Parallel workers. One shared filesystem.

Agent Harness: Plan, Delegate, Verify

Inside Agent Harness: why Pro keeps serious work in one agent, when Ultra workers help, and how exact-action approvals and verification shape the result.

#plugins #architecture #agents

Fragments become a maintained memory library.

AutoDream: Maintain the Memory

A close look at AutoDream’s reflective memory pipeline: complete-file inputs, validated changes, source lineage, archives, and the limits of recovery.

#plugins #memory #architecture

The graph stores topics. Recall scores words.

Ladybug Memory: Inspect the Recall Path

Look beneath Ladybug Memory’s graph view to its actual recall path: recent candidates, query-word overlap, substring bonuses, and independent storage.

#plugins #memory #architecture

Durable Execution title beside a journal.sqlite panel listing five completed steps with checkmarks and a sixth tool intent marked HOLD

Durable Execution for Agent Zero: Crash, Restart, Continue

A deep look at Alessandro Frau's experimental Durable Execution plugin: how it journals every model and tool call, recovers interrupted chats after a crash, and refuses to repeat actions it cannot prove are safe.

#plugins #architecture #agents

Tree Ring Memory for Agent Zero cover with a sliced tree ring illustration and the Agent Zero mark

Memory That Ages Like Wood: Tree Ring Memory for Agent Zero

Tree Ring Memory is a local-first, Rust-native memory lifecycle layer that keeps fresh memory sharp, compresses old memory into rings, keeps hard lessons visible as scars, and promotes verified truths to heartwood. Here is why it is a strong complement, or a full alternative, to Agent Zero's built-in memory.

#agents #architecture #context

A0 Self-Improvement cover with RLM and GEPA linked in a closed loop

A0 Self-Improvement: Why RLM + GEPA Beats Either Alone

How the A0 Self-Improvement plugin combines recursive long-context analysis with reflective prompt evolution—and why authority, replay, canaries, and a bounded autopilot loop matter more than the optimizer brand.

#agents #architecture #automation

Four isolated capability workspaces connected to a central execution path

Agent Zero v3: A Sneak Peek Inside the Sandbox

Agent Zero's third major iteration is moving agent-run code and heavyweight tools into isolated, disposable gVisor sandboxes.

#agents #architecture #security

Work in Progress: Spynel — One Conversation, Many Agents at Work

Jan Tomášek previews Spynel, a new orchestration layer that turns established AI coding harnesses into one durable, review-driven workflow from desk or phone.

#agents #automation #architecture