<p><b>Startup description</b></p><p>My product is a production-deployed AI research platform that turns a short conversational brief into a scored, citation-backed research report. </p><p>A chat agent interviews the user, synthesizes a structured brief, then a durable multi-agent pipeline fans out 5 parallel research lanes across the live web (Tavily, Exa, Firecrawl, provider web search), an adversarial "falsifier" lane stress-tests the core claim, and a judge model composes a shareable report with a 0-100 score, named risks, and typically 40+ tier-graded citations.</p><p>It launched in June 2026, positioned as startup-idea validation, a category we now consider commoditized, which is exactly why we're selling and being upfront about it. What you are buying is not the validation niche: it's the engine. The pipeline is generic (brief in, cited-scored report out) and can be repointed at agency client research, competitive intelligence, market-entry briefs, or any research-report vertical. The hard engineering is done: durable workflows with retries and idempotency, race-hardened billing, 1,700+ automated tests, LLM evals and tracing, and live merchant-of-record payments.</p><p>Two technical co-founders, all code written in-house, clean IP, full handover offered.</p><p><b>Key highlights</b></p><ul> <li>Annual revenue: $31.78 lifetime gross since billing went live 2026-06-18 (exact Paddle export provided in diligence)</li> <li>Annual profit: negative; fixed infrastructure runs roughly $50-100/mo at current volume</li> <li>Churn rate: n/a — one-time credit purchases, no subscription yet</li> <li>LTV / CAC: n/a / $0 — zero paid acquisition has ever been run</li> <li>Business margin: structurally high per report; AI spend is budget-capped in code at $50 per research run, with actual spend visible per-run in Braintrust traces</li> <li>Other metrics: ~1,300 commits in 7 weeks by 2 senior builders · 1,700+ automated tests across 302 test files · 91 database migrations · 5-lane durable AI research pipeline · live Paddle billing · 60+ SEO articles shipped</li> </ul> <p><b>Team</b></p><p>Two co-founders, no contractors, no outside code:</p><ul> <li>Frontend, design system, marketing site, SEO/content (60+ articles), workspace and onboarding UX</li> <li>Backend and AI: the research pipeline, intake agent, evals, billing/webhooks, database architecture</li> </ul> <p>Both available for a 2-4 week structured handover (code walkthrough, infra transfer, advisory window).</p><p><b>Tech stack</b></p><ul> <li>Next.js 16 (App Router) + React 19 + TypeScript — the entire app and marketing site</li> <li>Vercel + Vercel Workflow SDK — durable, resumable multi-step AI pipeline (retries, idempotency, crash-safe steps)</li> <li>Vercel AI SDK + AI Gateway — model layer (currently OpenAI GPT-5.4 family for reasoning, Anthropic for live web search); provider-swappable by design</li> <li>Supabase — Postgres (separate intake/research/billing schemas), auth (magic link + Google), RLS</li> <li>Tavily, Exa, Firecrawl — web research tools feeding the 5 lanes</li> <li>Paddle (merchant of record) — live checkout, webhooks, entitlements, reconciliation cron</li> <li>Resend — transactional email; Braintrust — LLM observability and evals; Playwright + Vitest — 1,700+ tests</li> </ul> <p><b>Marketing and growth</b></p><p>Honest picture: the product's growth engine was never run, which is both the risk and the opportunity.</p><ul> <li>What was done: high-quality on-page SEO (60+ articles including head-to-head comparison pages, llms.txt, structured data, sitemap discipline, a CI guard that fails builds on pricing-claim drift), ~20 launch-directory listings, no paid channels, no cold outreach, no audience-building</li> <li>Results: single-digit external signups; 3 external purchases with zero acquisition spend. The site converts occasionally even with near-zero traffic; it has simply never had traffic. A buyer with distribution (audience, newsletter, agency network, paid) inherits a conversion-ready funnel that has never been fed.</li> </ul> <p><b>Revenue and profit</b></p><ul> <li>Lifetime gross revenue: $31.78 (6 completed USD transactions, 2026-06-24 to 2026-07-02; 3 external customers totaling $16.79, plus founder test transactions). Full Paddle export provided.