Pretty wild how under-studied women’s health seems to be! >>> • From 1977 to 1993, FDA guidance recommended excluding women who could become pregnant from Phase I and early Phase II drug trials. • 8 of 10 prescription drugs withdrawn from the US market between 1997 and 2000 posed greater health risks for women than men. • Heart disease is the leading cause of death in women. Yet women are about 50% more likely to be initially misdiagnosed during a heart attack, and those misdiagnosed face about a 70% higher risk of death within 30 days. • In cardiovascular trials, women are still underrepresented relative to disease prevalence in 62% of studies, with no improvement over recent decades. • Women receive less pain medication than men at the same reported pain score. In one controlled study, nurses rated an identical case as less painful when the patient was described as female. • Endometriosis affects roughly 1 in 10 women and takes an average of about 11 years to diagnose in the US.

1/ Some things I've learned recently running coding agents on large-scale projects. Most of this contradicts advice from 6 months ago!

Coding agents are everywhere now, so Notion's primitives are going lower-level and more general. We just shipped a new concept called Workers – it's a hosted compute sandbox with convenient APIs for: - Sync: 2-way sync a Notion database with any external data source - Tools: build secure custom tools for your agents to connect to any external system - Webhooks: expose an endpoint any external service can call More soon!

Models are changing really fast, often in ways that will break your product and technical assumptions. There's no way around this – the only way forward is to relentlessly iterate to make sure you're on target with what's available now, and what you can see on the horizon. The good news is, it's easier than ever to do this if you properly apply coding agents!

Very excited to show you what we're brewing!

Some notes on how I got a coding agent to work for 13 days straight: Self-verification: the agent needs to be able to end-to-end verify everything itself. Design efficient testing layers that make sense for your project, and make sure the agent can effectively loop on them to prove correctness. Write spec documents: work with the agent to fully specify goals, full implementation details, and verification in a document. It's almost always fruitful to iterate on this multiple times. This could be anywhere: a Notion page, or just a simple .md file. Use a running to-do list: break down complex work into a to-do list that you can see and edit. As you think of more things for it to do next, you can just add more to-dos. This could be anywhere: a page or database in Notion, or a .md file. Adversarial review: as a step in the process, the agent should ask another agent to review the spec and implementation and make sure there are no gaps. Force the agent to loop on this until it's totally aligned. This can use sub-agents if that's supported, or you can design a simple CLI that calls another agent. The key thing is to invoke a fresh agent context. This was for a prototype of a new product. It's easier to do all of this on a small codebase, but the principles work anywhere.

The loudest story about AI is a lonely one. One person with an army of chatbots. Other humans are friction. That gets the future wrong. The best things aren’t built alone. In a moment of change, we want to remind the world (and ourselves) what Notion stands for: — Think Together

Sharing a few of my favorite prompt snippets for coding: For iterating on a spec: Please step back and think again. How can we make this SIMPLER and DUMBER while still achieving our goals? For planning: <goal here> Please research the codebase to make sure you fully understand how stuff currently works, and make a detailed plan to achieve the above. Your plan must be fully fleshed out and include only concrete actions such as editing files and running commands (ie. your plan cannot include actions like “read this file”). Your plan must also include how you will verify it is correct yourself, end-to-end. Don’t make any code changes yet. Research now as much as you need and then chat back with your full plan when ready. For debugging: Please think carefully and truth seek. Don't just blindly re-run things and hope they will work. Read the code, and liberally add temporary logging statements anywhere in the codebase to verify logic works as you expect (just remember to not commit them). Dig deep and take the time to really figure out the root causes of problems.

Our cofounder @SimonLast hasn't written a line of code since last summer. With custom agents running while he sleeps, entire products get built without him touching the keyboard. Custom agents run his inbox. Route bugs. File tasks. Search across Slack, Drive, and Notion—and find him answers, not links. He sat down with @Saranormous on @NoPriorsPod to share what it looks like when the entire company runs on agents.

Our cofounder @simonlast built the first Custom Agent at Notion. It had one job: route product feedback. Then he gave it a blank page and let it figure out the rest. Now the whole company relies on it 👀

With Notion Workers, you can build a custom integration that connects your custom agent to pretty much anything, usually in under 10 minutes!

Custom agents are here – I'm so proud of the team for this launch! My personal favorite: I set up an email triage custom agent a few months ago and it completely solved email for me. I get a daily recap of what I need to see, and anything I don't is automatically archived.