AI workspace and collaboration software
Notion
Notion builds a collaborative workspace where people write, plan and organize knowledge using documents and databases. Notion AI provides answers with citations, while Custom Agents automate workflows such as triaging product feedback and resolving support tickets in Slack. Notion Calendar extends the product family into scheduling, and its Developer Platform lets developers and agents extend the workspace. The Notion Mail inbox is scheduled to shut down on September 22, 2026; Gmail email history will remain, and Notion’s Gmail connector and agent email tools will continue operating.
Notion began in 2013 with Ivan Zhao and Simon Last; Akshay Kothari joined as COO in 2018 and became a co-founder. Zhao remains CEO. Its enduring engineering foundation is a graph-like block model: text, images, database rows and pages are movable, transformable units. This lets teams assemble customized roadmaps and organize project information, with blocks that can change presentation without losing their stored properties or content.
The company reported that AI-enabled customers accounted for more than 50% of ARR at year-end 2025, with that share more than doubling during the year. In 2026, it announced an approximately $270 million tender offer at an $11 billion valuation, involving GIC, Sequoia Capital and Index Ventures purchasing shares from current and former employees.
4 talks
Newest first
How to build world-class AI products — Sarah Sachs (Notion) and Carlos Esteban (Braintrust)
Sarah Sachs · Carlos Esteban · Doug Guthrie
AI Engineer World's Fair 2025 · 1:43:46
3 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- Notion's Token Town
Start here to understand why favorable per-token prices may not translate into sustainable product costs, and how provider dependence shapes model choices.
Sarah SachsAI Engineer World's Fair 2026
- The Hidden Life of Embeddings
Learn how a T5-based denoising autoencoder can recover text from embedding vectors, with related demonstrations using image models.
Linus LeeAI Engineer Summit 2023
- Understanding is the new bottleneck
See why checking correctness can leave gaps in developer understanding, and how educational practices can help address them.
Geoffrey LittAI Engineer World's Fair 2025
- How to build world-class AI products — Sarah Sachs (Notion) and Carlos Esteban (Braintrust)
In this hands-on workshop, Notion AI lead Sarah Sachs explains how observability, curated datasets, LLM-as-a-judge evaluations, and separate retrieval assessments support reliable AI products.
Sarah Sachs · Carlos Esteban · Doug GuthrieAI Engineer World's Fair 2025
Messages from the stage
Evaluation and provider flexibility
In the joint Notion–Braintrust workshop, Sarah Sachs explains curated datasets, LLM-as-a-judge evaluations, and separate retrieval assessments. In Notion's Token Town, she argues for matching model capabilities to task complexity and retaining provider options as output costs change.
Embeddings as editable material
Linus Lee demonstrates semantic direction editing, vector splicing, and text reconstruction, extending embeddings beyond retrieval. Image interpolation and multimodal vector arithmetic show how these representations can support exploratory interfaces.
Making generated software comprehensible
Geoffrey Litt presents cognitive debt as a problem in agent-assisted development. Collaborative explainers, comprehension checks, and generated microworlds offer ways to make software behavior easier to inspect and learn.
Affiliations reflect each recorded session, not necessarily current employment. The AI product workshop is a joint Notion–Braintrust session, with Braintrust presenters supplying the tooling demonstrations.