Post

Can the co-author be whatever Stackoverflow post/FOSS project the LLM was trained on? If it's going to geg-in and claim credit that should go to the actual source.

opt-in getting flipped to opt-out specifically for git history is a different category of mistake. devs treat that log as documentation.

the actual issue isn't copilot. it's that git history is sacred to a lot of devs and microsoft just decided to write to it without asking. that's a different kind of violation

I've been looking to drop VS Code for a bit now. It seems like there's a new update every other day for some AI feature I have absolutely no use for, and few other improvements.

I continue to use, and love, Panic’s Nova editor. It’s a native app, and feels like a spiritual successor to TextMate. They don’t seem to have any plans to incorporate AI stuff, which is fine with me. I prefer having Claude Code in a terminal window.

Nova is lovely, but I would like some *thoughtful* AI integration. I get enough productivity boost that I can't ignore AI tools. I don't want Nova to be left behind, and more than that, I think Panic are thoughtful enough to be able to do it well, I don't love any of the AI editors.

I do like Zed a lot, but it doesn't have the same feeling of Nova. I too often use LLMs in a terminal window, but Antigravity's UI for them is kinda nice. Just wish it was native and better designed.

But it’ll help with our metrics to write more lines of code with AI! (My work currently has a “Copilot gains” charge code - so you charge x hours in the week to project 1, y hours to project 2, and “number of hours saved by using AI tools” to “Copilot gains”. They must all sum to 40 hours.)

So… the more time copilot saves, the less time you spend working on any project? Pretty sure I could save 40 hours without burning a single token.

Excellent question! Also entirely unclear as to how this data will be used, what the “right” answer is, and how to meaningfully compare one developer’s self reported metric with another’s.