News for hackers. Stuff that matters.
Re:OpenClaw users, what is your token spend? by NetworkPerson on Monday February 02, 2026 @07:32PM (#46864545)
I found it far too expensive for Anthropic. Entire context of every conversation is sent each time you type anything. Switched to a local model running from Ollama. Not quite as smart as opus, but good enough for my needs.
Re:OpenClaw users, what is your token spend? by CjHuber on Tuesday February 03, 2026 @02:23AM (#46867675)
Does it not use prompt caching?
Re:OpenClaw users, what is your token spend? by atemerev on Tuesday February 03, 2026 @06:27AM (#46869576)
Had to write my own agent which is better with token economy: https://lethe.gg
Re:OpenClaw users, what is your token spend? by IhateAI on Tuesday February 03, 2026 @04:58PM (#46877909)
I like your website design, but hate llms. Good luck.
Re:OpenClaw users, what is your token spend? by Jeetu456 on Tuesday February 03, 2026 @08:58AM (#46871044)
Not much as i am not leaving for days
Re:OpenClaw users, what is your token spend? by runjake on Tuesday February 03, 2026 @02:23PM (#46875908)
My spend was similar. I only spent one day with it. Pretty neat but a security can of worms and I already do similar automations with Apple Shortcuts and APIs and the Apple Foundation models on-device. It's free and does a great job.

If I had the time and desire, I'd write up a little Swift app and an OpenClaw plugin to just use Apple Foundation models locally, by default. AFM still sucks for code, so proxy those requests out to Claude or Codex or whatever.