Cool - I can imagine myself using this. We usually use a local supabase for developing and playwright for UI validation tasks, could your sandbox accomodate this too?
Yes, If the low code solution has a good interface for LLMs. There are enough low code / no code solutions with a pure graphical interface and nothing else. Dinosaurs.
It will just iframe whatever page/app you would have been browsing anyway but potentially with ChatGPT directly being able to operate on the App state.
So if configured, I guess ChatGPT will be just a handy middle layer to your usual interfaces.
Nice one, I thought I need this and wanted to build something like that too couple weeks ago.
But the more I worked with Claude, I felt like *I am the bottleneck* and not the waiting times. Also waiting for more than (really max) 5 minutes is for my features just not happening.
I think remote claude code is nice if you start a completely new app with loads of features that will take a long time OR for checking pull requests (the remote execution is more important here)
I only know the old Github Copilot (like 2yrs+ ago) so cannot speak to it directly, but even the Cursor Agent (with Sonnet 4 or GPT-5) is IMO inferior to Claude Code (CC).
In my experience, it is faster and better performing.
CC seems to spend tokens more deliberately + gives superior coding tools to the model than other provider.
Recently my CC subscription ran out, tried 3 prompts with Cursor Agent and then went back to subscribing CC. I still use Cursor though for autocompletion.
It might look like you initially, but then some sites might block you out after you had some agent runs. I had something like this after a couple local browser-use sessions.
I think simple interactions like natural cursor movements vs. direct DOM selections can make quite a difference for these bot detectors.
When using spec writter and sub-tasking tools like TaskMaster, Kiro, etc. I've experienced Claude Code to take 30-60+ minutes for a more complex feature
I think it's less about the code output, but about the process of humans iterating and adjusting the LLM-drafted requirements and design.
Claude Code et al. are good enough, the bottleneck is IMO usually the context and prompt by now. So further improving that by optimizing for and collecting data about the human interaction seems like a good strategy to me.
Essentially, the user labels (accept/edit) data (design documents) for the agent (amazon)
I agree that the OS vendors are in a great position to add value via broad, general purpose features. But they cannot cover it all - it's breadth over depth. So I think the innovation for niches and specific business processes will be still owned by specialized 'GPT Wrappers'.