[e925]
🔮 frontier bets

AI companion / character-as-a-service

Niche AI companion w/ persistent memory + voice. Character.ai is saturated; specific niches still wide open.

🔴 hardfirst dollar in months$100 – $1k

🧾 receipts

  • Character.ai + Replika + Glow (the saturated incumbents)Character.ai $1B+ valuation pre-Google deal. Replika $20M+ ARR. Glow + Polybuzz hitting $5-30M ARR. The category exists at scale but the top of the funnel is dominated.
    source ↗
  • Niche AI-companion operators (mostly anonymous, Twitter / Reddit reports)Solo operators in narrow niches (language-learning companions, medical-anxiety conversation companions, foreign-cultural conversation practice) reporting $5-50K MRR. Most of these die — survivor bias warning. Asymmetric: huge if it works, $0 if it doesn't.
    source ↗

🚀 how to start

  1. 01Pick a NARROW niche where Character.ai's general-purpose offering is provably worse. Examples: 'practice Mandarin small-talk', 'rehearse breakup conversations', 'talk through a job-interview script'. NOT 'general AI girlfriend' — already lost.
  2. 02Tech stack: Next.js + Claude / GPT API + ElevenLabs voice + Hume TADA (zero-hallucination TTS, just dropped MIT) + persistent memory via claude-context MCP or vector DB. Voice is non-negotiable for retention.
  3. 03Pricing: $9-29/mo or token-pack $4.99-19.99 packs. Token packs convert better for reluctant-to-subscribe niches; subscriptions win for repeat-use cases.
  4. 04Distribution: niche TikTok + IG (the niche audience is on social, not Google). Demo videos of the companion doing the specific thing the niche needs. Reddit organic in niche subs.
  5. 05Honest risk: this category is the highest-mortality money move in the vault. Cap budget at $2-5K (8-12 weeks) before judging. If it doesn't pull at $1K MRR by month 4, kill it.

🧰 required modules

Modules you'll want under your belt before you fire this move.

💡 recommended modules

Helpful but not required — the polish-tier add-ons.

⚡ tools that power this