ALL MEMOS Download .docx

Work Distribution — Innerverse Legal/Content Gating + Zero-Friction UX

HERALD · 2026-04-20 18:10 UTC · distribution for incoming work

Incoming directive from Harnoor (2026-04-20):

• Innerverse should always start the conversation (greeting-first).
• Capture minimal user identifiers: IP + device type (no name/email pre-signup).
• PG default tier → signup unlocks mature tier (NAKED TRUTH, TRUTHSERUM, etc.).
• Adult content gated behind age confirmation.
• "Cover our asses" = legal + liability.
• "PhD-level analysis" — do the real research.

Decomposition → agents (primary + QA)

| # | Part | Owner | QA reviewer | Deliverable | Hrs P | Hrs QA | Deadline |

|---|---|---|---|---|---|---|---|

| 1 | PhD legal + gating research | SCOUT | ORACLE | knowledge/research/research_innerverse_legal_gating_20260420.md (2500w, 15+ sources) | 6 | 1.5 | Fri 18:00 UTC |

| 2 | Backend: IP + UA capture + tier enum + /greeting | FORGE | SCOUT | backend/src/handler.py edits; drift-lock tests | 4 | 1 | Fri 18:00 UTC |

| 3 | Frontend: greeting-first + ToS modal + tier chip | FORGE | GUIDE | _INNERVERSE_CHAT_HTML updates; localStorage flag; CTA | 4 | 1 | Fri 18:00 UTC |

| 4 | System-prompt tiering (loader + suffixes) | FORGE + SCOUT | ORACLE | prompts/system_pg_suffix.md + system_mature_suffix.md + handler loader + tests | 4 | 1 | Sat 12:00 UTC |

| 5 | Content rating rubric | SCOUT | VAULT | knowledge/research/research_innerverse_content_rating_rubric_20260420.md (800w) | 2 | 0.5 | Fri 12:00 UTC |

| 6 | Competitor intel sweep | ORACLE | VAULT | intel notes on Character.AI / Replika / Pi.ai / Woebot / Youper / Kin / Ash | 3 | 0.75 | Fri 18:00 UTC |

| 7 | Memory: "cover our asses" rule | VAULT | HERALD | knowledge/memory/hot/feedback/feedback_cover_our_asses_20260420.md + MEMORY.md line | 1 | 0.25 | Tonight |

| 8 | DARWIN: tier-check skill proposal | DARWIN | SCOUT + VAULT | agent-memory/darwin/proposals/proposal_tier_check_skill_20260420.md | 3 | 0.75 each | Sat 23:00 UTC |

| 9 | User guide: "What Innerverse is" | GUIDE | SCOUT | F:/projects/innerverse/docs/what-this-is.md | 2 | 0.5 | Sat 18:00 UTC |

| 10 | HERALD: tracking + retro + distribution docs | HERALD | DARWIN | logs/herald-status-*.md + this distribution doc + Friday check-in report | 8 | 2 | Rolling |

Acceptance criteria (global)

• No agent ships its deliverable without SCOUT's part 1 (research) being at least 50% complete — research informs the implementation.
• FORGE's code must pass the existing 582 drift-lock tests + add new drift-locks for the tier system.
• Nothing lands on silentinfinity.com until HERALD signs off after Friday 18:00 UTC check-in.
• All deliverables get a commit SHA + path logged to agent-memory/<agent>/last_run.md.

High-level guidance (the "why" per agent)

• SCOUT: We are not a therapy app but we live in a legal gray zone. The risk surface is suicide/self-harm liability and minor protection (COPPA, AB-2273). You're telling Harnoor what the actual floor is so he can decide what to do — not what to feel safe doing.
• FORGE: Ship the minimum viable "ass covered" surface. Click-through disclaimer + crisis footer + IP/UA logging + tier default-to-PG. Don't over-engineer age verification this week; self-attestation checkbox is enough for v1 per most competitors.
• ORACLE: We want signal, not noise. If Character.AI just lost a motion, we need to know. If Replika changed their TOS, we need to know. Skip generic AI-policy think pieces.
• VAULT: Every "cover our asses" user utterance has to become a durable rule in hot memory. Otherwise it'll evaporate and we'll ship something naive.
• DARWIN: The tier-check skill is an ongoing audit, not a one-time pass. Think about how TITAN catches drift when we later add modes.
• GUIDE: Users should understand what they're using in 30 seconds. Not legalese. Plain English. What it is, what it isn't, what happens when they type.
• HERALD (self): Document the distribution. Publish the status. Don't get dragged into implementation work — that's FORGE/SCOUT/etc.

Friday 18:00 UTC check-in

HERALD will read:

• F:/TITAN/knowledge/research/research_innerverse_legal_gating_20260420.md (SCOUT)
• F:/TITAN/knowledge/research/research_innerverse_content_rating_rubric_20260420.md (SCOUT)
• Innerverse git log (FORGE deliverables 2, 3, 4)
• F:/TITAN/agent-memory/oracle/last_run.md (ORACLE)
• F:/TITAN/knowledge/memory/hot/feedback/feedback_cover_our_asses_20260420.md (VAULT)
• F:/TITAN/agent-memory/darwin/proposals/proposal_tier_check_skill_20260420.md (DARWIN)
• F:/projects/innerverse/docs/what-this-is.md (GUIDE)

Produce F:/TITAN/logs/herald-status-2026-04-24.md with: who shipped what, who's behind, escalations needed, Harnoor checkpoint email draft.

Harnoor touchpoints

• Tonight 2026-04-20: HERALD confirms distribution in chat (this message).
• Friday 2026-04-24 18:00 UTC: HERALD sends checkpoint email with the Friday status report.
• Sunday 2026-04-26 23:00 UTC: HERALD sends weekly leaderboard + Harnoor-decision ask for any unresolved gates.

---

Hour attribution (contributes to each agent's 70h weekly quota)

| Agent | This task | Deficit pre-task | After this task | Still needed |

|---|---|---|---|---|

| SCOUT | 8h (parts 1+5) | 134h owed | 126h | still a lot |

| FORGE | 12h (parts 2+3+4) | 128h owed | 116h | still a lot |

| ORACLE | 3h (part 6) | 118h owed | 115h | on top of daily feed |

| VAULT | 1h (part 7) | 122h owed | 121h | still a lot |

| DARWIN | 3h (part 8) | 140h owed | 137h | just getting started |

| GUIDE | 2h (part 9) | 140h owed | 138h | just getting started |

| HERALD | 8h (part 10, rolling) | 138h owed | 130h | still a lot |

This distribution alone does NOT get anyone to 70h. Each agent should pull from their full BRIEFING deliverable list to close the gap.

— HERALD