Niche-gap negative-space analysis — 2026-05-25

Step 2 deliverable (b) of the Phase-2 research wave. Where is demand obvious but supply missing — and where does OpenAliceLabs's specific stack density let us close that gap before anyone else does. Twelve candidate niches scored on a Hormozi-style value-equation: REACH × MOAT × FIT / EFFORT. Top three identified as primary chase targets to anchor the three vertical pitch decks in steps 3-5.

TL;DR — 12 niches ranked

Recommended primary chase (top 3): Continuous-identity layer (1) + EU-sovereign avatar API (2) + Multi-character drama streams (3). These map cleanly onto the three vertical pitches coming in steps 3-5.

Mermaid view (lower-left = ignore, upper-right = pursue):

quadrantChart
    title Niche map — defensibility × leverage (top 12)
    x-axis "Low leverage" --> "High leverage"
    y-axis "Commoditised" --> "Defensible"
    quadrant-1 "Primary chase"
    quadrant-2 "Defensive niche"
    quadrant-3 "Skip"
    quadrant-4 "Easy money, low moat"
    "Continuous-identity layer": [0.85, 0.95]
    "EU-sovereign avatar API": [0.7, 0.85]
    "Multi-character drama streams": [0.55, 0.95]
    "Open-source robot SDK": [0.55, 0.85]
    "Brand-NPCs in-store SMB": [0.6, 0.55]
    "AI-musicology streamer": [0.5, 0.85]
    "VTuber stack mid-tier": [0.75, 0.45]
    "AI-augmented vibecode streamer": [0.55, 0.8]
    "Brand-mascot revival": [0.3, 0.55]

Methodology — the Hormozi value-equation, adapted

Each niche gets four scores (1-10 each):

• REACH — how large the addressable customer population is for us. 10 = billions of users / millions of orgs; 1 = bespoke single-buyer.
• MOAT — how defensible the niche is given our unique stack (VRM + voice + ZeroEGGS + persona + memory + spatial + A2A + Mollie + EU). 10 = no one else can replicate without our stack; 1 = trivially commoditised.
• FIT — how directly our shipped tech already serves the niche. 10 = we have the product 80 % already; 1 = we'd build something new.
• EFFORT — bootstrap difficulty (lower = harder). 10 = launchable in weeks with our current state; 1 = year + raise + ten hires required.

Composite score = REACH × MOAT × FIT / EFFORT (Hormozi value-eq pattern: big dream × likely to deliver × close to us ÷ work needed).

Lower-bound interpretation: anything below 50 = skip or defer. 50-80 = consider. 80+ = primary chase.

The top 3 — primary chase

🥇 1. Continuous-identity layer for personal AI — our core thesis (score 133)

Demand signal. Every human conversing with ChatGPT (800 M weekly) / Claude (500 M weekly) / Character.ai (200 M MAU) trains a relationship that is vendor-locked to that vendor's app. The moment they switch phone / get a robot / try a VR headset, the relationship resets. Replika 30 M users feel this acutely; the «my AI moved with me» feature is asked-for in every consumer-AI review forum.

Supply gap. Nobody offers portable identity across surfaces. OpenAI explicitly locks personality + memory inside ChatGPT. Anthropic same. Apple Siri is hardware-locked. Character.ai locks to web/app. Replika locks to app. The «portable soul» niche is structurally vacant — it requires either an industry alliance (won't happen, vendors don't share) or a third-party layer with a unique tech stack (us).

Our fit. Charter v2 explicitly stakes this position. Stack is already 80 % there: persona system + memory + VRM + voice + agent-protocol-rs (cross-platform polyglot). Need to add: explicit identity-portability SDK + auth flow + sample connectors.

Effort. Medium — not a brand-new product, more a packaging of what we have + missing connectors + outward-facing SDK + an iconic public demo («Alice moves with you»).

