Notion AI lead argues agents should run on CPUs, not GPUs
The insight is that agent reliability may come from boring determinism — deterministic code paths for repetitive steps — rather than routing everything through an LLM call, which cuts token costs and latency. Builders designing agent workflows should audit which steps can become plain scripts or CPU-served logic, and consider giving agents a local file system for intermediate state. The claim offers a practical counterpoint to GPU-heavy agent architectures gaining traction across the industry.
Notion AI lead Sarah Sachs said at MTS live that many workloads currently served on GPUs should be served on CPUs instead. She argued that executed code offers determinism and observability that LLM-generated processes lack, and that having agents reinvent processes each time is token-intensive and slow. Sachs also highlighted local file systems for storing agent thoughts and writing short Python scripts as among the most important tools Notion's internal agents rely on, besides edit and search.
In Sarah We Trust! MTS @MTSlive Notion AI lead @sarahmsachs argues that a lot of what runs on GPUs today should actually run on CPUs: "We have a strong perspective at Notion that a lot of things should be served on CPU that are served on GPU. There’s a lot of benefit and determinism in executed code." "Agents don’t need LLMs to reinvent that process every time, and that’s super token exhaustive, it’s super slow, and there’s no observability or determinism in it." "Giving the agent a local file system to store local thoughts, to write short Python scripts, that’s really important when there’s so much information. That has been probably one of the most important tools outside of edit and search that our internal agents rely on." @NotionHQ Your browser does not support the video tag. 🔗 View on Twitter 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 0 👀 128 ⚡
