Preserve first
Long critiques, investor notes, advisor comments, and user dictation land in data/inbox/ before atomization. Then they become feedback, claims, decisions, research goals, and version trace.
backend of the mind / compiled from data/*.json
This is the control surface for the project. New critiques and information should become atomic feedback, decisions, research goals, or version changes before they become prose in the public artifact.
| ID | Source | Source status | Confidence | Severity | Type | Summary | Implication | Status | Targets | Claims | Disposition |
|---|
| ID | Claim | Category | Source status | Confidence | Status | Sources | Verification needed | Risk if wrong |
|---|
| ID | Scenario | Posture | Price / option | Loan | Equity + closing | Debt service | Tax sensitivity | Other carry | Annual burden before revenue | Monthly break-even | Status |
|---|
| ID | Gate | Status | Owner | Proof required | Kill condition | Source checks |
|---|
| ID | Category | Question | Authority | Status | Needed evidence | Decision effect | Blocks | Target claims | Next action |
|---|
| ID | Date | Decision | Rationale | Status | Affects |
|---|
| ID | Date | Agent | Task | Status | Summary | Mapped to |
|---|
Long critiques, investor notes, advisor comments, and user dictation land in data/inbox/ before atomization. Then they become feedback, claims, decisions, research goals, and version trace.
| ID | Date | Title | Source | Status | Path | Next action |
|---|
Owns repo truth, compiles ledgers into HTML, updates version trace, runs verification, rejects incoherent agent output.
Maintains task state under docs/goals/field-house-project-mind. Scout/Judge/Worker receipts must land back in the data model.
Useful for slop-judge, investor red-team, language surgery, and alternate framing. Outputs receipts, not canonical prose.
Useful for web, exa, firecrawl, commercial-real-estate skepticism, and council passes. Outputs claims and gaps for Codex to vet.