Time window. 2026-2028. After 2028, multi-vendor identity standards emerge and the moat shrinks. Move now.

Vertical it anchors: any of three (streamers / brand-NPCs / digital workers) — continuous identity is the underlying capability, not the go-to-market. Probably anchors the brand layer above all verticals.

🥈 2. EU-sovereign avatar API for regulated industries (score 101)

Demand signal. Post-Schrems II (2020 + reaffirmed 2023), EU regulators have legally complicated US-cloud data-residency for healthcare, banking, insurance, government. Buyers in these segments need avatar interfaces but cannot procure HeyGen / Synthesia / D-ID / Soul Machines because they all run on US cloud. German Mittelstand specifically asks for on-premises or EU-region options that don't exist.

Supply gap. No tier-1 avatar-API competitor is EU-sovereign. Hetzner DE-hosted EU-AI-Act-aligned avatar APIs are a structurally empty market category — not because there's no demand, but because incumbents are US-cloud-locked and cannot pivot without rebuilding.

Our fit. We're already Hetzner DE, already GDPR-native (Mollie billing proves the discipline). Stack is real-time avatar + voice + persona + spatial. The only «product» work is the API surface + an enterprise onboarding flow + audit/compliance docs.

Effort. Medium-high — enterprise sales cycle is 6-12 months; healthcare / banking buyers require security reviews. But the buyers are explicitly looking for us in their procurement requirements; inbound is realistic from the moment we publish.

Time window. 2026-2030. EU AI Act enforces 2027-2028; window stays open through 2030 before US-cloud providers ship EU sovereign options (Microsoft Azure EU sovereign is in flight; AWS will follow).

Vertical it anchors: brand-NPCs vertical (step 4 pitch).

🥉 3. Multi-character drama / storytelling streams (score 90)

Demand signal. Solo VTubers (Neuro-sama, Hololive members, independent VTubers) are 99 % of the AI-streamer market. Multi-character drama / improv / banter streams don't exist because no platform supports two AI agents on one broadcast with coordinated state. Viewers explicitly ask for this in Neuro-sama chat regularly («can you do a podcast with another AI», «what if Neuro had a sister»).

Supply gap. Multi-agent infrastructure (A2A + agent-protocol-rs) is hard. No existing VTuber platform offers it. Vedal built Neuro single- instance; Cover Corp's VTubers are human-operated. The infrastructure gap is real.

Our fit. This is essentially uncopyable for us. A2A polyglot agent-protocol pair shipped; openalice-rt + openalice-realtime is the WebSocket lifecycle stack; multi-agent rooms work end-to-end. Stack is 90 % there for the product «two characters + chat in one stream».

Effort. Medium — scene composition + multi-agent state coordination (turns, interruptions, shared memory) needs polish. But the hardest part (A2A protocol) is done.

Time window. 2026-2029. By 2029 multi-agent VTuber infra becomes table-stakes. We have the «first to ship» window now.

Vertical it anchors: streamers vertical (step 3 pitch).

Honorable mentions — niches 4-6 worth tracking

4. Open-source robot personality SDK (score 69)

Demand signal. Hugging Face LeRobot + Pollen Robotics + K-Scale Labs

• hobbyist robotics community lacks an off-the-shelf «soul» — every

project rolls its own chatbot wrapper. Public Discord channels reflect this gap weekly.

Supply gap. No OSS robot personality SDK exists. Closest is ROS-based behavior trees + custom LLM glue, which are not «soul» in our sense.

Our fit. Persona + voice + memory + animation are all our stack; we'd repackage as an OSS-friendly SDK + reference integrations.

Effort. Medium-high — OSS community building is its own discipline; revenue is indirect (lead-gen for paid OEM deals).

Position: parking lot for 2027. Aligns with Soul-as-a-Service robotics thesis but premature to monetise. Build with one OSS partner (Pollen Robotics or LeRobot) as a credibility lever for paid OEM deals 2027+.

