Innerverse + LiveGrowEveryday Apps Studio — Master Index

Last updated: 2026-05-07

Purpose: Single recovery point for everything related to the 10-app studio build. Future sessions / agents read this first.

---

TL;DR — what this is

Two parallel app studios, each shipping 5 mobile-first single-file web apps. Built front-to-back using Bezos working-backward + Jobs obsessive-UX philosophy. All apps pre-render hyper-personalized content overnight via Bedrock + OpenAI + ElevenLabs.

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

---

Studio 1 — Innerverse Apps

Brand frame: introspective + ritual. Dark mode default (#0e1014), serif body for emotional content (Iowan Old Style / Source Serif), Inter for UI chrome, warm coral accent (#f78166).

The 5 apps

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

Files (Wave 1 builds — 5 forge teams, ~$25 each)


F:/TITAN/plans/innerverse-spinoffs/
├── dreams/      {press-release, journey, app.html, why-it-hooks, readme}.md
├── oracle/      {same 5 files}
├── timelines/   {same 5 files}
├── childhood/   {same 5 files}
├── arsenal/     {same 5 files}
├── USER-NARRATIVES-2026-05-07.md      ← reverse-engineered customer stories + 7 testimonials
└── FEATURES-CATALOG-2026-05-07.md     ← screen-by-screen feature spec per app

Wave 2 nightly content generation


Script:    F:/TITAN/scripts/innerverse_apps_nightly.py
Output:    F:/TITAN/state/innerverse-apps/{app}/{user}/{YYYY-MM-DD}/
Cron:      titan-scheduled-task "innerverse-apps-nightly" @ 03:10 ET daily
S3:        s3://innerverse-voice-scratch/innerverse-apps/* (public-read prefix)
Hub builder: F:/TITAN/scripts/build_innerverse_hub.py
Cost:      ~$2.50/user/day at premium quality

Infrastructure (already provisioned)


ACM cert:          arn:aws:acm:us-east-1:516005192254:certificate/6108606d-c9e3-45a1-981f-4ed1631fd20b
                   (wildcard *.silentinfinity.com, ISSUED)
Route 53 zone:     Z06831253QG9ZEXH69EC6 (silentinfinity.com)
CloudFront distros (6, all InProgress→Deployed):
  apps:       EJ2Y5FS2NAJXX  → d3cwqpeutyd4o5.cloudfront.net
  dreams:     EG4G64BVSHBF1  → d15i3obpddlohe.cloudfront.net
  oracle:     E1Y1AJVY6MKW72 → d2ciw5fyvdgdst.cloudfront.net
  timelines:  ENCGKCTQEHNQG  → d2i340v133dnl6.cloudfront.net
  childhood:  E2XCVIHF61KT85 → d33fo934grgp1a.cloudfront.net
  arsenal:    E20WKUJGXDMSUM → d1ybennpwb1qnm.cloudfront.net
S3 origin:         innerverse-voice-scratch/innerverse-apps/{slug}/index.html
Provisioning script: F:/TITAN/scripts/setup_innerverse_subdomains.py

Test user identity

Harnoor Mand · 36 · founder of Cloud 8 Data + Innerverse · DC area · Indian-American · 3yo daughter Aanya, infant son · spiritual (Reiki/chakras) · doomscroll/anxiety patterns · recent context: shipped major redesign, 53% Claude credits in 2 days, missed a swim lesson for a deploy, recurring word "stuck."

---

Studio 2 — LiveGrowEveryday Apps

Brand frame: practical + daily + bright. Cream background (#fafaf3), navy ink (#1a2548), signature orange (#f97316), success green (#22c55e). Inter font throughout (no serif — different psychological register from Innerverse).

The 5 apps + concept summary

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

In-flight work (as of 2026-05-07 — 5 forge agents, each ~$25 budget)

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

Each will deliver: press-release.md · journey.md · app.html · why-it-hooks.md · revenue-and-virality.md · readme.md.

Pending after agents complete

1. Read each agent's deliverables; if any failed, dispatch a continuation

2. Register livegroweveryday.com if not already owned (check Route 53)

3. Mirror the Innerverse subdomain provisioning: ACM wildcard cert + 6 CloudFront distros + Route 53 aliases

4. Build a parallel apps.livegroweveryday.com hub showing all 5 in one screen

5. Build a parallel nightly-batch script (livegrow_apps_nightly.py) — same shape as innerverse_apps_nightly.py

6. Schedule nightly cron @ 04:10 ET (1 hr after Innerverse so they don't compete)

---

Shared infrastructure / API stack (BOTH studios)


