AI search tool turns EAS build query into 2-hour detour through 2019 env vars hell – enter the age of mandatory prompt engineering

A recent experiment with Perplexity AI, a search tool, highlighted the importance of context engineering when using artificial intelligence to find information. The user attempted to learn about EAS Builds, a feature of Expo, a popular framework for building cross-platform apps, but was met with confusing and unrelated results. The AI tool failed to carry context across separate queries, instead treating each question as a fresh retrieval task. This led to irrelevant results, including a 2019 article on GitHub source code management. To overcome this limitation, the user found that pinning the domain and providing specific details, such as naming concrete artifacts like eas.json and explicitly excluding irrelevant meanings, improved the accuracy of the results. This experience underscores the need for users to understand how AI tools work and to provide clear and specific context to get relevant results, a key consideration for professionals working in the tech industry, particularly in fields like app development and software engineering.