No Priors: Artificial Intelligence | Technology | Startups

From Coder to Manager: Navigating the Shift to Agentic Engineering with Notion Co-Founder Simon Last

29 min

Notion co-founder Simon Last discusses the company's pivot to AI, driven by the capabilities of GPT-4, and its vision for a platform where humans and agents collaborate. He details the evolution of Notion's AI products, from writing assistants to autonomous custom agents, and how this shift has profoundly impacted engineering practices, leading to a new "agent manager" role for developers. Notion aims to be a "Switzerland for models," providing access to diverse AI capabilities within a collaborative workspace.

Summarized by Podsumo

🎙️ Ask about this episode

✨ Key Takeaways

1

GPT-4's capabilities in 2022 served as the pivotal moment for Notion, immediately sparking both short-term (writing assistant) and long-term (general assistant) AI product visions.
2

Notion rapidly launched AI writer (Feb 2023) and Q&A (Oct 2023), which involved building a robust, real-time semantic index across Notion, Slack, and Google Drive, a complex challenge many native products struggle with.
3

Notion embraces an "agentic engineering" philosophy, rewriting its AI harness every ~6 months to adapt to rapidly evolving models and technology, which is seen as crucial for staying state-of-the-art.
4

The advent of coding agents has dramatically increased individual engineer impact, transforming developers into "agent managers" who design, verify, and monitor end-to-end tasks rather than directly writing code, with Simon Last noting he hasn't written code since last summer.
5

Notion's custom agents enable powerful autonomous workflows, such as email triage and customer feedback routing, learning user preferences and rules on the fly, and are envisioned to bootstrap their own capabilities like building integrations.

💬 Notable Quotes

"I think this is honestly a really key thing and something that a lot of companies get wrong is just like doing one thing and then just like sticking with it you really do have to keenly aware of what the current state of the model and the technology is and then designing the harness and system in the product deeply around that and it basically means you have to rewrite it every six months."
"I haven't written code since like last summer. I don't type code anymore. ... I'm now like the agent manager, instead of the coder."
"I would say two big things. One is that it was just pretty smart. It could follow reasonably complicated instructions. It could write things for you, get edit things. And the second big thing was that the scope of its knowledge was extremely interesting. Super, super deep, like, and broad world knowledge."