LLM (text):       Bedrock Claude Sonnet 4.6  (cached system prompts)
Image gen:        OpenAI gpt-image-1         (Bedrock Nova Canvas is LEGACY-blocked)
Video gen:        Bedrock Nova Reel v1.1     (async; Veo 3 on Premium tier later)
Voice (TTS):      ElevenLabs Creator         (Bella default for Childhood + Oracle reveals)
Voice (STT):      AWS Transcribe streaming   (already in Innerverse main)
Music (later):    Bedrock Lyria / Suno      (not wired yet)
LLM (premium):    OpenAI GPT-5 + Claude Opus 4.7 (available, used selectively)
Image (later):    Vertex AI Imagen 3        (GCP service account ready, API enable pending)

.env keys (in `F:/TITAN/server/.env`)


OPENAI_API_KEY                       (sk-proj-..., 116 models incl gpt-image-1)
ELEVENLABS_API_KEY                   (sk_..., Creator tier 131K chars/mo)
ELEVENLABS_VOICE_CHILDHOOD           = hpp4J3VqNfWAUOO0d1Us  (Bella)
ELEVENLABS_VOICE_CHILDHOOD_PITCH_SHIFT = 1.15
GCP_PROJECT_ID                       = claude-code-495616
GOOGLE_APPLICATION_CREDENTIALS       = F:/TITAN/server/gcp-vertex-ai.json
GCP_LOCATION                         = us-central1
PERPLEXITY_API_KEY                   (existing)
RESEND_API_KEY                       (existing — email)
TITAN_API_TOKEN                      (existing)

⚠️ Both ElevenLabs and OpenAI keys were shared in chat unrestricted — rotate to scoped keys when production launches.

---

Key design decisions (immutable)

1. All content pre-rendered overnight — apps fetch static S3 URLs at runtime. No LLM API calls on the hot path. Apps load instantly even on bad networks.

2. Per-user namespacing — s3://{bucket}/{studio}/{app}/{user_id}/{YYYY-MM-DD}/ — every artifact is dated and addressable.

3. Idempotent nightly batch — re-running for the same user/date skips already-generated files.

4. Single-HTML-file apps — each Wave 1 app is one self-contained HTML file with inline samples for offline demo. Wave 2 wires fetch-from-S3.

5. Voice handled at edge, not at runtime — voice notes pre-rendered as MP3, served as static. Web Speech API only for STT (user → app input).

6. Cost discipline — Innerverse premium ~$2.50/user/day. LiveGrow target: similar or less since image-gen volume is lower.

7. Pricing tiers shared across studios — ~$19/mo Foundations, ~$49/mo Premium. LiveGrow apps cheaper individually ($7-14/mo) since they're more single-purpose.

---

Test user "Harnoor" — context all agents share

Use this verbatim across both studios. Forge agents have been briefed with this on every spawn.

> Harnoor Mand · 36 · founder of Cloud 8 Data + Innerverse · DC area, Indian-American · married, 3yo daughter Aanya + infant son · spiritual (Reiki + chakras) · runs multiple ventures · current state: just shipped a major redesign, burned 53% of Claude credits in 2 days then course-corrected with local-first directive · daughter started swim lessons, missed one for a deploy · recurring journal word: stuck · wardrobe leans practical-thoughtful · fridge stocked for South Asian cooking · finances: irregular income (founder), HYSA + 529 set up · workout: 35lb KB + bodyweight + stairs, no gym · sleep ~6.5h, walks 8k/day.

---

How to resume from this index

If a future session asks "where are we on the apps?":

1. Read this file

2. Check F:/TITAN/plans/innerverse-spinoffs/ for completed Innerverse builds

3. Check F:/TITAN/plans/livegrow-apps/ for LiveGrow builds (in progress as of 2026-05-07)

4. Check F:/TITAN/state/innerverse-apps/ for nightly-generated content

5. Check F:/TITAN/logs/innerverse-apps-nightly.log for cron run history

6. Check F:/TITAN/scripts/setup_innerverse_subdomains.py (template) before mirroring infra for LiveGrow

7. Check .env for which API keys are wired

If the LiveGrow agents are still running, their IDs are listed above; their outputs land in F:/TITAN/plans/livegrow-apps/{slug}/. Read each readme.md for a 200-word summary, then triage which to ship first.

---

Open questions / decisions deferred

1. livegroweveryday.com domain ownership — confirm if Harnoor owns this or needs to register. If not owned, alternative: use *.silentinfinity.com paths (silentinfinity.com/grow/assistant, etc.).

2. Real-child voice clone for Childhood — best-quality path requires recording a real 7yo with parental consent + ElevenLabs Instant Voice Clone. Deferred until first cohort feedback proves Bella+pitch-shift insufficient.

3. Vertex AI API enable — claude-code-495616 project needs aiplatform.googleapis.com enabled to unlock Imagen 3 + Veo 3. One click. Deferred until Premium-tier needs it.

4. Veo 3 allowlist — Google approval pending (~24-48h). Premium TIMELINES audio-narrated cinematic loops blocked on this.

5. Real-user cohort — currently only Harnoor as test user. First external cohort decision pending.

---

Single source of truth for the 10-app studio. If anything changes, update this file. Future agents grep here.