A look inside the Harmonic Mind.
Curated views of the Heinrich workspace in active development. Concept substrate, Knowledge State, Hebbian memory, Research Cortex, and proof receipts — the surfaces where a non-LLM architecture actually shows up in the interface. What you see here is real, presented with care for what is ready to show.
The Heinrich workspace, at rest and ready.
Conversation, plan, Knowledge State, evidence, and agent activity — in one coherent surface. Running on CPU. No GPU. No round trip.
Heinrich Workspace — Full View
Recommended: 1920 × 1080, sanitized labels
Heinrich workspace at ready state
The conversation surface, the active goal and plan, the Knowledge State strip, the receipt trail, and the running agent column — all visible at once on the local CPU. The screenshot is the easiest way to understand how the parts of a Harmonic Mind fit together.
The Harmonic Mind, shown in context.
Each view focuses on one surface of the system so the role it plays is unmistakable.
Goal & Plan View
Show: active goal, ordered plan, step status
Goal Orchestration in view
A single goal becomes a structured sequence — done, active, approval-gated, pending — with completion criteria the user can read at a glance.
Knowledge State Panel
Show: Known / Partial / Uncertain / Stale
Answer-readiness, before answering
Every concept tagged with confidence and evidence. The system decides whether it can answer or must research — before generation begins.
Proof Receipts
Show: checkpoints, evidence bundles, test reports
Evidence the user can open
Every meaningful step emits a reviewable artifact. The trail is part of the work — not an afterthought.
Hebbian Memory Map
Show: strongest links, activation history, decay state
Biological co-activation, visible
Concepts that activate together strengthen retrieval paths. Bounded. Decaying. Reviewable. A retrieval prior — never an automatic truth source.
Agent Coordination
Show: running agents, status, current step
Specialized work, organized
Worker agents under a single supervised goal — coordinated, visible, and accountable to the plan.
Approval Gate
Show: paused workflow awaiting user approval
Powerful, but supervised
Sensitive actions pause for explicit user confirmation. The system asks before it commits.
From request to verified outcome.
One arc, four moments — how Heinrich moves a real task from idea to finished work, end to end, on the CPU.
The user states the goal.
Heinrich receives the request, captures intent, and asks for any clarification needed before structuring a plan. The conversation surface stays simple. The Knowledge State underneath begins evaluating what is already known.
Goal Capture
Show: user prompt + intent confirmation
The plan is structured.
Heinrich proposes an ordered sequence with completion criteria. Where research is needed, the Research Cortex schedules it. Where approval is needed, a gate is placed. The user edits, accepts, or redirects before anything runs.
Plan Structure
Show: ordered steps with criteria + gates
Agents coordinate the work.
Worker agents pick up the steps they are suited for, in the order the plan defines. The supervised execution loop tracks state, surfaces blockers, and checks each step against its completion criteria as it moves.
Agent Execution
Show: running step + live progress
Receipts close the loop.
When the work is finished, Heinrich emits a verification report: what was done, what was checked, and what evidence is available. The user can open every receipt and confirm the outcome — not just trust it.
Verification & Receipts
Show: final report + receipt artifacts
Up close.
Focused panels and detail surfaces, captured at component scale.
Knowledge State Detail
portrait, 800 × 1000
Receipt Detail
portrait, 800 × 1000
Agent Card
portrait, 800 × 1000
Seen enough? Get on the list.
Heinrich is opening early access for users with real work to do. The screenshots are a preview — the Harmonic Mind itself is what comes next.