{"id":218212,"date":"2026-01-03T11:40:08","date_gmt":"2026-01-03T11:40:08","guid":{"rendered":"https:\/\/www.newsbeep.com\/il\/218212\/"},"modified":"2026-01-03T11:40:08","modified_gmt":"2026-01-03T11:40:08","slug":"why-notions-biggest-ai-breakthrough-came-from-simplifying-everything","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/il\/218212\/","title":{"rendered":"Why Notion\u2019s biggest AI breakthrough came from simplifying everything"},"content":{"rendered":"

When initially experimenting with LLMs and agentic AI, software engineers at Notion AI<\/a> applied advanced code generation, complex schemas, and heavy instructioning.\u00a0<\/p>\n

Quickly, though, trial and error taught the team that it could get rid of all of that complicated data modeling<\/a>. Notion\u2019s AI engineering lead Ryan Nystrom and his team pivoted to simple prompts, human-readable representations, minimal abstraction, and familiar markdown formats. The result was dramatically improved model performance.\u00a0<\/p>\n

Applying this re-wired approach, the AI-native company released V3 of its productivity software in September. Its notable feature: Cutomizable AI agents \u2014 which have quickly become Notion\u2019s most successful AI tool to date. Based on usage patterns compared to previous versions, Nystrom calls it a \u201cstep function improvement.\u201d<\/p>\n

\u201cIt’s that feeling of when the product is being pulled out of you rather than you trying to push,\u201d Nystrom explains in a VB Beyond the Pilot podcast<\/a>. \u201cWe knew from that moment, really early on, that we had something. Now it’s, \u2018How could I ever use Notion without this feature?\u2019\u201d<\/p>\n

\u2018Rewiring\u2019 for the AI agent era<\/p>\n

As a traditional software engineer, Nystrom was used to \u201cextremely deterministic\u201d experiences. But a light bulb moment came when a colleague advised him to simply describe his AI prompt as he would to a human, rather than codify rules of how agents should behave in various scenarios. The rationale: LLMs are designed to understand, \u201csee\u201d and reason about content the same way humans can.<\/p>\n

\u201cNow, whenever I’m working with AI, I will reread the prompts and tool descriptions and [ask myself] is this something I could give to a person with no context and they could understand what’s going on?\u201d Nystrom said on the podcast. \u201cIf not, it’s going to do a bad job.\u201d<\/p>\n

Stepping back from \u201cpretty complicated rendering\u201d of data within Notion (such as JSON or XML) Nystrom and his team represented Notion pages as markdown, the popular device-agnostic markup language that defines structure and meaning using plain text without the need for HTML tags or formal editors. This allows the model to interact with, read, search and make changes to text files. <\/p>\n

Ultimately, this required Notion to rewire its systems, with Nystrom\u2019s team focusing largely on the middleware transition layer.\u00a0<\/p>\n

They also identified early on the importance of exercising restraint when it comes to context. It\u2019s tempting to load as much information into a model as possible, but that can slow things down and confuse the model. For Notion, Nystrom described a 100,000 to 150,000 token limit as the \u201csweet spot.\u201d\u00a0<\/p>\n

\u201cThere are cases where you can load tons and tons of content into your context window and the model will struggle,\u201d he said. \u201cThe more you put into the context window, you do see a degradation in performance, latency, and also accuracy.\u201d\u00a0<\/p>\n

A spartan approach is also important in the case of tooling; this can help teams avoid the \u201cslippery slope\u201d of endless features, Nystrom advised. Notion focuses on a \u201ccurated menu\u201d of tools rather than a voluminous Cheesecake Factory-like menu that creates a paradox of choice for users.\u00a0\u00a0<\/p>\n

\u201cWhen people ask for new features, we could just add a tool to the model or the agent,\u201d he said. But, \u201cthe more tools we add, the more decisions the model has to make.\u201d<\/p>\n

The bottom line: Channel the model. Use APIs the way they were meant to be used. Don’t try to be fancy, don’t try to overcomplicate it. Use plain English.<\/p>\n

Listen to the full podcast to hear about:\u00a0<\/p>\n

Why AI is still in the pre-Blackberry, pre-iPhone era;\u00a0<\/p>\n

The importance of “dogfooding” in product development;<\/p>\n

Why you shouldn\u2019t worry about how cost effective your AI feature is in the early stages \u2014 that can be optimized later;\u00a0<\/p>\n

How engineering teams can keep tools minimal in the age of MCP;\u00a0<\/p>\n

Notion\u2019s evolution from wikis to full-blown AI assistants.\u00a0<\/p>\n

Subscribe to Beyond the Pilot on Apple Podcasts<\/a>, Spotify<\/a>, and YouTube<\/a>.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"When initially experimenting with LLMs and agentic AI, software engineers at Notion AI applied advanced code generation, complex…\n","protected":false},"author":2,"featured_media":218213,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[345,343,344,85,46,125],"class_list":{"0":"post-218212","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-artificial-intelligence","10":"tag-artificialintelligence","11":"tag-il","12":"tag-israel","13":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/218212","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/comments?post=218212"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/posts\/218212\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media\/218213"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/media?parent=218212"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/categories?post=218212"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/il\/wp-json\/wp\/v2\/tags?post=218212"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}