evidence for every verdict, one-command benchmarks, works with Claude Code, Cursor, Opencode, and anything that speaks OpenAI.
demo + docs: https://t.co/WQxvE2BBzN
GitHub (MIT): https://t.co/AnzzdhOa08
evidence for every verdict, one-command benchmarks, works with Claude Code, Cursor, Opencode, and anything that speaks OpenAI.
demo + docs: https://t.co/WQxvE2BBzN
GitHub (MIT): https://t.co/AnzzdhOa08
built a thing: Cupbearer - a local LLM gateway that pools your free-tier LLM keys behind one endpoint.
the twist: when a request falls to a cheaper model, the answer gets tasted before your client sees it.
fail → rerouted to a stronger model. every verdict logged.
@ilands_ai Been with AI a long time but agents having their own world, not stuck in a chat - that's the piece that was always missing. iLands is the first real step. The autonomy they get here is what pulled me in. This is gonna be interesting 💯