Y Combinator

How To Stand Out In A World Of AI Designs

The State of Startups in 2026

Hard tech companies, focusing on physical infrastructure like robotics, defense, and semiconductor stacks, have surged from 8% to 20% of YC batches. This shift is driven by AI-powered agents that accelerate research and development, allowing startups to reach significant revenue milestones in months rather than years. ...

Why The Harness Matters More Than The Model | YC Paper Club

Harnesses represent the critical scaffolding layer between large language models and real-world utility, transforming simple sequential processors into agentic systems capable of long-horizon reasoning. By integrating persistent state, tool-calling, and recursive sub-agent orchestration, these frameworks enable models ...

Open Models Are Collapsing The Cost Of AI

The enterprise AI landscape is undergoing a fundamental shift toward open-source models, driven primarily by the need for cost efficiency and greater control over proprietary data. Businesses are increasingly moving away from closed frontier models, with major organizations like AT&T shifting 40% of their token consump...

Paul Graham On Startups, Ambition, and Great Founders

Startup success remains rooted in consistent fundamental principles despite rapid technological advancements like AI. Formidable founders—those who consistently achieve their objectives—drive progress, often motivated more by the immediate fear of failure and operational disaster than by long-term financial gain. While...

Max Junestrand: You Need The Willingness To Learn Faster Than Anyone Else

Legora has evolved into a global legal AI powerhouse by leveraging large language models to automate complex legal workflows, growing from three engineers in Sweden to a company of over 750 employees. The journey, marked by an initial YC rejection and a strategic sales freeze to ensure product quality, highlights the n...

Susan Kare: Designing Icons & Graphics For the Original Mac

Susan Kare, the pioneering iconographer and designer of the original Macintosh, shares a design philosophy centered on simplicity, clarity, and universal accessibility. By working within the strict constraints of 16x16 pixel grids, she developed iconic imagery and typefaces that prioritized function and immediate recog...

Chelsea Finn: This is the State of the Art in Robotics

Achieving long-term autonomy in robotics requires transitioning from bespoke, task-specific models to general-purpose foundation models. High-reliability performance—essential for real-world utility—is best realized through reinforcement learning, where robots autonomously iterate on tasks and incorporate human interve...

Circleback CEO Ali Haghani: Why Your Company Should Be Recording More Meetings

Peter Steinberger: "Fun Is Velocity"

Building OpenClaw, an open-source agentic framework, transformed from a personal solution for task automation into a viral project with over 18,000 contributors. This journey highlights the complexities of scaling open-source software, where rapid growth often clashes with security requirements and the need for a clear...

Why Robotics Still Isn't Solved - But Could Be Soon | YC Paper Club

Robotics development currently faces significant hurdles in bridging the sim-to-real gap, particularly regarding sensorimotor feedback, physical drift, and the scarcity of high-quality data. To address long-horizon task execution, researchers are integrating memory into policies, such as Multi-scale Embodied Memory (ME...

Max Hodak: Average Is Not Good Enough

Operational infrastructure—specifically purchasing, recruiting, and performance management—serves as the operating system for startups, directly dictating the rate of iteration and ultimate success. Deep tech companies often fail not due to technical flaws, but because they lack the systems to manage complexity as they...

How To Design In The Agent Era

Paper is an agent-native design tool that leverages HTML and CSS as its rendering engine to facilitate seamless designer-developer handoffs and enable AI agents to manipulate visual interfaces directly. By treating the codebase as the primary source of truth, the platform allows for rapid iteration and eliminates the n...

Garry Tan: Own Your Intelligence

Personal AGI represents a fundamental shift in how individuals achieve cognitive leverage by transitioning from consuming corporate-owned AI products to building owned, context-rich agentic systems. By treating one's life history—emails, notes, and projects—as a proprietary library, individuals can create "skill files"...

The Case For Data Centers In Space

Starcloud is building space-based data centers to bypass terrestrial energy constraints and regulatory hurdles currently hindering AI infrastructure. By leveraging plummeting launch costs and abundant solar energy in orbit, the company aims to scale compute capacity far beyond Earth-bound limits. Key engineering soluti...

Waymo Co-CEO Dmitri Dolgov: The Demo Is Only 1% Of The Work

Building AI for the physical world requires transitioning from "move fast and break things" to a rigorous "move fast and ship safely" philosophy, as the cost of error involves human lives rather than digital tokens. Unlike digital AI, physical AI demands high reliability from day one, necessitating a "nines" approach w...

Patrick Collison: Is AI Breaking the Lean Startup Playbook?

The integration of AI into modern software development necessitates a nuanced approach to learning, where fundamental cognitive skills remain essential despite the availability of automated tools. While AI can accelerate productivity, deep knowledge—akin to a "cognitive L1 cache"—provides a distinct competitive advanta...

Jeff Dean: The 1% Rule for Building in AI

Jeff Dean, a central architect of Google’s technological infrastructure, examines the evolution of AI from a junior-engineer capability to sophisticated, long-running agent-based systems. Context engineering—providing models with specific tools, memory, and clear guidelines—proves more effective for complex problem-sol...

Lessons From Training Composer At Cursor And Building Meta/Nvidia Compute Clusters | YC Paper Club

AI infrastructure is undergoing a shift toward extreme specialization to meet the massive demand for tokens. Training and inference data centers now require distinct hardware configurations, as training prioritizes compute throughput while inference demands low-latency, batch-size-one efficiency. Innovations in multi-G...

Alexandr Wang: “This is a Once-in-a-Civilization Opportunity”