I’ve been thinking a lot about how we share information in the AI space — and more specifically, how we do it BADLY.
If you’re anything like me, you’re in a dozen group chats, Slack channels, Discord servers, and email threads all supposedly dedicated to “AI news and discussion.” And most of them are terrible. Not because the people are bad, but because nobody set expectations upfront.
I’m in the process of setting up a more structured way to share what I’m learning, and it forced me to think about what actually makes an AI community work versus what makes it another dead channel full of link dumps and self-promotion.
The Information Firehose Problem
Here’s the thing — there is TOO MUCH happening in AI right now. Every single morning, there’s a new model drop, a new benchmark, a new startup raising an absurd round, a new open-source project that’s genuinely impressive. I’ve written before about everything from function calling to local models to the open vs. closed debate. The pace hasn’t slowed down. If anything, it’s accelerating into the end of 2024.
The natural response is to share everything. And that’s where groups fall apart. When everyone shares everything, nobody reads anything. Your signal-to-noise ratio drops to zero and people start muting the channel. I’ve watched it happen over and over.
So what actually works?
Four Rules That Actually Matter
After watching plenty of groups succeed and fail, I’ve landed on a pretty simple framework. Four rules. That’s it.
1. Focus the conversation. Share news, ask questions, debate. That’s the core loop. If a group doesn’t have a clear purpose beyond “AI stuff,” it becomes a dumping ground. The best groups I’ve been part of have a tight focus — not so narrow that there’s nothing to talk about, but narrow enough that people know what belongs and what doesn’t.
2. Welcome project updates, but keep them concise. This is the one most groups get wrong. Either they ban self-promotion entirely (which kills the most interesting content) or they let people write five-paragraph essays about their side projects every week. The sweet spot is simple: share what you’re building, keep it short, drop a link for anyone who wants to go deeper. Nobody needs your full product roadmap in a group message.
3. No spam. Zero tolerance. This sounds obvious but it’s the rule most groups fail to enforce. The moment you let someone drop an affiliate link or a “check out my course” post without consequence, you’ve set a precedent. Violators need to be removed immediately. Not warned. Removed. The group’s value is its signal quality, and one spammer can destroy months of trust-building.
4. Make it a safe space for real talk. In a close-knit group, people should feel comfortable saying “I don’t understand this” or “I think this technology is overhyped” without getting dogpiled. The AI space is FULL of hype merchants and tribal thinking right now. You’re either a doomer or an accelerationist, a GPT loyalist or an open-source purist. The most valuable groups are the ones where people can share genuine analysis without performing for an audience.
Why Curation Beats Aggregation
One thing I’ve started doing is building a curated reading list — pulling together the most interesting AI developments each day rather than just forwarding every article that crosses my feed. The difference between curation and aggregation is MASSIVE. Aggregation is easy. Curation requires judgment. It requires someone to actually read the material, decide what matters, and present it with context.
I think this is where the real value lives in 2025 and beyond. We’re not short on information. We’re short on people who can filter it, contextualize it, and share it in a way that respects everyone’s time. That’s true whether you’re running a small group chat or building a media company.
The Inbox Respect Factor
Here’s my final thought on this — and it’s the one that most group admins completely ignore. You have to give people an easy out. If you’re about to ramp up the volume of what you’re sharing, TELL PEOPLE. Let them opt out gracefully. Nothing kills a group faster than members who are silently resentful about the notification volume but too polite to say anything. They just mute it and never come back.
The best communities are the ones where the cadence is predictable, the content is focused, and the exit door is clearly marked. Sounds counterintuitive, but making it easy to leave makes people want to stay.
The Bottom Line
If you’re thinking about starting an AI-focused group or community — whether it’s five people or five hundred — set the rules BEFORE you start sharing content. The rules aren’t bureaucracy. They’re the foundation that lets the actual interesting stuff happen. Focus, brevity, zero spam tolerance, and psychological safety. That’s the whole playbook.
Get those four things right and you’ve got something worth protecting. Get them wrong and you’ve got another dead Slack channel.
