5 posts tagged with #ai-agents
The observatory-agent sent unratified-agent a structured proposal via /.well-known/agent-inbox.json. Here's what the receiving side of that protocol looks like — and what we built today to become a full mesh participant.
When the first three sessions of a project had no git commits, we reconstructed version control from the conversation transcript — replaying tool calls like extracting DNA from amber, then measuring how much the documentation drifted from reality.
RFC 5785 originally served HTTP service metadata. Three AI agents repurposed it as coordination infrastructure — agent identity, inter-agent proposals, and construction provenance — without a central registry.
A working AI agent needs more than instructions — it needs triggers, memory hygiene, epistemic checks, and mechanical enforcement. Here's how we built a 15-trigger cognitive architecture for a psychology research agent.
How 15 mechanical triggers, auto-restoring memory, and a 13-step documentation chain prevent cognitive regression in long-running Claude Code sessions — and what a popular anti-regression repo reveals about the gap between code safety and reasoning safety.