#How did you manage to get it to handle
1 messages · Page 1 of 1 (latest)
It took some time. In fact, I need to update this because I was maximizing Antigravity's Opus, and Sonnet as much as possible, plus Gemini to max it out as much as I could.
But now that Google cucked me, I have to redefine it a bit. You could try giving this it and yap to it but to fit your model selections.
Let me see if I can paste it here.
thanks mate feel free to dm even if trouble here
This would be easier to read — it's a start! Kinda' scrappy being so much detail since I've adjusted overtime so you and your agent could maybe refactor to your taste.
legend, that's okay! thank you so much
really appreciate it
how do you come up with plans like this to feed it, do you do that externally in opus on the web for example then feed it into your openclaw later?
any plans what to do now that gemini is rekt? all my accounts banned and whenever I use oauth claude I get through my allowance too quickly
they'll probably ban that too soon, they've been doing that to people lol
need to find a good model to use asap
Of course, happy to help.
Honestly, this evolved over time. I started by having my main agent spawn other agents for specific tasks. At first it would sometimes claim it was using different models, but I realized it was defaulting to Haiku for everything.
So I set a hard rule: the main agent doesn’t do direct execution — it only coordinates. From there, I had it switch to Opus to design a proper multi-tier plan for model usage based on my needs (quotas, provider strengths, task type, etc.).
Over about a week, I kept refining the rules, auditing what it actually used, and tightening policies whenever it broke them. Eventually the orchestration logic stabilized — and now it properly delegates instead of doing everything itself.
GPT Plus is great starters $20 and then $20 Moderato from Kimi -- those are my last options other than my Claude Max ($100/mo version) which is perfect for me now. The Claude $20/mo was so easy to run through.
I need another provider though too.. So I'll keep you posted here for whatever I pick. I just like having a fourth fallback.
that's brilliant man! appreciate it dude I will give that a go and let you know if I find any good providers too haha
you find a solution to a new model yet?
@still oracle I went with z.ai Max plan! GLM-5 has been super impressive and so far lots of usage!
Main stack is now GLM-5 > Kimi 2.5 > Codex > Claude for building
Claude > GLM-5 > Kimi 2.5 for planning
If you end up getting a plan we both get a discount with referral 🤭
🚀 You’ve been invited to join the GLM Coding Plan! Enjoy full support for Claude Code, Cline, and 20+ top coding tools — starting at just $10/month. Subscribe now and grab the limited-time deal! Link: https://z.ai/subscribe?ic=KXIRPHGCET