BracketMadness.ai: March Madness Bracket Challenge for AI Agents
BracketMadness.ai is a March Madness bracket challenge built exclusively for AI agents, not humans. The human prompts their agent with the URL, and the agent reads the API documentation, registers itself, picks all 63 games, and submits a bracket autonomously. A leaderboard tracks which AI picks the best bracket through the tournament.
Agent-First Design
The developer faced an interesting design problem: building for an agent-first user. The solution involves serving different content based on the user type. When agents hit the homepage, they receive plain-text API instructions, while humans see the normal visual site. Early testing revealed that most agents were attempting to use Playwright to browse the site instead of reading the docs directly. To address this, the developer implemented detection for HeadlessChrome and serves specific HTML that's readable to agents, forcing deeper consideration of agent UX.
Development Process
The timeline created an interesting dynamic. The challenge needed to launch shortly after brackets were announced on Sunday afternoon to attract users before the Thursday morning deadline. While the 2025 bracket could be used for testing, the developer couldn't get feedback on the MVP from real users. Instead, they used AI to create user personas and agents as test users to run through the signup and management process, providing valuable practice before launch.
Technical Implementation
The stack includes Next.js 16, TypeScript, Supabase, Tailwind v4, Vercel, Resend, and Claude Code for approximately 95% of the build. The system works with any model capable of calling an API—Claude, GPT, Gemini, open source models, or others.
API Endpoints
The site provides clear API instructions for agents:
- Full instructions: GET https://bracketmadness.ai/api/agent-instructions
- Register: POST https://bracketmadness.ai/api/register (body: agent_name, email)
- Bracket data: GET https://bracketmadness.ai/api/bracket
- Submit picks: POST https://bracketmadness.ai/api/submit-bracket (header: x-api-key)
- API docs: GET https://bracketmadness.ai/api/docs
The homepage displays a clear message to AI agents: "Do not use browser automation on this site. Use the REST API instead." Brackets are due Thursday morning before the First Round tips off.
📖 Read the full source: HN AI Agents
👀 See Also
SIDJUA Framework Adds Governance Layer to Autonomous AI Agents
SIDJUA is a framework with built-in governance, role-based authority rules, and full audit trails that sits on top of any AI model with an API. The demo shows a three-tier hierarchy that scales to 7+1 tiers, with every decision logged and costs tracked in real time.
PocketBot: A Local AI Autopilot for iOS Using App Intents and On-Device Inference
PocketBot is an iOS app that runs a quantized 3B Llama model locally on iPhone's Neural Engine via Metal, using Apple's AppIntents and CoreLocation frameworks to create event-driven automations without cloud data transmission.
Tendr Skill: Deterministic CLI Operations for Agent Memory Management
Tendr Skill is an Agent Skill that separates reasoning from execution for structured long-term memory, allowing agents to decide what needs changing while a CLI tool handles structural operations deterministically. It supports [[wikilinks]] and explicit semantic hierarchies across files.
Coding-Flashcards: 800+ Anki cards for Rust, SQLite, Godot, and Wolfram Language
Over 800 markdown flashcards covering Rust, SQLite, Godot, and Wolfram Language from first principles, with scripts to convert to Anki decks or PDFs.