Explore Notion’s journey to build 2025 Custom Agents. Learn why Simon Last and Sarah Sachs chose a massive software refactor to create autonomous AI workflows.

The core directive for any developer right now should be: strip away the internal system complexity. If you're constantly fighting the model's limitations instead of riding the current of what it's actually good at, you're going to lose.
A detailed lesson on the 2025 updates to Notion's Autonomous custom agents, specifically focusing on complex architectural changes and advanced workflow capabilities introduced during the 2025 refactor.








The Notion Autonomous Agent Refactor refers to the intensive three-and-a-half-year engineering journey led by Simon Last and Sarah Sachs to redefine how AI interacts with software. Instead of settling for initial versions, the team deleted and rebuilt their product multiple times to move beyond simple chatbots. This process was essential to transition from human-centric interfaces to a system where autonomous agents can effectively perform complex tasks like coding and triaging.
The development of the 2025 Custom Agents was driven by Notion’s AI leadership, specifically co-founder Simon Last and Sarah Sachs. Their vision centered on the realization that traditional software design, which is typically optimized for human eyes, does not work for autonomous agents. By treating agents as a different species of user, they pivoted the architecture toward a 'software factory' model where AI handles heavy lifting and manages other agents.
Notion’s early attempts at building coding agents in 2022 failed because the existing architectural models were not suited for the way machines process information. The team discovered that proprietary JSON models were insufficient for autonomous workflows, leading to a complete refactor. This failure was a turning point that prompted a shift toward more robust tools like SQLite, allowing the system to support agents that can code, triage, and manage workflows autonomously.
Unlike standard chatbots that primarily summarize notes, Notion’s autonomous agents are designed to function within 'software factories' where they perform active work. These agents are capable of coding, triaging tasks, and managing other agents within a workflow. This shift required a fundamental refactor of their software engineering approach, moving away from human-oriented interfaces to an architecture that treats AI agents as the primary users of the system.
From Columbia University alumni built in San Francisco
"Instead of endless scrolling, I just hit play on BeFreed. It saves me so much time."
"I never knew where to start with nonfiction—BeFreed’s book lists turned into podcasts gave me a clear path."
"Perfect balance between learning and entertainment. Finished ‘Thinking, Fast and Slow’ on my commute this week."
"Crazy how much I learned while walking the dog. BeFreed = small habits → big gains."
"Reading used to feel like a chore. Now it’s just part of my lifestyle."
"Feels effortless compared to reading. I’ve finished 6 books this month already."
"BeFreed turned my guilty doomscrolling into something that feels productive and inspiring."
"BeFreed turned my commute into learning time. 20-min podcasts are perfect for finishing books I never had time for."
"BeFreed replaced my podcast queue. Imagine Spotify for books — that’s it. 🙌"
"It is great for me to learn something from the book without reading it."
"The themed book list podcasts help me connect ideas across authors—like a guided audio journey."
"Makes me feel smarter every time before going to work"
From Columbia University alumni built in San Francisco
