# Grant OS Public System Note

Generated at: 2026-03-15T01:31:54.916Z
Radar snapshot: 2026-03-14T17:21:24.298Z

## Scope

Grant OS is a working operating layer for Japanese grant applications. It collects current public-call intelligence, ranks candidate programs, generates draft submission bundles, and prepares portal adapters while separating human-only truth from AI-generated content.

## Totals

- Tracked programs: 11
- Open programs: 3
- Upcoming programs: 2
- Closed programs: 6
- Scenarios: 2
- Routed bundles: 4

## Problem framing

- Grant acquisition is often skewed toward organizations with stronger application machinery.
- Consultant-driven or paperwork-heavy processes can divert time and money away from actual production and research.
- Grant OS exists to move repetitive filing work into AI-assisted tooling so that more attention can stay on the project itself.

## Ownership Lanes

- Human Required: total 48, ready 26, missing 6, placeholder 16, capture_required 0
- Human Seed + AI Refine: total 70, ready 52, missing 0, placeholder 18, capture_required 0
- AI Autodraft: total 54, ready 24, missing 0, placeholder 0, capture_required 30

## Automation workspaces

- Creative / public-call workspace: 2 workspace(s), 14 human issue(s), 2 pending browser mapping(s)
- Research / portal workspace: 2 workspace(s), 8 human issue(s), 2 pending browser mapping(s)

## Adapter Status

- Ready: 0
- Tentative: 0
- Pending capture: 4

## Pipeline

1. Radar ingest collects publicly available call pages, then normalizes deadlines and portal conditions.
2. The shortlist layer ranks which calls should be acted on now using creative-track and research-track strategy profiles.
3. The router turns top-ranked calls into application packets, reviews, checklists, and proposal drafts.
4. The WebUI schema splits submission work into Human Required, Human Seed + AI Refine, and AI Autodraft lanes.
5. The portal adapter planner builds selector seeds and fill plans from public pages or authenticated snapshots before autofill runs.
6. The action runner brings together OpenAI Responses API operations and Playwright-oriented automation for drafting, rewriting, and adapter rebuilds.

## Public / Local Boundary

- The public export explains the social and technical rationale without naming individual target grants.
- The local app shell and browser layer remain responsible for AI-assisted drafting, bundle state, browser control, adapter rebuild, and autofill preview.
- Legal submission, portal authentication, verification handling, and final confirmation remain human-gated.