I'd like to share an updated methodology for building agents.[P]

Hi guys, been exploring here for a while, wanted to share something we've been working on. It's called Spice, an open-source decision layer above agents. We have tons of great execution agents now — Claude Code, Codex, hermes, etc. They're good at doing stuff. But they're terrible at deciding WHAT to do and WHEN to do it. Right now the "decision" layer is basically you typing a prompt. The agent doesn't know your context, your priorities, your constraints. It just does whatever you tell it. What Spice does: It's a lightweight runtime that acts as a "brain" above your agents. Instead of you deciding what to delegate, Spice observes your context, detects conflicts, simulates options, and dispatches tasks to the right agent. The core loop: perception → state model → simulation → decision → execution → reflection It allows AI systems to: understand context (Decision relevant state) reason about possible futures (simulation) make structured decisions (decision) delegate actions to agents (execution) learn from outcomes (Decision Evolution) Spice does not replace agents like Claude Code, Codex, Hermes, or OpenClaw. It gives them an auditable, traceable, and evolving decision layer before execution. Github: https://github.com/Dyalwayshappy/Spice Feel free to fork, star the repo, or share any feedback and ideas. Would love to build this together with the community.