Hi! Ive been writing code since i was 8 and still love doing that. I use replit mostly when im trying to get my 1-year old son to sleep. This activity leaves me one free hand to do some coding! Ive tried many apps and methods to do coding on iphone, but replit is the best one i found so far.
I find agent (and AI in general) an interesting product, but so far it hasnt been useful to me. Everytime i ask it to build something, it builds the boring parts pretty well: login screens, database integrations, html forms. But some aspect will not work correctly. When i ask it to fix that, it breaks something else. And now im left debugging ai generated code. No thanks
So I’m mostly focussed on the non-ai parts of replit. They are mostly fine, although sometimes i think they are also ai-generated: scrollbars that dont work. Buttons that disappear after an update.
My biggest current gripe is that the shell restarts everytime i switch to it, and forgets all history.
So my cry for the replit team is: please dont forget us lowly programmers that just want to keep smashing the keyboard instead of letting the AI have all the fun!
1 Like
I have done coding off and on but I find you comment about breaking the code each time it’s fixing something else! The shell restart and constant port issues are a lot of my checkpoints!
The key with using ai is planning, documentation, and clarification. The best way to always start is to know exactly what you want to create. Write it down thoroughly for documentation. TELL replit agent your IDEA and allow it to create your 1st initial build. Give documentation to replit AFTER it creates your initial build. Now it has a clear plan. Have it create all files necessary to create YOUR IDEA. WATCH THE MAGIC.
Hi @richmans I feel the same. I’ve been using replit on and off for many many years. I use it for trying out different technologies and languages, and agree that it’s the only usable coding tool for iPhone. (Because of my usage profile the assistant is irrelevant) The fact that the shell restarts on mobile even if I just click away briefly makes it very hard to use the shell and also makes using language REPLs similarly limited. I find that using a “main.py” or “main.lua” file and putting a sequence of lines in there the most effective approach. (Can change the run parameter in .replit) Then just click run to see the result. I really hope that REPLIT will keep some focus on mobile!
I feel you on the agent always breaking things and needing debugging. I’ve found that using Assistant for basically everything after initial agent build is the way to go with the AI. But, I do find myself manually debugging and troubleshooting a LOT. And frustratingly, it’s all stuff that if it were programmed right, should be able to see in logs and console on it’s own and continue to work through the issues. But, instead, it says “Hey if this looks good I can start working on more stuff!” when the server sits there errored out and unable to start 