Perplexity with Claude 3 Opus for day-to-day questions; Replit (which uses either GPT-4o or GPT-4) for development.
If I weren't using Replit for software related tasks, I would only be using Claude models.
Discussion
is replit/gpt better for code? why do you use replit instead of gpt directly?
They have their own Copilot implementation and LLM chat built into the IDE, so it's a smoother workflow for me, and it's nice to only have to pay for one product to get everything. Context handling for prompts related to specific files or multiple files is also very simple; just select the files needed from a drop-down, and the LLM can analyze, edit, and drop-in code where needed and across several files. Deployment management is baked into the IDE as well (with a set amount of free cloud credits every month), and they have a mobile app, so I can code on the go whenever needed.