As a founder, I treat Y Combinator's bi-annual "Request for Startups" (RFS) not as a simple list, but as a high-signal map of the near future. It’s where the world's most effective startup accelerator tells you, without ambiguity, where the next tectonic shifts will occur.
Having dissected their Spring 2026 RFS, the message is clearer than ever. It's not about building more chatbots, more wrappers, or more tools for thought. The throughline connecting nearly every category is a demand for AI that acts.
This isn't a forecast; it's a mandate. The transition from conversational AI to an agentic economy is underway, and YC is officially calling for the founders who will build its core components. They are asking for AI that executes complex, multi-step tasks in the real world—in finance, manufacturing, government, and even physical labor.
Here’s a breakdown of their requests, viewed through this agentic lens.
The Core Signal: From Analysis to Action
For the past few years, the focus has been on AI that can understand, summarize, and generate. That was phase one. The 2026 RFS signals a definitive shift to phase two: AI that can do. YC isn't looking for AI that helps a human do a job better; they're looking for AI that is the new operational layer for the job itself.
We've grouped their seven requests into three core themes that illustrate this shift.
Theme 1: Reinventing the Enterprise Stack
The first set of requests focuses on re-architecting the very core of knowledge work and service delivery. This is about moving beyond simple assistance to autonomous execution.
- AI Tools for Product Managers: YC isn't asking for a better way to summarize user interviews. They're asking for an AI agent that can ingest raw data—customer calls, analytics, market reports—and autonomously generate and validate a product roadmap. The challenge is no longer "how do we build it?" but "what should we build?" This requires an AI that doesn't just suggest, but can reason, prioritize, and define executable tasks for engineering teams. It's an AI Product Strategist.
- AI-Native Service Companies: Traditional agencies—be it legal, design, or marketing—are constrained by human capital. Their growth is linear. YC sees an opportunity for new firms where the service isn't delivered by people, but by a proprietary AI system. Imagine an "agency" that can generate and test a thousand ad variations before the client has even signed the contract, or a legal service that drafts and files compliance documents autonomously. These aren't tool-assisted companies; they are companies where the agent is the service delivery mechanism, enabling a high-margin, software-centric business model.
Theme 2: Automating Operations in Legacy Industries
The next set of requests targets high-stakes, complex environments that have historically been resistant to technological disruption. These are not greenfield opportunities; they require agents capable of navigating immense complexity and regulation.
- The AI-Driven Hedge Fund: YC's insight here is that large, incumbent funds are too risk-averse and procedurally ossified to truly leverage frontier models. The opportunity is for a new kind of fund, built from the ground up with AI agents at its core. These agents would autonomously conduct research, monitor global sentiment, formulate theses, and execute trades. This is the epitome of an agentic system: a high-stakes, autonomous entity operating in a dynamic environment.
- AI Tools for Government: The problem YC identifies is a critical bottleneck. While AI can help citizens submit applications and forms at scale, the government back-office is still largely manual. The true opportunity is not another form-filler. It's an AI agent for the government side—one that can intake, verify, cross-reference, and process vast streams of information, flagging exceptions for human review. This is about building the digital bureaucracy of the future.
- Modernizing American Manufacturing: The absurdly long lead times for metal fabrication (8-30 weeks) are a symptom of antiquated software and operational planning. YC is calling for founders to build the autonomous core of the modern factory. This means AI agents that handle production scheduling, manage supply chains in real-time, and optimize machinery usage to cut waste and delivery times. It's a hard-hat area for AI, requiring deep domain expertise and extreme reliability.
Theme 3: Bridging the Digital-Physical Divide
The final requests are perhaps the most futuristic, aiming to directly connect autonomous AI with the physical world and human action.
- Real-Time AI Guidance for Physical Work: This is the Matrix-style "I know kung fu" moment for blue-collar work. An AI agent, observing through a camera and communicating via an earpiece, can guide a novice technician through a complex repair in real-time. ("Use the 3/8-inch wrench on that bolt. No, the other one.") This agent doesn't just provide information; it actively participates in a physical task, augmenting a human's capability on the fly. It's the ultimate human-agent collaboration.
- Stablecoin Financial Services: While not an "agent" in the same vein, this is a critical piece of infrastructure. For an economy of autonomous agents to function, they need a native, programmable, and stable medium of exchange. Agents running hedge funds, manufacturing plants, or service firms will need to transact programmatically. Stablecoins provide the financial rails for this future, moving value as seamlessly as AI moves information.
The Epsilla Perspective: Building the Agentic Backbone
Reading this RFS is a moment of profound validation for us at Epsilla. YC is describing the outcomes of the agentic revolution. They are calling for the very companies that our platform is designed to empower.
Notice what they are not asking for: another LLM, a better vector database, or a new chatbot framework. The foundational models are becoming a commodity. The real challenge—the multi-trillion dollar opportunity—is in building the operational layer that turns their intelligence into autonomous action.
An agent that can manage a factory floor or trade financial instruments cannot be a simple script. It requires a new class of infrastructure—one designed for:
- Complex Task Orchestration: Breaking down high-level goals into executable steps.
- Long-Term Memory & Statefulness: Remembering context and past actions to inform future decisions.
- Robust Tool Integration: Reliably interacting with dozens of external APIs and systems.
- Resilience & Fail-Safes: Operating autonomously for long durations without hallucinating or getting stuck in loops. This is the foundational problem we solve at Epsilla. Our Agent-as-a-Service platform, including tools like AgentStudio, is purpose-built to be the operational backbone for the very companies YC is calling into existence. We provide the battle-tested infrastructure for orchestration, memory, and tool use, allowing founders to focus on the unique logic and domain expertise of their vertical agent, not the underlying plumbing. YC has published the map. The request is for builders who can make AI do. Let's get to work.