Case study 02
Ìhìn
An AI-native creator distribution platform: businesses launch creator-led campaigns, and invited creators with verified audiences carry them.
Problem
Good products often fail to reach the people who would care about them. Ads read as noise, and creator marketing is hard to trust from both sides: businesses can't easily tell whether a creator's audience is real and relevant, and credible creators lack a fair, structured way to price their work and find the right campaigns.
Product direction
Two connected sides on one platform. Launch is where businesses brief and run creator-led campaigns. Circle is an invitation-only home where creators establish verified audience profiles and publish rate cards. Campaigns follow DAETO (Discovery, Approach, Execution, Tracking, Optimisation), with a person making every approval rather than automated matching.
My role
Founder: product strategy, brand and design direction, and engineering. Ìhìn is a BlindspotLab Limited product; I own the product, the system architecture and the build.
Engineering approach
- One Cloudflare Worker serves React Router SSR with host-aware routing for
ihin.si,circle.ihin.si,launch.ihin.siandpartner.ihin.si, on D1 and private R2. - Creators join by invitation: single-use invitation tokens stored only as hashes, email one-time-code sign-in through Better Auth, and an application state machine in which only the founder can approve or decline, always with a reason.
- Audience evidence stays private. Analytics screenshots are type-checked and kept in private R2; an AI pipeline (through OpenRouter) extracts audience insights and suggests USD rates, with budget and attempt caps and a manual-entry fallback, before a person reviews the result.
- Campaign coordination and approval workflows follow the DAETO method, alongside Dispatches, an editorial publication that will only report consented campaign results.
- Security in the platform itself: per-request CSP nonces, CSRF origin checks on every mutation, staff MFA, and preview hosts that refuse account and write endpoints.
Stack and constraints
- Edge
- Cloudflare WorkersD1R2Cron Triggers
- App
- React Router 8 SSRReact 19TypeScriptDrizzleZod
- Identity
- Better AuthEmail one-time codesGoogle sign-inStaff MFA
- AI
- OpenRouterAudience extractionRate suggestions
- Private by default: creator identity and audience evidence are sensitive, so nothing becomes public without approval.
- No fictional social proof: the site shows no invented creators, campaigns or results.
- One founder-engineer building both sides of a marketplace.
Design decisions
Invitation before scale
A private Circle and invitation-only onboarding keep quality high while the operating model is proven, instead of opening a public waitlist.
AI suggests, people decide
The model reads evidence and proposes rates, but every creator approval and campaign decision is made by a person, and the pipeline falls back to manual entry when extraction fails.
Truthful empty states
Dispatches launched with a real but empty case-study archive rather than borrowed or imagined results.
System
Outcome and status
- Live at ihin.si, with sign-in surfaces at circle.ihin.si and launch.ihin.si (checked 3 November 2026).
- 188 commits between 2026-10-24 and 2026-11-02 in a private repository.
- Identity, creator onboarding, business onboarding, the campaign engine, finance, measurement and release hardening are implemented and covered by automated tests. By the repository's own status report, several external integrations are not yet verified live.
- Early stage: no campaign results are published yet.