Published a longrid on agent stack configurations. The dispute usually boils down to a choice between “own OpenCode” and “alien Claude Code or Codex”, but this is a false fork: the binding, model and code are selected separately.

Published. longrid about the configurations of the agent stack. The dispute is usually reduced to a choice between “own OpenCode” and “alien Claude Code or Codex”, but this is a false fork: the binding, model and tools are selected separately, and control is needed over the entire chain from data and identity to actual action in the system. I analyzed eight stack options: their TCO, lock-in, characteristic failures and security boundaries. Sovereignty and security are properties of the entire architecture, not a single component.

Tonight, 17:00 We will discuss these ideas on the stream Misha Trifonov: what to choose for the pilot, corporate platform and adjustable circuit.

#AI4SDLC #Agents #Architecture #PlatformEngineering