Options Bot via LangGraph - the dream

In this episode of The Build, Cameron Rohn and Tom Spencer explore an options trading agent prototype built with LangGraph and demonstrate how Langsmith orchestrates model evaluation and traces. They begin by unpacking the AI agent development and architecture, outlining choices between monolithic LLM calls and modular agents, and why pipelines with MCP tools and Langsmith enable debugging and observability. The conversation then shifts to developer tools and workflows, with practical notes on deploying frontend interfaces on Vercel and persisting signals and state in Supabase for realtime experimentation. They explore building in public strategies, detailing trade-offs of releasing MVP telemetry, community engagement, licensing, and monetization—highlighting startup tactics for attracting users and contributors. Next, they analyze technical architecture decisions: vector stores, chain-of-thought instrumentation, test harnesses, and cost controls for inference at scale. Transitions emphasize reproducibility, CI for models, and frameworks for local-first iteration across various development approaches. Throughout, entrepreneurship insights surface about roadmaps, pricing experiments, and partnership plays. They end with a forward-looking charge: they encourage builders to iterate publicly, use modern stacks like Langsmith, Vercel, Supabase, and MCP tools, and treat architecture decisions as experiments toward sustainable AI products.