Skip to content
El Observatorio

Building & Agents · 1 min de lectura

When Agents Break Your Flow State

Paul Graham's essay "Maker's Schedule, Manager's Schedule" argues that creative workers need long, uninterrupted blocks to do real work, while managers run efficiently on hourly slots. A single meeting dropped into a maker's afternoon doesn't cost one hour — it costs the whole block. Founders building AI-native companies face an updated version of this problem: the interrupter is no longer just a calendar invite, it's every agent handoff, review request, and evaluation failure that pulls human attention into a broken loop.

Con · estudiado y reformulado para builders AI-native“Maker’s Schedule, Manager’s Schedule” — Paul Graham ↗

The Maker's Block Has a New Shape

Your agents are the makers now. Like human makers, they need complete context, clear scope, and the runway to finish a task before something interrupts the chain. An agent handed a vague goal, a thin tool surface, or a token dump instead of curated context will stall or drift — and then it pings you. The interruption tax Graham described hasn't disappeared; it has moved from your calendar into your architecture. Every poorly scoped workflow is a meeting dropped into the middle of the afternoon.

Founders Still Live on the Maker's Schedule

The judgment calls that only you can make — setting goals, drawing boundaries, deciding what good looks like — require the same deep focus Graham was defending. If your agents are unreliable, you slip into manager mode by default, triaging outputs instead of thinking. The way to protect your maker blocks is to build evaluation harnesses before you build the agents themselves, so failures surface cleanly and rarely, rather than as a steady drip of context-switching across the day.

Schedule Design Is System Design

Graham's real insight was that schedule design is output design — how you arrange time determines what gets made. That logic extends directly to agentic systems. A workflow with tight tool definitions, structured results, and measurable success criteria runs without you. An agent built on vague autonomy becomes a recurring interrupt. Decide, as Graham advised makers to do, where your protected blocks live, and then engineer your agent layer to defend them rather than consume them.

Prueba el Desktop OS

Explora la escena startup de Atlanta, en vivo.

Todo lo que acabas de leer vive dentro de nuestro Desktop OS — un espacio de trabajo interactivo que se abre directamente en tu navegador. Su mesa de Atlanta mapea la escena startup de la ciudad: quién está construyendo, qué está pasando y dónde. Ábrelo y échale un vistazo.

Abre el Desktop OS →