Over the weekend, another social media platform exploded into the fray: AirChat. The app is like a combination of Twitter and Clubhouse. Instead of typing a post, you speak it. The app quickly transcribes what you say, and as your followers scroll through their feed, they’ll hear your voice alongside the transcription.

Built by AngelList founder Naval Ravikant and former Tinder exec Brian Norgard, Airchat takes a refreshingly intimate approach to social media. There are people I’ve known online for years, and only after following each other on AirChat did I realize I’d never heard their actual voices. The platform makes it feel like we’re actually having conversations with one another, but since AirChat is asynchronous, it doesn’t feel as daunting as joining a room on Clubhouse and having live conversations with strangers.

Posting with your voice may sound scary, but it’s not as intimidating as it seems — you can re-record your post if you misspeak. But if you’re someone who loves sending your friends three-minute voice memos instead of typing (or if you have a podcast), AirChat feels intuitive.

AirChat wouldn’t be worth using if the transcriptions were sub-par, but it’s the best speech-to-text product I’ve ever used. It almost always hits the mark in English… it even transcribes Pokémon names correctly (yes, I tested this extensively). It also seems to be doing well in other languages — I found it functional in Spanish, and TechCrunch reporter Ivan Mehta said that the app did a decent job transcribing Hindi. Sometimes, the app will translate speech directly to English, and while the translations were generally correct in our testing, it’s not clear why or when the app translates instead of transcribing.

So, is AirChat here to stay? That depends on what kind of people can find community on the platform. For now, the feed feels like a San Francisco coffee shop — most of the people on the app have some connection to the tech industry, which could be because tech enthusiasts are often the first to jump on new apps. This wasn’t the case for Threads when it launched (it’s just an extension of Instagram), or even Bluesky, which developed an early culture of absurd memes and irreverence. Right now, the app has paused invites, so this won’t improve in the near future.

The app’s current culture could also be a reflection of its founders, who are influential in Silicon Valley and venture capitalist circles. But it’s telling that when AirChat introduced a channels feature, two of the first to spring up were “Crypto” and “e/acc,” which stands for effective acceleration, an aggressively pro-tech movement.

This doesn’t have to be an automatic red flag — I (somewhat reluctantly) use Twitter/X every day, and the tech industry also feels especially loud there. But at least on X, my feed also contains posts about my favorite baseball team, the music I like and the ongoing debate over adding more bike lanes in my neighborhood. So far on AirChat, I haven’t seen many conversations that aren’t about tech in some way.

What I do consider a red flag is AirChat’s naïve approach to content moderation.

“We’re going to try and put as many of the moderation tools in the hands of the users as possible. We want to be as hands-off as possible. That said, sometimes you just have no choice,” said Ravikant on AirChat.

The phrasing of “hands-off” is reminiscent of Substack, a platform that lost popular publications like Platformer and Garbage Day after it refused to remove pro-Nazi content proactively.

AirChat did not respond to TechCrunch’s request for comment.

Ravikant argues that AirChat should function like a dinner party — you won’t kick someone out of your house for partaking in a civil debate. But if they start violently screaming at you, it would be wise to intervene.

“We don’t want to moderate for content, but we will moderate for tone,” Ravikant said.

In real-life social situations, it’s very normal behavior to disagree with someone and explain why you think differently. That’s a pretty manageable situation to handle at your own dinner table. But AirChat isn’t a normal social situation, since you’re in conversation with thousands of other people; without more robust content moderation, this approach is like hosting a big music festival, but with only one person working security. One might hope that everyone will enjoy the music and behave themselves without supervision, but it’s not realistic. Just look at Woodstock ’99.

This is another way in which AirChat parallels Clubhouse. Clubhouse’s approach to content moderation was even more permissive, since there was no way to block people for months after launch — AirChat already has block and mute features, thankfully. Clubhouse repeatedly played host to antisemitic and misogynistic conversations without consequence.

With this minimalist approach to content moderation, it’s not hard to see how AirChat could get into hot water. What happens if someone shares copyrighted audio on the platform? What about when someone doxes another user, or if someone uploads CSAM? Without an actual plan to navigate these situations, what will happen to AirChat?

I hope that people can behave themselves, since I think the concept behind AirChat is brilliant, but we can’t be so naïve. I would like to know that if neo-Nazis tried to politely explain to me why Hitler was right, the platform would be able to protect me.