
Naver salon owner
Posts model-cut style images weekly but still recruits real hair models.
- Evidence
- Naver Place and Instagram posts show repeated promo-photo production.
- Next contact
- Public comment first
Customer growth agent for AI-built products
BFF analyzes your product, creates the highest-probability customer hypothesis, finds the first 30 people worth talking to, profiles them with public evidence, and turns the conversation back into product work.
curl -fsSL https://breakfastfactory.ai/install | sh
BFF_BIN="$(go env GOPATH)/bin/bff"
# archived demo command
"$BFF_BIN" customers find --count 30
Posts model-cut style images weekly but still recruits real hair models.

Asks for faster before/after content to fill empty booking slots.

Pays for model shoots and runs paid ads for seasonal style menus.

Needs portfolio content before having enough paying customers.

Multiple branches need consistent promo images without extra shoot ops.
The app exists. Paid ads created 30 monthly subscribers and roughly $600 MRR. The founder still needs to know which salons, stylists, and studios are worth contacting next.
paid subscribers from paid ads
paid subscribers after sharper lead loops
First 30 customer leads
Every lead has a public signal, fit reason, profile summary, recommended contact action, confidence score, and spam-risk note. No evidence, no lead.

Posts model-cut style images weekly but still recruits real hair models.

Asks for faster before/after content to fill empty booking slots.

Pays for model shoots and runs paid ads for seasonal style menus.

Needs portfolio content before having enough paying customers.

Multiple branches need consistent promo images without extra shoot ops.
From lead to conversion
Warm conversations and meeting reminders stay attached to the lead.
Reply from the exact post where visual-production pain is visible.
Start with a helpful public comment when the norm makes DMs spammy.
Move qualified accounts into founder-approved sales follow-up.
Customer-driven product loop
BFF turns qualitative customer conversations and quantitative app telemetry into structured insights, then spawns Codex or Claude Code, or exports a clean Markdown handoff for the founder's coding agent.
Salon owners do not understand credit usage before first render.
Fix onboarding copy48% drop on reference-image selection after upload.
Simplify reference flowUsers want style examples by gender, length, and treatment type.
Add preset galleryAnnual plan interest appears after third successful generated image.
Trigger upgrade prompt laterinsights/modelcut-onboarding-loop.mdAttach this to Codex, Claude Code, or the BFF runner to ship the next iteration from customer evidence.
Opportunity, profiling, contacted, active customer, and power user are tracked as one customer memory.
Research progress, tool output, queue status, retries, artifacts, and logs stay visible in real time.
Accepted evidence becomes product learning for the next search, message, meeting, and build loop.
Archived project
The dashboard and backend are offline. This landing page remains available for context, demos, and references.