The startup world is a risky business, and sometimes, those risks pay off in unexpected ways. Take the story of Sarthak Dhawan, a 21-year-old co-founder of Turbo AI, who, in a single month, accidentally spent a staggering $30,000 on AI tokens. This might seem like a costly mistake, but Dhawan sees it as a valuable lesson in the fast-paced world of AI-powered startups.
The Fast-Paced Learning Curve
Dhawan and his co-founder, Rudy, launched their AI learning tool app in January 2024 while still in college. They dropped out to pursue their business full-time last year, and their focus has been on shipping speed. In Dhawan's words, their bottleneck is shipping speed, not the cost of AI tokens. Slowing down to manage token spending would have cost them momentum, and that's a risk they were willing to take.
The AI landscape has evolved rapidly. When they started coding the app, AI code-generation models were rudimentary. Dhawan wrote the code himself from scratch. Now, the task of engineering has transformed, and AI has become an indispensable tool. Dhawan's day-to-day involves planning and reviewing AI-generated code, a process he calls 'vibe-checking.' This shift has led to concerns about the atrophying of coding skills, as engineers become reliant on AI.
The AI Shift and Its Implications
Twenty years ago, a group of engineers deeply understood the ins and outs of their codebases. They wrote the code themselves, making architectural decisions with full knowledge. However, as AI takes over the coding process, the codebase becomes an entity that engineers may not fully comprehend. This raises questions about the future of software engineering jobs and the role of human expertise.
Budgeting and Flexibility
Turbo AI's approach to AI spending is flexible. They don't have a set budget for AI token spending, instead keeping a loose eye on it. This flexibility has allowed them to adapt to the evolving needs of their team. With about 10 people on their team, costs have crept up as more AI is used, but they're fine with it as long as it drives output. Their average monthly AI tooling costs for software development are around $20,000.
The Easy Fix: Fast Mode
In April, Turbo AI's AI token bill jumped to $30,000, partly due to Dhawan's unintentional use of fast mode in Claude. This setting, while making the AI feel faster, significantly increases token costs. The team quickly identified and fixed the issue, turning off fast mode when it wasn't necessary. This simple adjustment highlights the importance of awareness and vigilance in managing AI spending.
Balancing Token Savings and Productivity
When it comes to saving tokens, Turbo AI takes a pragmatic approach. They prioritize easy wins, such as defaulting to standard mode and using lighter models for simple tasks. However, they don't stress over every dollar. With a strong focus on productivity and revenue growth, they believe that AI investments are worth it in the long run. This mindset has helped them cross $13 million in lifetime revenue this year.
The Future of AI-Powered Startups
Dhawan's story is a testament to the challenges and opportunities in the AI-powered startup ecosystem. As AI continues to evolve, so must the strategies and mindsets of entrepreneurs. Turbo AI's journey is a reminder that embracing the risks and learning from mistakes can lead to significant growth and success in the tech industry.
