The latest advancements in AI models have led to increased adoption of AI-generated code among developers. A notable example is the current batch of startups at Y Combinator, a renowned Silicon Valley startup accelerator. According to YC managing partner Jared Friedman, a quarter of the W25 startup batch have 95% of their codebases generated by AI, as mentioned in a conversation posted on YouTube<\/a>.<\/p>\n Friedman clarified that the 95% figure excludes code written for importing libraries and focuses on the comparison between human-typed code and AI-generated code.<\/p>\n \u201cIt\u2019s not like we funded a bunch of non-technical founders. Every one of these people is highly technical, completely capable of building their own products from scratch. A year ago, they would have built their product from scratch \u2014 but now 95% of it is built by an AI,\u201d he said.<\/p>\n In a video titled \u201cVibe Coding is the Future\u201d, Friedman, along with YC CEO Garry Tan, managing partner Harj Taggar, and general partner Diana Hu, discussed the trend of using natural language and instincts to create code.<\/p>\n Last month, former head of AI at Tesla and ex-researcher at OpenAI, Andrej Karpathy described the term \u201cvibe coding\u201d<\/a> to describe a way to code using large language models (LLMs) without focusing on code itself.<\/p>\n Code generated from AI is far from perfect, though. Studies and reports have observed that some AI-generated code can insert security flaws in applications<\/a>, cause outages<\/a>, or make mistakes<\/a>, forcing devs to change the code or debug heavily.<\/p>\n During the discussion, Hu said that even if product builders rely heavily on AI, one skill they would have to be good at is reading the code and finding bugs. <\/p>\n \u201cYou have to have the taste and enough training to know that an LLM is spitting bad stuff or good stuff. In order to do good \u2018vibe coding\u2019, you still need to have taste and knowledge to judge good vs bad,\u201d she said.<\/p>\n Tan also agreed on the point of founders needing classical coding training to sustain products in the long run.<\/p>\n \u201cLet\u2019s say a startup with 95% AI-generated code goes out [in the market], and a year or two out, they have 100 million users on that product, does it fall over or not? The first versions of reasoning models are not good at debugging. So you have to go in depth of what\u2019s happening with the product,\u201d he suggested.<\/p>\n VCs and developers have been excited about AI-powered coding<\/a>. Startups including Bolt.new<\/a>, Codeium<\/a>, Cursor<\/a>, Lovable<\/a>, and Magic<\/a> have raised hundreds of millions of dollars in funding in the last 12 months.<\/p>\nUnderstanding AI-Generated Code<\/h2>\n
Technical Founders and AI-Generated Code<\/h2>\n
The Concept of Vibe Coding<\/h2>\n
Definition of Vibe Coding<\/h2>\n
Limitations of AI-Generated Code<\/h2>\n
Importance of Code Review<\/h2>\n
Classical Coding Training<\/h2>\n
Sustaining Products in the Long Run<\/h2>\n
Debugging and Maintenance<\/h2>\n
The Future of AI-Powered Coding<\/h2>\n
Conclusion<\/h2>\n