Today I discovered that Perplexity even has trouble understanding the more complex prompts. But that was for only one style of prompting. Other than that it worked really well. For me personally, since being introduced to it, Perplexity is the most efficient LLM I have used. One big flaw I have realized is how bad Perplexity is to following directions. I have had it follow directions for about 4 prompts until it resorts to a different method, I would then tell it to go back to my directions. So, for me that is my failure of the blog post. I still have high hopes for it as the LLM develops through time. Since being introduced, Perplexity has been more efficient in following directions, though still lacking behind other models.
For me Perplexity has been the most effective for helping me with my course work. Mainly for helping me build a guide for papers or schedules. I started using Claude to assist me write code for my CSci project. Then I started only using ChatGPT for day-to-day questions and finding articles for sports and fitness
Perplexity is really good for summarizing. I use it most of the time when I have to summarize some article. The downside is that other than summarizing, using Perplexity for other purposes is not optimal, especially for solving math problems or generating code.
I heard people say Perplexity is a pretty good AI app, but I haven’t used it yet. Maybe I’ll give it a try after hearing your opinions!