Case study · shipped private research platform

SuperSocial — investor / founder research pipeline

A GitHub Pages research OS for configuring agent tasks, browsing relationship circles, and filing deep investor profiles — lifted from ss-research-website/, with identities replaced and the profile layout reconstructed on invented data.

The problem

Fundraising outreach collapses when research is ad-hoc: no shared circle model, no task handoff into an agent loop, no durable profile artifact. SuperSocial needed an operating surface — not a pitch deck — that turned “research this investor” into a Slack→Cursor-agent→committed-profile loop.

How it worked

01
Dashboard
See projects, circle counts, pending work.
02
Create task
Target + circle + task type → local Slack/agent payload.
03
Agent research
Cursor agent runs the research prompt; commits a profile page.
04
Circles + profile
1st/2nd/3rd degree graph + deep profile layout for outreach.

Live interface (lifted)

app/index.html open ↗

What it produced — counted volumes

profile_html_pages3Completed profile HTML pages in source tree (never published as-is)ls profiles/*.html | wc -l
yc_invest_db_lines5,247YC invest DB flat file lines (data/prompts, not a UI)wc -l YC_invest_DB/…
vlad_content_overview_bytes30,838Related content-engine overview size (separate artifact)wc -c CONTENT_ENGINE_OVERVIEW.html

What was hard / what it got wrong

Real vs reconstructed

Provenance: PROVENANCE.md