Ep 20 - The build - Claude Skills, minimax m2, LangChain Agent Builder, Options trading MCP.

In this episode of The Build, Cameron Rohn and Tom Spencer dissect the latest advances in AI agents and developer tooling for practitioners and founders. They begin by surveying recent model and agent developments, comparing Claude Skills, minimax m2, and LangChain Agent Builder while noting how Langsmith informs observability and evaluation workflows. The conversation then shifts to tooling and deployment patterns, with practical notes on Vercel for edge deployment, Supabase for developer-friendly data layers, and how MCP tools support options trading use cases and risk-aware agent behavior. They explore architecture decisions next, weighing memory systems, vector stores, and the trade-offs between stateful agents and ephemeral chains. The hosts examine developer workflows—local testing, CI, and the role of Langsmith and LangChain in reproducible agent orchestration—and offer guidance on integrating MCP tools for domain-specific trading agents. They also discuss building in public strategies, open source composition, and startup monetization, emphasizing rapid iteration while avoiding hype-driven mistakes. Through temporal transitions and concrete examples, the episode balances technical depth with entrepreneurship insights for teams making architecture choices and product bets. The key takeaway: build transparently, instrument carefully, and prioritize composable tooling to move from prototype to production.