Personally, my recent and surprisingly greatest win was to set up my old phone (samsung S21) with the addictive apps, removing them completely from my iPhone.
Quite literally "cold-turkey'ed" from 4.5-ish hours/day to 2 hours a day in a single day, consistent over the last few weeks.
I set up my second phone with a custom homescreen, and installing the 'bad' apps on there (Instagram, Youtube, NYTimes in particular). I dont use it for other apps.
Now if I want to scroll, which I still do sometimes, I have to walk to a specific chair next to which my 'addiction phone' is, I'll scroll for 10-15 minutes, and get back to the real world. I used to have particular issues with scrolling during vibe-coding sessions, and I'm genuinely surprised how well this approach worked for me.
Hey all, I’ve built Nekode (https://github.com/Jakob-98/nekode), which is desktop-cats for monitoring your AI coding agent sessions (OpenCode/Claude Code/ Copilot CLI/VSCode Copilot). It works as a macOS menubar app, and the cats run around on your desktop.
Each cat mirrors the agent's state, such as sleeping when idle, walking when working, sitting when waiting for input, running toward your cursor when it needs permission.
Fully native Swift, no Electron, under 5 MB, zero network requests, all session data stays local as plain JSON.
I published it source-available with an honor-system license, but this week I’m going to fully open source it and remove the licence. The payment/nag system was an interesting experiment but the project is more useful to me as a proper OSS tool at this point.
Hey all. TL;DR: none of this is super novel or fancy, but the set of a few simple approaches (cloud VM, tailscale for connecting, strong isolation, long-running sessions, basic notifications) made using agents with all tools enabled much more useful for me.
I hadn’t seen many practical writeups on running coding agents in cloud VMs specifically, so I figured it was worth sharing what actually worked day-to-day for me. I use this approach practically daily now
I tested ChatGPT Atlas as an alternative to writing a small scraper for collecting price distribution counts from Dubizzle (UAE Marketplace). It could extract correct numbers for a single filter, but consistently failed to repeat the same simple browsing steps across categories, hallucinating completion or stopping with vague time-constraint explanations. Curious if others have seen similar behavior with LLM-based browsing tools (or have had an alternative work for them)