5. Brand-NPCs for in-store / POS retail SMB (score 67)

Demand signal. Saudi giga-projects + UAE hospitality + EU boutique retail all ask «can we put a virtual brand rep on this tablet?». HeyGen-grade pre-rendered video doesn't fit (need real-time interaction); Soul Machines is too expensive ($100k+ per deployment).

Supply gap. Mid-market in-store avatar segment (€500-5000 / mo per location) is empty — pricier than chatbots, cheaper than enterprise Soul Machines.

Our fit. Stack ready; need a vertical-specific product wrapper (touchscreen-friendly UI, payment-integration if needed).

Position: secondary opportunity under brand-NPCs vertical (step 4). Worth a sub-tier inside the brand-NPCs pitch rather than its own pitch.

6. AI-musicology streamer mode (score 81 — higher than I expected)

Demand signal. Music-creator content on Twitch is a known traffic generator (lofi-girl style streams, Suno-AI-react streams, etc.). No streamer-friendly «AI composes a track for chat's prompt, live, on-screen» product exists.

Supply gap. Suno + Udio are batch generation; nothing streams the composition process. Openalice-media has 9 genre presets + MIDI export

• pure-Rust DSP — we are essentially alone in the streamable AI music

niche.

Our fit. 9/10. The product is largely written; needs an OBS scene + streamer integration + monetisation hook.

Effort. Medium — scene work + Twitch / YouTube integration polish.

Position: strong sub-tier inside streamers vertical (step 3). A possible signature feature that makes our streamer offering uncopyable.

Skip or defer (niches 7-12)

• VTuber stack for mid-tier creators (#7) — high reach BUT commoditised; Hololive-tier production beats us on character work. Don't pursue as primary.
• AI-augmented vibecode streamer (#8) — interesting but small market (vibecoders watching other vibecoders); fold into the streamers pitch as a niche segment.
• Personality-driven kiosk worker (#9) — overlaps with brand-NPC POS; not big enough to anchor its own pitch.
• Independent VTuber-as-micro-business (#10) — service business with high per-unit work; not scalable as a product.
• VTuber JP/EU market crossover (#11) — distribution problem, not a technology problem; skip.
• Brand-mascot revival (#12) — IP-licensing nightmare; skip.

Synthesis — how this maps to the 3 vertical pitches

The 3 primary chase niches (1 + 2 + 3) map cleanly onto the 3 vertical pitches in steps 3-5:

flowchart LR
    layer["Continuous-identity layer<br/>(brand-level positioning)"]
    layer --> streamers["Streamers vertical<br/>(step 3 pitch)<br/>= multi-character drama + AI music"]
    layer --> brand["Brand-NPCs vertical<br/>(step 4 pitch)<br/>= EU-sovereign avatar API + POS SMB"]
    layer --> workers["Digital workers vertical<br/>(step 5 pitch)<br/>= embodied flavour vs Polsia head-on"]

Continuous identity is the platform-level positioning above all three; each vertical is a go-to-market wedge that the brand thesis supports.

The decision matrix in step 6 will weigh these three verticals on REACH × MOAT × FIT × EFFORT (now with niche-specific data); NAO picks the 1-2 verticals to lock for Phase 3.

Open questions for future iterations

• Friend / companion B2C — parked as sub-project F. Could re-rank if NAO wants to greenlight it; current SOM (low single-digit M) doesn't justify the marketing burden.
• Education / K-12 / corporate training — entire vertical not considered yet. Would benefit from a 2026-Q4 quarterly refresh.
• Cultural-heritage / dying-language preservation — long-tail recurring + good PR. Could be a non-profit / grant adjunct rather than a primary commercial chase.
• Negative niches (deliberately skipped): adult content (NAO policy line), real-money gambling (regulatory), political / electoral content (reputational risk).

Refresh cadence: re-run this article quarterly OR after each vertical pitch closes with new market data.