Perplexity's engineers use 2 AI coding tools, and they've cut development time from days to hours

Perplexity AI is making waves in the tech industry by leveraging advanced AI coding tools, significantly enhancing engineering productivity. The company’s founder and CEO, Aravind Srinivas, shared insights at a recent Y Combinator event, stating that the implementation of AI tools has become a requirement for their engineering staff. This mandate typically involves utilizing tools like Cursor or GitHub Copilot, or a combination of both. Srinivas highlighted the remarkable reduction in prototyping time, noting that tasks which once took three to four days can now be accomplished in just about an hour. This efficiency isn't confined to complex algorithmic work; even team members without a technical background are successfully using these tools to make swift adjustments to user interfaces. He illustrated this by explaining a scenario where he provides feedback on an iOS app by simply taking a screenshot and suggesting changes. Those adjustments are then rapidly implemented using Cursor, showcasing the transformative power of these AI tools. However, Srinivas also cautioned that these tools are not without their drawbacks. While they expedite processes, they can inadvertently introduce new bugs, posing challenges that users may not know how to resolve. The growing trend of AI coding tools is evident across the tech landscape, with a recent report indicating that major companies like Visa and DoorDash are actively seeking candidates experienced with these technologies. A survey conducted by Jellyfish revealed that a staggering 90% of engineering teams have integrated AI into their workflows, a significant jump from 61% just a year prior. Of those surveyed, nearly half reported using multiple AI coding tools, indicating a shift towards a more varied and exploratory approach in software development. Despite the benefits, industry leaders have also expressed concerns about the potential downsides of relying on AI coding tools. GitHub’s CEO, Thomas Dohmke, noted that for seasoned developers, these tools might slow down their workflow, as they may need to provide feedback in natural language rather than directly in code. OpenAI co-founder Greg Brockman echoed these sentiments, suggesting that the current state of AI coding has relegated humans to the less enjoyable aspects of programming, such as code review and deployment, which many find far from fulfilling.