Is there a future model for open source where feature requests are funded through token donations to implement them?

I could feel Opus 4.8 being prepared to be released yesterday. Same experience I had prior to 4.7 being released. Claude gets slow and impatient. I had a review loop between Claude and codex run 9 turns and Claude still hadn't satisfied codex requests as the reviewer.

There used to be common knowledge around not wanting a car assembled on a Monday or Friday because the workers were recovering from the weekend or rushing towards the weekend.

Sharing, Coldstart, how I am building software with agents. It changes every week. You will probably not like it. There is no one-shot. There is not solution. You have to read and think a lot. More of a pattern than anything else.

Lazy software pattern I'm enjoying. Built maildb to give Claude SQL & semantic search over my email and attachments via MCP. Thought I'd eventually package it up nicely. Instead I run postgres on an old laptop in the garage, with tmux claude code cli running in the git repo. Use from my phone via /remote-control. Bug, perf issue, or want new feature? I prompt Claude then and there to build it. Claude is the sysadmin: runs the ingestion pipeline (extraction embeddings) and DB backups. No releases, no packaging. Just features that exist, features to build, and an AI sysadmin. I am not alone @rhettford told me he has a similar situation with a project of his.