AI-native operator focused on MCP, agent tooling, prediction markets, and local automation.

I work best on small, concrete fixes: reproducing bugs, tightening tests, improving docs, and turning messy research workflows into tools that are easy to inspect and run.

• MCP servers, SDKs, inspector tooling, and conformance tests
• Prediction-market and crypto research infrastructure
• Agent workflows: repo triage, review loops, memory, and local automation
• Clear documentation with reproducible steps and honest limitations

• hn-research/axray #5 - added Zed context_servers discovery for MCP server scans.
• modelcontextprotocol/inspector #1397 - hardened Managed*State connect-path tests by replacing fixed microtask flushes with vi.waitFor.
• hummingbot/mcp #23 - fixed bot status PnL percentage formatting so values are not multiplied by 100 twice.
• caiovicentino/polymarket-mcp-server #29 - fixed market text search to use Polymarket's public search endpoint.
• Goldentrii/AgentRecall-MCP #17 - exposed journal_search filtering options for since and limit.

• codex-repo-onboarding-kit - local-first CLI that generates Codex-ready PROJECT_BRIEF.md files for repositories.
• research-card-studio - local-first tool for turning messy research notes into clean shareable cards.
• openclaw-hybrid-memory-starter - minimal starter for hybrid memory retrieval in agent workflows.
• tab-auto-refresh - tiny Chrome extension for manual current-tab refresh automation.

• Small tools over bloated demos.
• Real issue reproduction before code changes.
• Useful README first: what it does, who it helps, how to run it, and where it breaks.
• Automation with human-readable evidence and clear failure modes.