State
Engine Room
Records, documents, instrument docs, nudges, diffs, and decisions live in the hub.
Philip OS
Private studio console
The Engine Room is not a CRM and not a chatbot. It is the shared hub where Philip's reads, meetings, decisions, edits, deck work, and weekly reviews become calibrated instruments.
Engine Room hub
State
Records, documents, instrument docs, nudges, diffs, and decisions live in the hub.
Agents
Mildred/OpenClaw and Cowork should read/write through MCP or REST.
Boundary
Personal and Hawco records can cross over now, but provenance stays attached.
Exit
Workspace export/delete proof is an early build requirement.
Vertical slice
Hub health
Loading local hub...
No nudge selected.
syntheticLatest journal entries
Instrument library
AI extracts descriptive fields. Philip writes live wire, main problem, engine, defense, risk, bucket, and next action.
Deck skeletons, design kits, drafts, finals, and edit diffs accumulate in the hub instead of staying stranded.
Relationship-aware notes stay Philip's until the corpus earns trust. AI can organize and polish, not replace judgment.
A delivered memo pulls Philip back into the system: what changed, what is stuck, what needs a decision.
Coverage workbench v1
This keeps the boundary clean: extraction first, judgment only from Philip until the coverage corpus earns the right to draft evaluative work.
Logline, synopsis, format, genre, comps, character list, submitted material facts.
Live wire, main problem, TV engine, primary defense, kill risk, bucket, next action.
Decision why, edit diff, outcome, and similar past cases become the instrument memory.
Writer, rep, buyer, obligation, and history stay visible without turning into CRM clutter.
Instrument 1 detail
AI descriptive
Extracted from material and editable before anything becomes part of the record.
AI descriptive
Neutral summary with citations to pages/scenes once parser support exists.
Philip evaluative
The thing worth protecting. Philip writes it, the system learns from it.
Philip evaluative
Blunt internal decision support, not writer-facing diplomacy.
Philip evaluative
Strong, maybe, unclear, or N/A, with the one-line reason.
Philip evaluative
Pass, read more, develop, clarify rights, meet writer, take to buyer, or hold.
Instrument 2 preview
The deck instrument should migrate Cowork's existing deck work into the hub, then capture draft-to-final deltas so the system learns Philip's pitch language and visual standards.
Import from Cowork
Reusable deck structure, page roles, and narrative sequence.
Per project
Tone, references, typography, palette, image rules, and layout constraints.
The Vig
Known benchmark, treated as asset and taste reference, not a final template.
Calibration
Every draft-to-final change becomes deck-instrument memory.
Nudge flow
Philip answers a message with two sentences. The hub files and links it.
Meeting ends, debrief nudge appears before relationship detail evaporates.
A CRM project, material, person, or deck asset stays attached to the note.
Nudges have status, owner, and result, so both agents know what happened.
Recent residue
JournalEntry
Unstructured note classified after capture; linked to people and projects later.
CoverageRecord
Descriptive extraction stays separate from Philip's evaluative fields.
EditDiff
Before/after correction becomes signal, stronger than a feedback button.
Nudge
A relationship obligation captured without becoming another CRM row.
What changed
Hawco CRM sync
Surface only if a decision, follow-up, or rights risk changed.
Canada-first radar
Market news should become project-specific implications, not a news feed.
Instrument library
No evaluative AI coverage until the instrument has real correction history.