My ActivityPub pet project is getting super close to being kinda ready for public consumption.
Honestly, I’m just being picky about one last feature, being able to quote people, that I think would make it a real contender from the get go, AKA table stakes.
Sure I could have gone as is without, but I’m so close! Once done I think I’ve got a pretty decent MVP IMO (in my opinion).
More on AI
A few weeks ago I switched from Gemini to Claude. And while it’s nice, I really find myself watching the clock. I’ve got the usage meters up and running and watching and refreshing like a hawk.
I usually hit my quota’s hard and twiddle my thumbs do something else while waiting for the next window to open.
But their lower paid subscript still makes me feel like a second class citizen, enticing me to hit upgrade from $20 to $100 / month.
Change Models, Opus v Sonnet
One thing I did learn from last time, is to mix up the models I’m using. Seems the default was set to their Opus model instead of their other Sonnet model. Which has significantly decreased my quota usage.
Someone equated using Opus to driving a Ferrari in rush hour. Sure, you could, but why?
So now I make the plan with Opus, ensure it’s documenting every nook and cranny, then switch to Sonnet once I start getting into the nitty gritty.
Which One is Better?
The fact that I kind of micro-manage my developer agent I don’t know if the flavours and nuances really matter that much. Maybe if I was in total “vibe” mode and didn’t care under the hood the havoc and mess that could be made, then maybe Claude would win, because it oddly does “feel” a bit better.
However using a concise GEMINI.md, a detailed PLAN.md and, a concise TODO.md – combined with using agent skills, I question whether I would even notice.
That along with not feeling I needed to watch the clock, and a second tier citizen – I’m currently likely to go back to Gemini once I use up my monthly allowance.