Notion just turned your workspace into code

Hey,

I've been working on something new, and this is the first time I've shared it with readers of this newsletter.

I'm working with a couple of rock star developers, and we are going to be launching a series of live, hands-on workshops designed to help people turn AI from something interesting into something genuinely useful. You bring a real part of your work that could be better, and we build a working first version together.

The first workshop I want to introduce is How AI Actually Works and How to Use It in Business. In ninety minutes, you'll learn a practical way to understand modern AI, identify where it could help in your work, and build a starter system around one of your real tasks. It's live Thursday, July 30, 10:00–11:30 AM CT, recording and templates included even if you can't make it live.

Sign-ups are open until today at noon CST. We'd love to have you!

Let's get into it.

- Nick

In Today's Email 👀

• Workers now live in your credits dashboard

• Notion goes full DevOps: workspace as code

• The agent that fixes your wiki instead of just answering

• Notion Computer, a new template, and quick links

3 Tips / Updates

1/ WORKERS NOW SHOW UP IN YOUR CREDITS DASHBOARD

Notion shipped a small but telling update on July 24: Workers usage now shows up in the credits dashboard, right alongside your other AI spend.

Quick refresher on what Workers actually do, since a lot of you are still sleeping on this. They run custom code on Notion's own infrastructure, and they power three things: database sync (pull Zendesk tickets, Salesforce accounts, or any outside data into a Notion database and keep it live), tools for Custom Agents (querying a data warehouse, generating assets, taking action in an app that isn't natively supported), and webhook triggers (an event in another tool kicks off a workflow in Notion). Available on Business and Enterprise, on by default for workspace owners.

The catch: Workers are free during the beta, but that ends. Notion's own pricing page reportedly says credits kick in August 11. The Help Center says October 15. Same company, two different dates, on a feature that runs your automations. Whichever one is real, this dashboard update exists so you can "size your credits before the beta ends" that's Notion's own phrasing, and it's a hint worth taking seriously.

Open your credits dashboard → find Workers usage → note your baseline now, while it's still free. If you've got a Worker running on a tight schedule, this is also the week to check run frequency. Notion's own troubleshooting doc says if cost looks higher than expected, the fix is simple: lower how often it runs.

Run custom code with Workers →

2/ NOTION AS CODE: DEFINE YOUR WORKSPACE IN TYPESCRIPT

Notion opened a beta on July 23 that lets you define an entire workspace, teamspaces, databases, custom agents, all of it, in TypeScript, then deploy it through the API.

Write the setup once and deploy it anywhere: the same script across dev, staging, and production, or a tested setup reproduced in a brand-new workspace without rebuilding from scratch. Check it into git and you get a real change history instead of a wiki page nobody trusts.

This is Notion admitting workspace sprawl is an ops problem now, not a wiki problem. If you're the person who gets pinged every time someone reorganizes a database and breaks three dashboards downstream, this beta is built for you specifically.

Apply for the beta here, then while you wait, list the three teamspaces or databases you rebuild most often by hand. Those are your first candidates for a workspace script once you're in.

3/ THE NOTION AGENT THAT FIXES YOUR WIKI INSTEAD OF JUST ANSWERING QUESTIONS

Most team wikis rot the same way. A question comes in, gets a half answer, and nothing gets written down to fix it for next time.

A setup making the rounds this week treats that as the real problem, not something to bolt a chatbot onto.

One agent, three jobs: answer questions from verified documentation, check live product data when the wiki might be stale, and flag documentation gaps as they surface. Ask it something with no clean answer and it won't guess. It tells you the info is unverified or missing, then opens a task with the original question attached.

The part worth stealing is the wiring. Instructions stay light, and everything routes through separate skills instead of one long prompt, since longer instructions make agents worse at following any single part of them. Four databases carry the weight: questions asked, skills available, the docs themselves, and the tasks the gaps create.

Run it long enough, and the same questions stop generating gaps at all.

Poll / From The Community

There's a feature quietly living inside Notion AI that Notion hasn't really talked about: an alpha called Notion Computer. Ask Notion AI what tools it has access to through it, and the answer isn't a list of prompts. It's a real sandboxed environment, its own machine, that Notion AI can operate directly: writing and running actual code, reading and editing files, installing from a full library of preloaded tools.

What that unlocks in practice is Notion AI actually building things rather than describing them. People have gotten it to generate working Excel sheets with real formulas, styled and printable PDFs pulled from a page, PowerPoint decks, Word docs built from workspace content, even rough first attempts at video. One person used it to export timesheets into the exact PDF format a client's procurement process required, the kind of formatting work that used to mean leaving Notion entirely.

It's still alpha and still rough in places, but it's a different kind of capability than a chat window that summarizes your notes.

Quick Links & Tools

Kimi K3 now live in Notion → Notion quietly added Kimi K3, an open-weight model, to its AI model picker hosted by Fireworks, a US-based provider. One more option if you're chasing model diversity over the usual GPT/Claude/Gemini lineup.

Style your workspace so it actually feels done → A walkthrough on picking one theme, matching cover images, consistent icons, and using headings, callouts, columns, and spacing so pages stop feeling like a wall of text.