We're building Forum Atlas at single-FTE + AI-agent scale.

The conventional way to build an industry-intelligence platform is to hire a large analyst team, pick a vertical, sell it at enterprise seat prices, and grow by adding more analysts as the corpus expands. The established products in this category were built that way, and they are good products. Their price follows from how they are produced: the unit cost is an analyst's time.

The agent-built answer is not to displace them at the human-analyst layer. It's to skip the human-analyst layer entirely. Forum Atlas runs as a continuously-ingesting corpus across twelve commercial sectors (defense, cyber, space, materials, tech, health, energy, finance, comms, infra, auto, ag) under a single founder, with a multi-agent ensemble producing per-persona briefs, Monte-Carlo forecasts, and a cross-sector entity graph. The marginal cost per customer-seat is the AI inference + storage + serving cost, not an analyst's time. The pricing reflects that.

Operating commitments

1. Public information only. No paywalled / proprietary data redistributed without license. Sources disclosed by URL on every claim.
2. Citation by default. Every brief, paragraph, analytical claim attributed to source. No anonymous synthesis.
3. Sector-native, not generic. Each sector speaks its industry's vocabulary.
4. Mid-market by design. Five tiers, self-serve from $0. No minimum commitment to start.
5. US-based, US-hosted. Delaware C-corp. US-region cloud.
6. Bootstrap by intent. No outside investors (until revenue justifies). Single FTE + AI agents.
7. Agent-built, not AI-bolted-on. Conversational interface, scheduled synthesis, multi-agent ensembles.
8. Free-data-first ingestion. EDGAR + gov APIs + Wayback + GitHub + arXiv + OpenStreetMap + FOIA foundational.

The company

Forum Atlas is operated by Forum Atlas, Inc., a Delaware C-corp. Justin is the founder + CEO + sole employee. The build is intentional single-FTE for at least the first eighteen months — proving that AI agents can do the work historically done by analyst teams is the core thesis, and a real test requires not having analyst teams.