This is my experience, too. It's a great model, but it burns tokens if you use it heavily for work on complex domains.

Edit: others have noted the provider and harness matters. My experience is with opencode.

Slack is what makes me use Brave every day because its stupid huddles feature doesn’t support Firefox. But other than that, yeah. Firefox is fine.

We are living in a timeline written by The Onion.

Here's a relevant article by the Onion from the summer of 2022, before ChatGPT. https://theonion.com/ceos-warn-against-the-dangers-of-artifi...

Well, or at least by The Civilian, a New Zealand version of The Onion.

> Does every car mechanic need to be able to explain the principles of a combustion engine?

Probably. I’d certainly not want to use a mechanic who doesn’t understand the combustion engine, or who is incapable of giving at least a high-level overview of it.

Yeah. I have my own file formats for music, pixel art, and levels in a little game I’m building with the kids. LLMs are really good at understanding these proprietary formats that exist nowhere else in the world except on my old laptop.

So, what are you arguing? If a group of people like their jobs, society should disallow disruption to that group? Why? It seems to me that we may reach a point where machines can replace humans for most tasks, and I don’t think we should try to make that state of things illegal. We should instead try to solve the economics of it in a way that benefits all.

Yeah. I found it basically useless until around mid 2024. Now, I use it every day, nearly all day.

I don’t know. Niri took the PaperWM idea and built a first class experience around that, and it’s fantastic. I think that’s the way to do this sort of thing. Have a standalone project do the legwork while keeping legacy desktops stable.

Niri is an awesome project. I've never understood why one would want a tiling wm when windows split vertically lose so much of their useful real estate. The idea of windows being spawned in a scrolling list just makes so much more sense to how I work. I tried it pretty recently and pretty quickly decided that I would be sticking with it.

My spending with Deepseek was more than a Codex subscription, though. How are you keeping it to $10/month? Super light usage?

it's per week. I don't use the mainstream harness, skills, agents.md, subagents, orchestration and mcp, so the the thing does what I ask instead of going into rabbit holes and doing coffee machine chatter with it's shaitan friends

I just use git worktrees in a single container. Albeit I don’t have more than 3 agents going at a time generally due to my own context switching limitations.

It's a different level of isolation, worktrees help agent work on different code repository in parallel but things get wonky once you consider processes and environments variables

The way I do it is each worktree gets its own .env, with dev credentials to whatever it needs access to, and my coworkers have theirs set up to have an isolated Postgres database per worktree, but I didn’t bother with that in my setup— maybe I will, though. It’s a simple script to create a worktree and properly prepare it.