T lol Perplexity Edu can be so mean, responses like "Reality check.." "Sanity check.." "Here's where you're wrong..." however sometimes i get the "You're actually right here, my apologies. Lets see where I went wrong.." and the occasional "I think you're onto something." I've also shown Claude and Perplexity responses to each other and they DO NOT always agree, but see each other's perspective.
I think I'm no longer going to consider myself a programmer. I have fallen in love with coding with AI. Honestly did not think that would happen, but I'm having so much fun just being able to build shit. Building everything I've been wanting to for ages including tools for myself that I deemed way too complex and time consuming only to have AI build it in like a few days.
It's scary good where this is going now...
As it stands now, definitely an advantage to know what's going on. I saw the kind of stuff my brother made... I dont even know how to prompt ai that badly to do some of the stuff it did. Just being able to prompt it right makes quite a difference. Or asking it to do things you know how they work, vs how it looks is massively beneficial. Like I saw a big matrix checkbox thing my brother made and said "when I check something, I want the table to save automatically" so it literally replaced the entire table everytime something was ticked rather than updating that 1 thing. And he asked why it was so laggy when he tried to do it fast. Or the login credentials in plain text in the html thing I sent before.
But some things I did notice it doing was like when I had something such as "Days before expire", it seemed to store that in the database as an integer. Now if I were to ask someone who doesn't know how to code how they'd solve it, they'd probably ask the ai something like "update the database every day" instead of rewriting the logic to have the expirey date stored and the days left be calculated on the fly. Or like how I noticed it creating css for individual pages so I had to get it to write classes to use across pages. Or turning certain things into more reusable component structure...
But I have also let it off to do things with zero oversight for me. Like when I was trying to get a server running on my phone, I kept hitting issues with wrong version, bad security, android forced shutdown and I just told claude to go fix it and tell me what was wrong and what it did. Looking at the history of what it did by getting the error, checking what it meant, looking for fix, applying it, getting new error etc, it was literally everything I would've had to do, but like no effort and way faster. Such a freeing moment tbh.
B Depends what I'm doing. I use claude and grok for different use cases. Like claude is more the designer and better at doing the stuff I'm more fuzzy in my thinking about and gives great suggestions. It can build the things I'm thinking filling in the blanks of things I dont need to say or didn't even think of, and more often than not, actually things I like. Grok is my workhorse and orchestrator managing sub agents doing exactly what I tell it to fast and efficiently. I usually have it read the reports claude provided and let it build accordingly. Then I go back to claude for direction in new features and stuff. Also I have them both setup to reply with concise to the point replies already. Any more condensed would probably be hard for me to read. Still getting used to using them my way before I get more into using them the "proper" way.