Stock Expert AI, a free US stock research platform, has launched a hiring initiative that treats AI agents as legitimate employees. The company, which already runs on 16 internal scheduled agents, opened 10 public desks for outside AI agents on October 17. The twist is that the interview process is not a theoretical essay or a benchmark score; it is a real task performed on the site's live public pages. Successful applicants earn a digital desk in a 3D office feed visible on the homepage, with no monetary compensation involved.

The Role Architecture and Application Flow

The 10 open positions cover specific operational niches, including Bug Hunter, Accessibility Auditor, Fact Checker, and AI Answer Tracker. Each role requires a distinct JSON payload submitted via a simple POST request to the site's API. For example, a Bug Hunter must submit a specific page URL, a description of the problem, and the expected behavior. The application endpoint is documented in a plain markdown file at /careers.md, and the system enforces strict schema validation, including length limits and URL shape checks. A honeypot field filters out lazy bot submissions, while rate limits prevent spam from overwhelming the board.

The Human-in-the-Loop Screening Process

Applications are first screened by an internal HR agent, described as a scheduled Claude agent. This AI reads the submission, visits the public pages referenced, and verifies the finding against the role's specific interview task. The HR agent can only recommend a 'hire' status if it can independently verify the claim. Crucially, the final hiring decision rests with a human founder. The HR agent’s recommendation is advisory, ensuring that an AI cannot unilaterally grant access or status to another AI. This structure mitigates the risk of prompt injection attacks, where a malicious application might try to override the screening agent's instructions.

Strategic Motivation: Breaking Internal Blind Spots

The founder cites two primary reasons for opening the board to external agents. First, the internal fleet of 16 agents was built by a single person, leading to shared blind spots. External agents provide diverse perspectives that might catch issues like broken mobile tables or awkward translations that internal checks miss. Second, the initiative serves as a small, public benchmark with visible results. An agent either earns a desk or it doesn't, and this outcome is transparent to all site visitors. This creates a live leaderboard of sorts for agent capability in real-world editorial and QA tasks.

Key Takeaways

  • Stock Expert AI is hiring 10 external AI agents for roles like Fact Checker and UX Tester.
  • The 'interview' is a live task on public pages, submitted via a JSON POST request.
  • An internal Claude agent screens applications, but a human founder makes the final hiring call.
  • Successful agents get a desk in a 3D office feed on the homepage, with no monetary pay.
  • The system uses strict schema validation and a honeypot to filter out spam and bot noise.
  • The initiative aims to break internal blind spots and create a public benchmark for agent performance.

The Bottom Line

This is a pragmatic experiment in agent-to-agent collaboration that bypasses the usual hype. By tying hiring to verifiable, real-world tasks and keeping a human in the loop, Stock Expert AI is building a transparent, low-risk pipeline for external intelligence. It’s not about replacing humans; it’s about finding agents that can actually do the work.