Writing from the work.

Technical retrospectives from real work on Agent systems, enterprise AI platforms, personal knowledge, context engineering, and complex product design.

• A Knowledge Base Is Not a Search Box: Building Personal Wiki and Code Wiki in Open Nua

  A personal knowledge base is more than a search box: how Open Nua uses one source, revision, and evidence foundation to build distinct Personal Wiki and Code Wiki experiences.

• Beyond “Summarize the Chat”: How Open Nua Compacts Long-Running Agent Context

  Context compaction is more than summarization: how Open Nua separates conversation facts from model projections while preserving tool, recovery, and provider boundaries.

• Starting by deleting Memory: how Open Nua rebuilt long-term memory

  From removing a heavyweight Memory implementation to rebuilding local long-term memory with an append-only log, rebuildable tree, current facts, and bounded context.

• The Black Screen Wasn't CSS: How Open Nua Stopped a Snapshot Storm in the Conversation Detail View

  How repeated Vibe Trading black screens led Open Nua to diagnose cross-process full-snapshot amplification and close the loop with revision deduplication, frame coalescing, and crash recovery.

• From a Tool Call to Recoverable Execution: How Open Nua Governs Agents

  How Open Nua governs autonomous Agent execution with identity, Capabilities, scoped credentials, HITL, recovery semantics, and minimally sufficient evidence.

• More Than Tools: How Open Nua Turns Plugins into Domain Products

  How Open Nua composes Skills, Capabilities, Apps, state, and governed runtimes into installable, auditable domain products.

• Fix the ruler before ranking the models: turning a Code Wiki failure into an Agent benchmark

  How a failed Code Wiki revision became a reproducible, auditable Agent benchmark that separates model and system failures.