I had a positive experience reading Nikhil Gupta's article on Instagram system design, but I am skeptical about two key decisions Nikhil made. First, I wonder if DynamoDB is the best choice for user feeds. While DynamoDB is scalable and performant, there may be other databases or technologies that could be more suitable for this particular use case. Second, I question whether Redis caching is necessary in this context. While Redis caching can provide significant performance improvements, I wonder if DynamoDB caching would be sufficient and more cost-effective. Ultimately, it's important to consider the specific requirements and constraints of the system and continuously evaluate and optimize the system architecture as needed. Nonetheless, I appreciate Nikhil's thorough explanation of the Instagram system and look forward to reading more from them in the future.