H01Most users don't have an interface to extract AI value — The Wall
Validated
People try AI, can't put their needs into words, get a flat text response, and quit. The problem isn't access — it's that text-in/text-out doesn't fit how the inert 88% live.
How big it is
The demand is consumer — the interface isn't.
How they use it (when they do)
~12 minutes per session, 25–37 minutes per day for daily users (Semrush, SimilarWeb 2025–26). Compare: TikTok ~95 min/day, Instagram ~30–50 min/day. AI is used transactionally, not habitually.
80% of conversations are 'Practical Guidance,' 'Seeking Information,' or 'Writing' (OpenAI/Harvard, 2025). People want outcomes — they get text.
What it costs
How we remove the wall
Yoks — micro-skills with built-in outcomes. No prompting (it's a feed, tap to run). No flat text (outcomes in any medium). No isolation (remixable and shareable).
H02Herd mentality drives mass adoption — and AI products have no herd
Validated
Every consumer category that crossed the chasm did it socially. TikTok, Instagram, Snapchat, WhatsApp — adoption ran on visible peer use, not feature lists. The 88% installs what their friends installed. They don't read changelogs.
ChatGPT has 700M+ weekly users and no friend graph. Claude, Gemini, Perplexity — same. Every conversation is a private 1:1. No feed, no remix, no 'what my friends are running.' That's why penetration plateaus at the agentic 12%: power users find tools through search and docs; the 88% needs the herd.
YokYak is built herd-first
Yoks are public by default. Every successful Yok shows usage counts, remix chains, and creator identity — the same social signals that made TikTok sounds and Instagram filters spread. Permission flows on consensus: '10,000 people trust this Yok with their email' is a different signal than 'this app needs Gmail access.'
This is the moat
A standalone LLM can't simulate consensus. It can be smart; it can't be trusted by association. Trust-at-scale is the only AI moat that doesn't commodify with the next model release.
AI is the first consumer category that hasn't invented the social layer yet.
H03Creators have no AI revenue model — the whole industry is monetizing the wrong side
Validated
The whole industry is monetizing the wrong side.
The doom loop
Platforms build utility tools → the 88% uses casually but won't pay → platforms paywall harder → churn rises → creators get nothing.
That last number flows the wrong direction — creators paying to learn monetization, because no platform offers it.
YokYak inverts this
We don't try to monetize the 88%. We accept they won't pay. Instead we sell brands agency-level integration — pay-per-outcome placement inside the default flow.
- Brands pay. Creators get a share. Users get free outcomes.
- The model only works because we deliver outcomes, not chat.
- Utility platforms can't pivot here — they have no flow for brands to integrate into.
Consumption is free for the 88%. Value is paid by brands. Creators earn from brand budgets — not user wallets.
H04Brand ads have no AI-native infrastructure — and the giants can't build it
Critical
~$700B of digital ad spend is sitting on a substrate that's about to disappear.
The market is about to break
Search ads work because users search. Feed ads work because users scroll. When users delegate to one AI assistant, both behaviors collapse.
The giants face a conflict of utility
If ChatGPT recommends a product, users immediately ask: optimized for me, or for the bidder? The moment a user senses paid placement, the assistant breaks. OpenAI, Anthropic, Google, Apple — all structurally barred from putting ads inside the answer without destroying the product they're selling.
Brands are racing — without infrastructure
Perplexity Shopping, OpenAI's instant checkout, Amazon Rufus, 'generative engine optimization' SEO firms. Every brand is trying to install itself as the AI's first answer.
- None is a marketplace.
- None has auction-based pricing.
- None offers measurable outcome attribution.
Yok is the AI-native brand integration
Brands don't pay for impressions or clicks — they pay to be installed as the default flow inside a Yok. Travel Yok defaults to a brand. Shopping Yok defaults to a brand. Auction-based, outcome-priced, opt-in by user consent.
We move the industry from pay-per-click to pay-per-outcome.
Will brands see fast enough ROI on AI-native placement to redirect substantial budgets, not just pilot dollars?
H05Tech stack is finally ready
Critical
Four primitives shipped or generally available in 2026 that make YokYak buildable by a small team. None existed 12 months ago.
Skills are portable
Claude Skills, GPT actions, and MCP tools standardized AI capabilities as installable units. Thousands now exist; none have a consumer surface.
Apps are agent-callable
Android AppFunctions (Android 16, EAP) and iOS App Intents expose app capabilities to system-level AI agents as first-class tools. The OS plumbing is shipped.
Agents can transact safely
Stripe Link for Agents (April 2026) and x402 + Google's AP2 (Foundation includes Coinbase, Cloudflare, Google, Visa) enable human-approved per-task payments. Atomic permission, exactly what the 88% needs.
Inference is cheap and routable
Together.ai, Fireworks.ai, and others let any model serve any task at near-commodity cost.
All four are production-grade in 2026. YokYak sits at the intersection.
Are there infrastructure gatekeepers we can't get past?
H06The market window is open
Validated
Skills exist. Distribution doesn't. Discovery is the open lane — and nobody is filling it.
