A Collective of Curiosity: Community Building in the Age of AI, Part II
Field notes on how we can create belonging in a world where technology is reshaping what it means to be human.
Part I (which you can read here) covered a lot on how communities have shifted from broadcasting to belonging. Some practitioners who run AI communities reached out with a validation of the same while a few mused at the obvious: What do we actually do with this?
Part II is my attempt to answer that.
These are field notes on what seems to be working in AI communities right now—not prescriptive advice, but observations from running community for an AI fund, building events, and learning from people like Annie Warner at Lenny's Newsletter.
So, here’s what seems to be working!
1. Communities Are About People and Their Curiosity
We need to realize that communities are about people and their curiosity to get information. We all like to be in the inner circle. We all like to know.
The best way to do that is to ensure that people are able to feed into that curiosity with intent but also genuinity.
What does that look like for an AI startup? What’s the next build going to look like? What’s the next problem you’re solving? Whether it’s a feature, a platform extension, maybe a pivot.
Getting some hint into that helps people reimagine—curiosity feeds into imagination, and you’re essentially doing more than just telling them about a product.
Test it: Share one behind-the-scenes update this week (a feature in progress, a problem you’re stuck on, a decision you’re wrestling with) and ask your community: “What would you do?”
2. Build Around What Feels Natural
Build around what feels natural and mirror your personality—or give it shape and let it roll.
It’s really important to recognize this. It’s all right if you are not the most extroverted extrovert out there who can run community, but you have to also recognize that building around what feels natural to you—if you don’t have resources—or giving it to somebody who can emulate and understand your end goal and letting them roll with it, matters.
You are the conductor, not the orchestra.
Test it: Identify one community activity that drains you. Either delegate it or kill it. Double down on what feels natural to your energy.
3. Have One Foot In, One Foot Out (When It Comes to Creative Control)
Have one foot in the door, one foot out of the door when it comes to creative control. I know it is really hard to let go, but oftentimes having things take shape and form is the best way to let community build.
Annie told me something interesting during the conversation we had: they don’t have tons of rules in their Slack, but they’re ruthlessly stringent about self-promotion.
Most channels are completely free from it, which means folks don’t feel like they’re just in a channel where people drop their own LinkedIn stuff all the time.
That single constraint creates psychological safety. People show up to genuinely help, not to pitch.
For AI communities, this is crucial. Your early members are likely building their own startups or side projects. If every conversation becomes a disguised sales pitch, trust evaporates.
Test it: Pick ONE behavior you want to eliminate from your community. Enforce it consistently for 30 days and watch if the quality of conversation changes.
4. Build Experiences and Put Effort In
People want to see you work hard. It’s literally like a romantic relationship.
They want:
In-person experiences
Online experiences
Something unique, something exclusive, something curated—the whole spectrum
And the truth is you have to identify what that balance looks like for you as a founder and as a business.
The best communities figure out their channels. Email brings people back. Slack or Discord facilitates real-time help and async discussion. In-person events create high-touch, curated experiences. Each serves a different function.
The communities that feel most alive aren’t the ones with the most channels or members. They’re the ones where every interaction feels intentional. Quality over quantity—turn down partnerships and growth opportunities if they don’t hit your bar.
Small gestures compound too. Physical mugs for custom emojis. Handwritten notes. Small tokens that cost $15 but signal: We see you. We appreciate the tiny things you do to make this place feel like ours.
Test it: Pick one small behavior you want to encourage (intro posts, helping others, creative contributions). Reward it with something tiny but personal and see if that behavior increases.
5. Don’t Control the Narrative—Shape-Shift Around a Theme
The reason I say this is because most AI companies and most businesses believe that if you’re investing in something, you need to control the narrative. The truth is you cannot keep controlling all narratives—because then there is no voice. It’s just an agenda, like a corporate deck.
So you have to build around a theme.
If you’re an AI company that is, say, building the infra layer for cybersecurity and you’re running evals, you need to ensure that your community theme isn’t just about your product.
Maybe it’s linked to a broader problem that people are often struggling with or the biggest problem that you want to solve in your line of business. Or it could be something as technical as a research paper-driven discussion, or even a mental model for thinking about the space.
For instance, Anthropic’s community isn’t just “here’s Claude”. It’s “here’s how we think about AI safety and alignment.” The theme is bigger than the product, which gives people room to engage beyond just feature requests.
Set the initial conditions, then let emergence happen. The best communities surprise you. If yours doesn’t, you’re holding too tight.
Test it: Identify the bigger theme your product sits within. Start one conversation around the theme, not your product, and see if it unlocks different kinds of participation.
6. Run Tiny Experiments (And Run Many of Them)
And the last bit is run tiny experiments. I am a big, big believer of Anne-Laure Le Cunff ‘s work. And ever since I started reading Tiny Experiments, I have been running them in literally all aspects of my life (Definitely counterintuitive to what the book suggests - but I’m a little excitable about these things. Whoop.)
What I’ve learned from the book is that it allows you to not overthink. It allows you to test something and then decide if it’s working for you or not. You don’t have to commit. You don’t have to get married. People are going to move on if something doesn’t work.
And the beauty of building in AI today is that the world is really forgiving. We are all trying very hard. And we’re all trying to make it. And we understand that things are so new—you cannot know what’s best unless you do it repeatedly.
A good filter: Will this help us grow? And can we execute it at our quality bar? If the answer to both isn’t yes, don’t do it.
When I look at startups today, we’re in the best place to experiment. So try something. If it works, great. Scale it. If it doesn’t work, move on to the next thing.
Test it: Pick one thing you’ve been “thinking about doing” for your community. Run it as a one-time experiment this week and decide after: scale or kill?
The Meta Observation
I realize that while we talk about the future of technology and an emulation of human behavior run purely by artificial intelligence, and how agentic systems are taking away jobs and there is this whole high-noise high-signal thesis in the world around us, we also have to acknowledge that AI is a way of life. We cannot ignore it.
We cannot pretend like it’s not going to exist and shape-shift our jobs and our personal lives for the foreseeable future.
So when I look at community, I really believe that it’s an opportunity to synthesize people. An opportunity to (in real-time) synthesize our existing existentialism, our need for belonging and our dependency on technology — especially when technology will be doing things before we can even think them.
That predictive nature of what we are building through code is going to need a balance of the unpredictive nature of human behavior and human nurturing.
And that’s what communities in AI represent: They’re microcosms of what we’re all building toward. Technology that feels less like extraction and more like collaboration. Products that grow with us, not at us.
The startups that understand community isn’t a feature - it’s a moat - will be the ones still here in five years.
So yes, start that Discord. Launch that WhatsApp group. Host that first awkward meet up where only three people show up.
Because the ones who do show up are not there for the free swag or the LinkedIn post. They’re there because they want to be part of something that hasn’t been built yet.
And that’s the entire point.