People who have followed my blog for a while know that I have been interested in working while walking, but software development has always required too much direct interaction with a computer for it to be a serious option.
That may now be changing for me because of two developments.
First, coding agents and language models are improving rapidly. They can solve more problems independently, require less guidance and produce better results than they could even a month ago. GPT 5.6 sol is amazing at completing tasks without having a human in the loop all the time.
Second, I recently migrated my entire development environment from my Mac to a powerful Linux server. It has everything I need: VPNs, Docker, and all my development tools. I can connect to it from anywhere from any device, including through my phone.
As a result, I can now do meaningful work from my phone. The agents I control from my phone have access to all the work resources I want -- something that isn't yet possible with regular Cursor cloud agents.
Some tasks still work much better on a computer. Testing desktop interfaces from a phone is inconvenient, and reviewing large pull requests is easier on a large screen.
At the same time, manually testing uis and manual code review are becoming a smaller part of my work. Instead of inspecting every implementation detail, I increasingly only focus on whether the right problem was solved and whether the architecture still makes sense. So not being able to do detailed reviews on my phone is not a big deal.
The main limitation is now the voice interface. Voice typing in the Cursor app is still pretty bad and requires you to constantly look at your phone.
ChatGPT's live voice mode shows how much better this interaction could feel. You talk into the microphone and an AI talks back through your speaker. The voice agent is designed to be indistinguishable from a real human: it can listen and speak at the same time, and will say stuff like "ehm, let me think for a moment" while it's secretly delegating that thinking to a more powerful model in the background.
If such a voice experience could be connected to my work agents, I can see myself walking outside all day with my AirPods on, being "on the phone" with my agents.
I am very enthusiastic about working while walking becoming possible, but I have doubts too about whether this is a healthy direction. With this setup, your work is always in your pocket so the distinction between work and private life is at risk of getting lost. On Hacker News, I read a comment arguing that walking outside should be a time to let your mind wander and enjoy your surroundings, rather than doing intense work while being absent-minded about the environment around you. There is something to that. Still, I am going to try it and see how it feels.