been doing this for over a year with termius and tailscale to a server with grapheneos. every time a new product tries to do the same stuff

TBH I'm not sure this is where dev is going. It's pretty annoying to have to read what can be pretty verbose LLM output on a 5 inch screen.

It's useful for certain tasks if you don't have to read too much and just need to give approval or review relatively concise output etc but otherwise I've found it to be a pretty terrible experience. I stopped doing this after just a couple days as a result.

An exception might be if you don't have access to your computer for hours or days at a time in which case I'd probably be willing to put up with coding on a 5 inch screen but otherwise I'm super curious how other people are able to be productive with such a small screen.

Only if you don't care about output at all.

Even with my eyes constantly in the logs window, credo + ex_dna + ash_credo [1] and looking at all the output, I still miss a bunch of shit that LLMs produce. Can't imagine just ssh'ing from a phone and letting an LLM lose (though I've tried it once or twice).

[1] Tools for Elixir to catch various code smells:

- Credo https://hexdocs.pm/credo/overview.html

- ex_dna https://hexdocs.pm/ex_dna/readme.html

- ash_credo https://hexdocs.pm/ash_credo/readme.html

It honestly feels like a segue to just talking to the agents that won't really have much of a long time horizon. People who either prefer or are professionally required to maintain control over software aren't going to be working from a 5-inch screen while people who use the iPhone as a dev environment just want a simpler interaction with their agentic fleet for creation.

Ad. Title of page is "Onepilot: The mobile SSH & AI agent IDE"