Vibe Coding Is The Future

Founders describe a shift toward product engineering, where taste and user understanding matter alongside technical ability. One founder says he mostly thinks and reviews code now, while another says faster coding makes it easier to scrap and rewrite. Teams can also use multiple Cursor windows to work on separate features in parallel.

The founders say current coding tools struggle with debugging and often need explicit, step-by-step instructions. Some developers instead ask the model to regenerate code from scratch when fixing it seems slower. Diana says Cursor leads the surveyed batch, with Windsurf gaining ground because it can find relevant files across a codebase; founders also use reasoning models such as o1 and o3 for debugging.

One quarter of surveyed founders estimated that more than 95% of their codebase was AI-generated, despite being technically capable of building products themselves. The discussion suggests these tools can help people with backgrounds in fields such as math and physics become productive programmers faster. Engineers still need enough experience to read code, spot poor solutions, and debug.

The partners distinguish getting a product from zero to one from scaling it to millions of users. They argue that vibe coding can help founders ship early features, while growth may require systems engineers who can address architecture and performance. They point to Facebook’s custom HipHop compiler as an example of specialized systems work that followed rapid early development.

The discussion says engineering assessments should match the job: some roles call for building products quickly, while others call for systems design and deeper computer science knowledge. With AI tools, interviews may need to test code review, debugging, and the ability to judge generated code. The partners argue that becoming exceptional still takes deliberate practice and enough technical depth to evaluate both human colleagues and AI agents.