Lens

Most ideas fail not because they were bad, but because nobody challenged them early enough. I built Lens to fix that: a way to pressure-test any idea before spending weeks building it. The core is a LangGraph pipeline where six specialized agents collaborate. A Parser structures the raw input, then three agents run in parallel: the Critic looks for weaknesses, the Researcher maps the competitive landscape, and the Opportunity agent finds angles others might miss. A Feasibility agent evaluates execution realism, and a Synthesis agent pulls everything into a final verdict. The real engineering challenge was not the individual LLM calls, but designing for partial failures, managing parallel execution, and making each agent's output feed cleanly into the next.