Engineering decisions, architecture deep-dives, and what we're learning about scaling agents.
The origin story — why we chose to build a platform engine that can be a CLI, IDE, personal agent, and orchestration system, instead of another single-purpose AI tool.
EngineeringHow one engine handles the progression from a single coding agent to orchestrating hundreds of agents across multiple environments.
ArchitectureWhy plugins in Codebolt don't need separate agent instances, how the three communication channels work, and what this means for building autonomous companies.
ArchitectureA research-backed comparison of coordination strategies. Why indirect communication through shared state scales better than direct messaging.
ResearchReal problems we encountered scaling agent coordination — conflicts, drift, memory exhaustion — and the system-level solutions we built.
EngineeringInside the deliberation council system — proposing, debating, voting, and reaching consensus without human intervention.
Architecture