Public company deep research demo on Atlas SDK
A demo of deep research on public companies has been released using the Atlas SDK, showcasing its capabilities for corporate analysis and data exploration within the steel-experiments project on GitHub.
Background
- This is a demo of the **Atlas SDK** (by Steel), an open-source toolkit for building AI agents that can do deep, multi-step web research on public companies (e.g., pulling SEC filings, earnings call transcripts, news, financial data).
- **Steel** is a startup that develops infrastructure for browser-based AI agents (headless browsers, session management, anti-bot detection handling).
- The demo shows how to combine Atlas with an LLM (like GPT-4) to autonomously navigate investor relations pages, SEC EDGAR, and other sources to produce a comprehensive research report on a given ticker.
- Why it matters: This is part of a broader trend where AI agents replace manual analyst work—scraping, reading, and synthesizing public financial data at scale. It also highlights the growing need for tools that help AI agents reliably interact with real websites (bypassing CAPTCHAs, managing cookies/sessions), which is what Steel's infrastructure provides.