Android Wireless debugging is already a major pita to work with.
Requires a wifi connection to even work, so I need to carry a mobile wifi router with me, instead of having it bound to my secure WireGuard connection.
Next it assigns a random port number every time the wifi connection is interrupted, or when it considers now is a good time to reset the port number. It should be possible to set a stable port and and have it available, perhaps behind a few awkward UI/adb toggles.
I'm still wondering if this is something the big AI companies will come out with tomorrow, if it has business potential (it's extremely niche I think) or if I should put it on GitHub and build a community around it. This is what the Android app looks like:
https://imgur.com/a/qy8B2gthttps://imgur.com/a/C0eTPyn
I've been building out the Android app that works as the wrapper between codex and me this way.
ADB debugging access over WireGuard. I ask to add a local wake phrase detection for the phrase "Hey codex" to start dictation. Codex tells me it thinks Vosk or OpenWakeWord might be a good fit, I voice select Vosk.
Codex starts working. Codex deploys (updates) the app on my phone and restarts it and I hear the task complete response.
I say "Hey Codex" to test and ask for some tuning/improvements, or move onto the next task. All while keeping my phone in my pocket.
But there are countless projects I work on this way. Eg, I got an email from person X, it says it encountered a bug when doing x and y. Codex reads the email (using nvidia/gliner-PII to hide PII data) and investigates the reason for the bug and proposes a solution. I ask it to implement the fix on my dev server and increase test coverage. I enjoy my walk and after say 7 minutes get a an overview of the fix and can decide to deploy to production.
I have my phone in my pocket, no screen interaction is required. I use a headset (Shokz OpenComm2) with wind muff (when cycling).
I made an Android app that listens for codex turn-complete or intermediate updates and plays them back to me. My answer is transcribed and pasted back to the relevant codex (tmux) session on the server (which I can select by voice) a tiny layer helps with things like /new, /plan, answer selection, etc.
I do a ton of coding (codex) with a tts/stt wrapper. During walks, cycling, in the car. Not every task is suited to this style of interaction, but many are. Long form codex replies are condensed, code blocks are suppressed all in the name of making it work for tts feedback. So it works best on well defined projects with guardrails, where you know the agent can perform well.
Cursor was nice when I was still meticulously hand coding my stack, fantastic autocomplete. With today's top models, I barely write code myself, just review commits. Cursor eats Opus credits like there is no tomorrow. Composer has been a net negative in my experience. All in on Codex with GPT 5.5 on high using /fast.
I've been trying out various mobile, ai-assisted coding workflows.
Packing a Linux mini-pc in my rucksack, connected to display glasses, and voice-to-text with handy. Voice to text gets injected into a remote (Docker) codex session, running a hot reload web stack. I prompt to implement various features in an existing code base, where codex understands the structure and requirements.
If a feature is done, I take a moment to inspect the results on the display glasses, then move onto the next feature or keep iterating. It's not perfect, but I was able to implement a couple of not too complex features while walking my local national park.
The display glasses have a built-in 4-microphone array, and solid speakers. No need for a bulky headset or earbuds. Glasses come with monochromatic dimming, you can easily switch between dimming and see through.
If this comes with Linux integration, I will certainly give it a try.
I'm also using this project. Easy to configure and operate.
I am feeling a slight unease using such a recent project for things as important as the database. But the polished interface combined with the easy docker deployment made me use it anyway. Restores need some permission tuning on PostgreSQL but otherwise happy.
They are very proud of their github star acquisition curve [0], the "blessing" by Anthropic [1]
I've been using the Xreal One Pro display glasses since July last year. No more physical monitor display for me. They do feel a bit warm when the ambient temperature is over 25 degrees but overall very happy.
It's interesting that cloud providers are unable to provide stable S3 as a service. Hetzner is unable to deliver stable object storage, but given the article neither are OVHCloud and UpCloud.
The Linux terminal app on Android reddits are full of reports of instability. It is far from being useful as far as I understand. I had so much hope for this being a good way to use my phone as a portal for development, but it's a dud. At least we have termux and proot.
I think many saw this coming and it's probably a good thing. Personally I went with Kysely as a thinner abstraction with less risk of being acquired and turned into something that no longer aligns with my objectives. Still this is probably a nice boost to the major rewrite the Drizzle team is working on.