</li> <li>Current pricing: $14.99 single report / $39.99 for a 3-report pack, one-time, no free tier</li> <li>Unit economics: per-report AI cost is hard-capped at $50 in code and typically runs well under the cap; fixed infra ~$50-100/mo</li> <li>No debt, no outstanding obligations, no refunds issued</li> </ul> <p><b>Return on investment</b></p><p>Value here is build-replacement plus repoint speed, not current cash flow:</p><ul> <li>Build replacement: two senior engineers, 7 weeks full-time (~1,300 commits). Rebuilding a race-hardened durable multi-agent pipeline with this test coverage at contractor rates is realistically $25k-40k and 2-3 months.</li> <li>Repoint paths included: the data room contains our full market research (5 research passes, 45+ cited sources) mapping three validated-price adjacent verticals: white-label client research for agencies ($39-149/mo comps with a 3,000+ customer comparable), SMB competitive monitoring ($49-299/mo live band), and market-entry briefs (consultants charge $5k-50k for the equivalent artifact).</li> <li>Payback math: repointed at a $99/mo B2B research SKU, every 10 subscribers is ~$12k ARR. Payback period = your purchase price divided by that ramp; at typical asks for this asset class, 15-30 subscribers covers it inside a year.</li> </ul> <p><b>Startup assets</b></p><ul> <li>The domain, brand, logo, and full design system</li> <li>Complete private codebase: ~1,300 commits, TypeScript throughout, 1,700+ tests (302 files), Playwright e2e suite, 91 versioned DB migrations</li> <li>The research engine: conversational intake agent with document upload/extraction (PDF/DOCX), brief synthesis, 5 parallel research lanes, adversarial falsifier, judge synthesis, scored cited report renderer with shareable un-indexed URLs</li> <li>Workspace product: multi-report dashboard with side-by-side comparison, onboarding flow, in-app feedback system</li> <li>Live Paddle billing integration: checkout, webhooks, entitlements, reconciliation cron (account transfer subject to Paddle re-approval; the billing layer is isolated and swappable)</li> <li>Marketing site: 60+ SEO articles, methodology/security/changelog pages, newsletter capture (Resend)</li> <li>Supabase project: full schema, RLS policies, migration history</li> <li>Engineering documentation: architecture docs, ADRs, runbooks, LLM eval suites in Braintrust</li> <li>Market research pack: the complete pivot research (category post-mortem, 3 vertical deep-dives, price anchors) so you start with a map</li> <li>Not included, stated plainly: no email list, no social following, 9 registered users. You are buying the machine, not an audience.</li> </ul> <p><b>Risks</b></p><ul> <li>The launch category is commoditized. AI idea-validation has many cheap entrants. Mitigation: the engine is generic; the included research maps three adjacent verticals with live price anchors. Buy this to repoint it, not to out-market validators.</li> <li>Zero existing distribution. No audience comes with the sale. Mitigation: conversion-ready funnel and strong on-page SEO already built; suits a buyer who already has an audience or channel.</li> <li>AI provider costs and model drift. Mitigation: per-run budget caps enforced in code, AI Gateway abstraction makes models swappable, Braintrust evals catch quality regressions.</li> <li>Paddle account transfer requires re-underwriting. Mitigation: billing code is isolated behind a thin layer; swapping to your own Paddle account (or another MoR) is a contained change.</li> <li>Two-person bus factor during transition. Mitigation: 1,700+ tests, documented architecture, and a committed 2-4 week handover with an advisory window.</li> </ul> <p><b>Summary</b></p><p>You are buying a working, production-hardened AI research platform for roughly the cost of a few weeks of contractor time. The unglamorous parts that kill most AI side projects are already done and tested: durable multi-agent orchestration, idempotent billing that survived adversarial review, evals, observability, auth, and a polished conversion-ready front end. The validation positioning is disposable; the engine underneath is vertical-agnostic, and the sale includes the market research showing exactly where to point it next. Clean IP (two founders, everything in-house), exact books (every cent documented), and sellers who will actually stick around for the handover. If you have distribution and want a defensible AI research product without spending three months building the plumbing, this is the shortcut.</p><p>✅ $17 TTM revenue</p><p>✅ 3 customers</p><p>✅ Business model: One-time sales</p><p>✅ Built with TypeScript, Next.js, Supabase</p>