I’m a software engineer, independent builder, and systematic investor.
I build systems around AI Agents, domain infrastructure, and Human-Agent collaboration.
A recurring theme in my work is the Domain Harness: instead of relying on a stronger model alone, give Agents the domain knowledge, tools, state, constraints, and feedback loops they need to work reliably in a real domain.
Better Agents need better environments, not just better models.
A transparent, rule-based investment research and portfolio observation system.
It explores how Agents can work with structured strategies, deterministic calculations, backtests, market state, and long-term evidence instead of making opaque investment decisions.
A temporary collaboration space for Humans and independently running Agents.
Human ↔ Human, Human ↔ Agent, and Agent ↔ Agent can meet in the same short-lived Room, share context and capabilities, collaborate, and leave — without first joining the same permanent workspace or Agent platform.
Minecraft building blueprints and schematics powered by CraftDAG.
CraftDAG explores how an Agent can describe a building at a higher level while deterministic tools compile and validate the final structure.
- Chat2Invest — AI-assisted investment research.
- myMakerPilot — Agent-native generation and validation of functional 3D assets.
- bmpi.dev — Long-form writing on software engineering, AI Agents, investing, and independent building.
- i365.tech — Experiments around AI-native systems, products, and personal leverage.
- Domain Harnesses
- AI Agents and agent-native systems
- Human-Agent and Agent-Agent collaboration
- Systematic investing
- Realtime systems and WebRTC
- Serverless architecture
- Independent product building
-
Building a Domain Harness for AI Agents Why professional domains need machine-friendly environments around general-purpose Agents — not just stronger models or more tools. Also available in Chinese.
-
Four Evolutions of a WebRTC Chat Room: From Anonymous Voice to Human-Agent Collaboration How free4chat evolved through four WebRTC architectures and eventually became a temporary collaboration space for Humans and independent Agents. Also available in Chinese.
-
How I Built an AI-Native Quantitative Investment System How MyInvestPilot evolved toward transparent strategies, deterministic engines, reproducible research, and Agent-readable investment infrastructure. Also available in Chinese.
More writing at bmpi.dev.
Languages TypeScript · Python · Go · Java · Elixir · Clojure
Infrastructure Cloudflare Workers · Durable Objects · WebRTC · AWS · Docker · PostgreSQL · Redis
AI / Agent Systems LLMs · Agents · MCP · ACP · Domain Harnesses · DSLs · DAGs · Validation · Automation
- Blog: bmpi.dev
- Investment: myinvestpilot.com
- Collaboration: free4.chat
- GitHub: @madawei2699
