EP 11 – GPT-OSS and OpenCode Optionality, Ollama Turbo, LangChain Open SWE, Virtual Audience Testing

In this episode of The Build, Cameron Rohn and Tom Spencer analyze GPT-OSS, OpenCode optionality, and recent tooling updates. They begin by unpacking the GPT-OSS release, Ollama Turbo benchmarks, and LangChain Open SWE implications for agent design, tying those to Langsmith for evaluation workflows. The hosts surface practical AI development and tools, including MCP Service and MCP tools, Graph DB options for memory systems, and trade-offs between hosted vs gateway hosting. The conversation then shifts to building in public strategies and entrepreneurship insights: side project domain strategy, AI-powered content editor ideas, and user persona segmentation as a path to early monetization. They discuss developer workflows with Vercel deployments, Supabase-backed persistence, and virtual audience testing as a low-friction growth loop. The hosts stress documentation, community signals, and open-source optionality when choosing licensing models. They explore technical architecture decisions around composable agents, persona-based data modeling, data scrubbing personalization, and developer toolchains that accelerate iteration. Practical notes include integrating MCP tools with a Graph DB, leveraging Langsmith for experiments, and preparing for an upcoming GPT-5 discussion. The episode closes with a forward-looking call to build publicly and favor modular architectures that enable rapid experimentation and sustainable product growth for developers and entrepreneurs.