Skills exist, distribution doesn't
Thousands of AI capabilities are being published into developer libraries (Claude Skills on GitHub, skills.sh, refero, MCP servers everywhere) with no consumer surface.
AI giants optimize for the agentic 12%
Anthropic, OpenAI, Google build for enterprise, coding, and research. Their consumer products are chat boxes. None has shipped a feed-based discovery layer for the inert 88%, and their UX DNA is text-conversation.
Entertainment giants are format-locked
TikTok, Meta, Netflix own attention but their core loops are video / social / long-form. Pivoting to outcome-feeds requires rebuilding the consumption habit, not adding a feature. TikTok's 2026 forecast doubles down on video commerce, not agentic outcomes.
Adjacent startups solve different problems
- Wabi ($20M pre-seed) — on-demand app generation from prompts. Closest analog. They generate; we curate and distribute pre-validated outcomes.
- Patina — spontaneous software for individuals. No social layer, no creator economy.
- Oasiz ($2.5M seed) — AI-generated games. Mono-format, entertainment-only.
- Playroom ($2M seed) — collaboration tools and social games. Different category.
Four quadrants — discovery layer, consumer-feed UX, outcome-not-content format, agentic-not-passive. They intersect only at YokYak.
H07The cold start: the flywheel can actually start spinning
Critical
Sequenced GTM. Each phase unlocks the next. Brands won't pay without users. Users won't come without creators. Creators won't build without proven format.
Phase 1 — Founder-built proof (months 0–3)
Founder builds 1,000 Yoks personally to prove the format. No creators, no brands, no users at scale. Goal: validate that the Yok format delivers outcomes people actually want.
Phase 2 — Creator seed (months 3–6)
Recruit 50 creators with revenue guarantees. They bring AI/automation expertise. Goal: prove a creator can build a Yok that 10k users install. This is the make-or-break test for the entire flywheel.
Phase 3 — User acquisition via creators (months 6–12)
Creators bring their TikTok audiences into the app. CAC stays low ($0.1–0.5) because distribution is creator-owned, not paid. Goal: 500k DAU.
Phase 4 — Brand integration (months 8–18)
Once user volume justifies it, brands sign default-flow contracts. Auction-based pricing, pay-per-outcome. Goal: 20k brands subscribed.
Founder-built Yoks unblock the chain.
Can founder + 50 seed creators drive 1M organic MAU in 9 months?
H08The giants won't crush the company in 18 months
Validated
We don't need to be unbeatable forever. We need 18–24 months. After that, the moats are structural, not temporal.
Market direction
One assistant per person → skills replace apps → discovery becomes the bottleneck → inert users can't evaluate skills cognitively → they need pre-validated, low-stakes, entertaining exposure to skills → feed format is the only proven mass-market mechanism for that → whoever builds the skill-feed first owns the default install path.
The giants' hands are tied
OpenAI, Meta, Google can't bolt a feed onto ChatGPT or Gemini without breaking the trust relationship that makes the assistant work (the conflict of utility). Pivoting requires a separate product, separate brand, separate UX DNA — a multi-year overhaul we've already done.
Four moats stack by the time they ship
- Validated Yok library — consensus that single LLMs can't simulate.
- Permission graph — users have granted scoped access bit-by-bit; switching means rebuilding trust from zero.
- Creator relationships — top creators locked into our economics.
- Brand integrations — exclusive default-flow contracts. Once a brand is the default in a Yok, replacing them costs the user friction.
The growing threat of competition transforms into the growing price of acquisition.
H09Unit economics work
Critical
Premium to TikTok ($0.75/mo) and Snap ($1.10/mo global). Pay-per-outcome justifies the premium over pay-per-impression.
Per user / month at maturity
Gross revenue $2–6 — premium to TikTok ($0.75/month) and Snap ($1.10/month global), approaching Snap North America ($3/month). Pay-per-outcome justifies the premium over pay-per-impression.
Annualized
Margin compounds with inference deflation. The unit math gets better every quarter we exist.
Will inference costs collapse 10×/year as a16z projects?
H10The founder can execute
In progress
3× $1M+ crowdfundings. 3 exits. Twice a millionaire, twice bankrupt. I know what the chasm feels like from both sides.
Operating record
- 10 years building tech startups — B2B, B2C, 6 different domains.
- eBay and Amazon strategic partnerships — the brand-side relationships experience.
- Hands-on TikTok production at scale — creator-distribution muscles for Phase 3 are already trained.
- 120k consumer devices sold + 3 exits — shipped consumer products that crossed the chasm.
6,000 paid sessions with the inert 88%
U.S. women aged 35–54, college-educated, $100k+ household income — the highest-spending consumer segment in America. They came to me crying about decision fatigue: too many tabs, too many micro-anxieties, no mental capacity left for their own lives.
The pitch that landed every time wasn't 'smarter answers' — it was 'someone else handles it for me.'
The convergence is the unfair advantage
A founder who has sat with the 88%, shipped consumer products to mass market, knows creator distribution firsthand, and holds enterprise commerce relationships. This exact stack is what the problem demands.