YC Summer 2026 RFS Calls for Agent-Native Software and Company Brains
Y Combinator's Summer 2026 Requests for Startups is, by The VC Corner's analysis, the most specific and partner-attributed RFS the accelerator has ever published — 15 categories, each naming the YC partner behind it and each framing a structural gap that AI, robotics, or biology have made newly solvable. Three of those categories name the agent-infrastructure layer directly, and they read like a capital-backed roadmap for the primitives the next AI wave runs on.
What the Source Actually Says
Aaron Epstein's Software for Agents makes the most direct case: while every founder races to build agents, the larger opportunity is the software agents depend on — machine-readable interfaces, APIs, MCPs, and CLIs with thorough documentation and programmatic signup flows requiring no human in the loop. Epstein's logic: the next trillion internet users will be AI agents, and today's software was designed for humans clicking buttons. Tom Blomfield's Company Brain is the organisational counterpart: a system that pulls knowledge from people's heads, Slack threads, email archives, and support tickets; structures it; keeps it current; and exposes it as an executable skills file. Blomfield is explicit — this is not a chatbot over documents, it is a living operational map of how a company actually works.
Gustaf Alströmer's AI-Native Service Companies closes the loop: the era of AI copilots (2023–2025) is ending. The next era skips the human and delivers the service directly. The categories Alströmer names — insurance brokerage, accounting, audit, compliance, healthcare administration — are already-outsourced service markets, structurally easier to replace with an AI-native product than entrenched SaaS. Meanwhile Diana Hu's Inference Chips for Agent Workflows names the hardware gap underneath all of it: current GPUs operate at only 30–40% peak utilization on agent workloads because agents branch, backtrack, and hold long-range context across dozens of steps rather than doing prompt-in/response-out inference. NVIDIA's $20 billion Groq acquisition is cited as the leading confirmation that this investment thesis is already closing.
Strategic Take
"Software for Agents" and "Company Brain" are, in YC's investor-facing language, exactly what MCP orchestration and context engineering are building toward. The RFS signals that capital is flowing to the agent substrate, not just the agents themselves — positioning agenticonsult can reference explicitly in founder and enterprise client conversations.
