Black Swan

Company: Black Swan

Self-reported basis: The description is entirely self-reported and unverified, based on a Devpost submission for the OpenAI 2026 hackathon. No external corroboration or additional evidence is available.

What it appears to be: A system that uses adaptive agent swarms to test decisions under uncertainty, with a focus on stress-testing assumptions and evaluating recommendations in complex environments.

What changed: The project was submitted as part of a hackathon; no prior development or commercial activity is evidenced.

Most important open question: What is the actual mechanism by which "adaptive agent swarms" evaluate or challenge evidence, and how does this translate into actionable decision support?

The description states that Black Swan is “chaos engineering for decisions: adaptive agent swarms stress-test assumptions, challenge evidence, and show why a recommendation survives or changes.”

• Claimed function: The system uses "adaptive agent swarms" to evaluate the robustness of decisions under uncertainty.
• Inference: This implies a simulation or modeling framework where multiple agents interact dynamically to test decision outcomes.
• Not evidenced: No details on how the agents are implemented, what kind of assumptions they stress-test, or whether this is a tool for humans or an automated system.

The tagline positions Black Swan as a tool for decision-making under uncertainty, using chaos engineering principles.

• Claim: It applies concepts from chaos engineering (typically used in systems reliability) to decision-making processes.
• Inference: This suggests a novel or cross-domain application of chaos engineering — likely targeting complex environments where traditional decision support tools may fail.
• Not evidenced: No indication of prior positioning, evolution of claims, or how this differs from existing decision-support frameworks.

The description does not specify target customers or ideal customer profiles (ICP).

• Claim: The system is for users who make decisions under uncertainty and need to stress-test those decisions.
• Inference: Likely targets teams or individuals in high-stakes domains like strategy, risk management, or AI/ML decision systems.
• Not evidenced: No evidence of customer personas, use cases, or target industries.

No business model or pricing information is provided.

• Claim: Not stated.
• Inference: If commercialized, it might be a SaaS or consulting offering, but this is speculative.
• Not evidenced: No indication of monetization strategy, pricing tiers, or revenue streams.

The project was built using a range of technologies including FastAPI, LangGraph, OpenAI agents SDK, React, Next.js, Python, Rust, and others.

• Claim: The system is built with modern tooling for AI/ML, web interfaces, and agent-based systems.
• Inference: Suggests a hybrid technical stack combining backend APIs, frontend UIs, and AI agent orchestration.
• Not evidenced: No evidence of architecture, deployment strategy, scalability, or delivery mechanism beyond the tech stack.

The project was submitted to a hackathon (OpenAI 2026).

• Claim: It is a prototype or proof-of-concept.
• Inference: No evidence of traction, users, revenue, or product-market fit.
• Not evidenced: No data on adoption, usage metrics, or post-hackathon development.

No competitive analysis or market positioning is provided.

• Claim: Not stated.
• Inference: The concept of applying chaos engineering to decision-making is novel but not clearly differentiated from existing AI/ML decision support tools or simulation platforms.
• Not evidenced: No evidence of competitors, market size, or competitive advantages.

• Risk: Lack of clarity on how the system works — the term “adaptive agent swarms” is not defined.
• Red flag: The project is a hackathon submission with no evidence of further development or commercial viability.
• Not evidenced: No indication of technical feasibility, scalability, or alignment with real-world decision-making needs.

1. What exactly are the “adaptive agent swarms” in this system? How do they function?
2. What kind of decisions or recommendations does it evaluate, and how is that different from existing tools?
3. Is this a tool for humans to use or an automated system?
4. What assumptions or evidence does it challenge, and how does it show why a recommendation survives or changes?
5. Are there any real-world use cases or early adopters?