@gabztodaro tentei rodar ollama com o modelo gemma4:12b-mlx para tarefas simples de código + https://t.co/FVrRItLnkp. talvez o contexto de skills/AGENTS md que eu tinha deixava o negócio mto sobrecarregado. pra chat normal e estudo de coisas low level deve funcionar bem msm
working more than ever is prettry true to be fair, it’s expected for you to take more threads in parallel because you can delegate to AI. while also having a lot of PRs to review
A advantage that I see from Cursor compared to other AI subscriptions for personal projects is having just the monthly quota instead of 5h/weekly quotas. Of course Claude/Codex have a lot of subsidies but it's annoying to wait until your quota resets.
@thayto_dev Memory files from Claude Code probably has a great impact here, try to summarize the tips from Claude /insights output into codex AGENTS md
I’m building a homelab and having 48 gigs of ram is quite fun, not gonna lie
I literally allocate 24 gigs to colima + k3s so they can have all needed resources.
It never runs out of capacity while I run worktrees with coding agents in the side lol