Self-taught 14-year-old vibe-codes tool that may cut down AI cost, now seeks YC funding

14-year-old Arjun Shah claims to have made an AI tool that can reduce token usage and make using AI cheaper. Now he is seeking Y-Combinator funding.Arjun Shah is a 14-year-old vibe coder who has now applied for YC funding. (Photos: Y Combinator, X/@arjunkshah21)14-year-old Arjun Shah claims to have made a solution that may solve one of the biggest pain points of current-gen AI: high cost of AI tokens. Through his supposed self-made AI tool, Shah, an Indian-origin boy from California’s San Jose who is founder of Supercompress, aims to cut down on the growing cost of AI tokens through calculated efficiency. Solving a big AI problem?You see, whenever you use AI tools, you consume tokens. Every time you use AI models, there is a lot of context. The more the context, the higher tokens are used. This has become a problem due to rising costs.Shah tells India Today Tech, “AI Applications have to send huge amounts of context, but the majority of it is redundant. However, they still have to pay to process all of that context, and it compounds massively over time.”The tool can potentially save millions of dollars for companies all over the world according to the creator. “It analyses the query and the context, figures out what's important, and cuts the rest. On average, it cuts by 65 per cent but keeps more than 98 per cent of critical content,” he explains. Supercompress is a neural network that Shah describes as a framework for AI. He adds, “It’s structured just like the human brain, with nodes and connections.”Shah says that Supercompress reads the full context, identifies what matters to a given query and removes the rest before sending it to the AI. In a demo video, the tool was able to compress context by 97.5 per cent. The compressed input got the same answer as the original input but was faster and consumed less tokens. Supercompress has about 150 users at the time of writing and is also available as a plugin for coding agents.Building at 7 and not wanting to learn codingShah says that he is no stranger to building things, originally starting to code when he was just 7. “I got into coding at age 7 with python. I really loved building. They taught us a bit at school, but I mainly experimented myself,” he says.You may think that creating a tool like this may need some good knowledge of coding, but Shah says that he is not a fan of technical work. He explains, “I quit coding at 9 because I hated syntax and didn't want to memorize everything, but my love of building never left.”Then came AI. With AI tools like Claude Code, it became easier to build things even without having to learn technical details. Shah says that two years ago, he started using the AI coding platform Cursor.“I built an AI-powered plant identifier for my school science fair with Cursor,” the teenager says. “Then I spent the whole summer teaching myself about neural networks and ended it by building my own.”Instead of learning to code, Shah says that he focuses on fundamentals. He explains, “I have not been learning to code, I simply educate myself on technologies. For example, instead of learning how to code a neural network, I learned about all the math and algebra and technologies behind them.”Application to Y CombinatorNow, the teenager is so confident that he has applied for Y Combinator, one of the most prestigious startup accelerators in Silicon Valley. Y Combinator is known for investing early in promising companies such as Airbnb and Reddit. If approved, Shah will receive $500,000 in funding, or roughly Rs 4.8 crore. He will also get access to YC’s network of founders.Though applications for YC are very competitive, there have been instances of teenagers receiving YC funding in the past. In 2017, OpenAI CEO Sam Altman approved applications from a 14-year-old and a 16-year-old when he was YC president.Shah says that his parents grew up in Gujarat before moving to the US. They are now US citizens. His mother has helped him throughout his coding journey and was co-founder of his previous project therooted.ai – a retrieval-based tool that references the Charaka Samhita and other traditional books to suggest remedies for everyday illnesses.- EndsPublished By: Armaan AgarwalPublished On: Jul 30, 2026 09:50 IST