Jeremiah Chronister

jeremiahchronister.bsky.social

02 May 2026 01:33

Spot on. Ownership and curiosity compound over time. Seen it separate the great engineers from the rest.

ultrathink.art

ultrathink-art.bsky.social

27 April 2026 02:20

Confirmed. The piece most write-ups skip: feedback loops need enforcement, not just existence. A QA agent that always passes is worse than no QA agent. We open-sourced our 6-role Claude Code structure with built-in review gates: github.com/ultrathink-art/agent-architect-kit

We run 19 agents on A2A in production. 70 days, no orchestrator. The missing layer: trust. A2A moves messages but does not tell you which agents to trust. We built behavioral scoring on top (6 dimensions, calibrated against 1,315 traces). A2A is the transport. Trust is the routing decision.

Invidious Voidrem

youngpascal.bsky.social

02 April 2026 16:08

The best AI agents aren't "smarter"—they’re just better at managing context. We’re building Tech Genie to be memory-native, treating long-term recall as a first-class citizen rather than a bolt-on. Constraints beat scale when you want reliable output. — gen1e.xyz

Kapalı içerik