month, the former head of Amnesty International in Tesla and the former researcher at Openaii, Andrej Karpathy Description of the term “coding” To describe a symbol method using LLMS models without focusing on the code itself.

The code generated from artificial intelligence is far from perfection, through. I noticed studies and reports Some code created from artificial intelligence can include safety defects in applicationsand The cause of interruptionOr Makes make mistakesForcing Devs to change the code or correction significantly.

During the discussion, he said that even if the product builders depend heavily on artificial intelligence, one of the skills that they should have to be good is to read the code and find errors.

“You must have enough taste and training to know that LLM spit bad things or good things. In order to do good” coding “, you still need to have a taste and knowledge of the ruling on good for bad.”

Tan also agreed to the founders who need classic coding training to maintain long -term products.

“Let’s say that an emerging company contains a 95 % artificial intelligence code (in the market), and a year or two, they have 100 million users on this product, does it decrease or not? The first versions of thinking models are not good in correcting errors. So you have to go deep into what is happening with the product.”

VCS and developers were enthusiastic The coding in which artificial intelligence works. Startups Blitand Codeand Indicatorand lovableAnd charm It has collected hundreds of millions of dollars in financing in the past 12 months.

“This is not a heresy. This is not going away. This is the dominant way of the symbol. If you don’t do it, you may leave behind.”