We want Bonzai to help people find their people, not just write posts
While packing the dishwasher, I talked through what could happen on Bonzai after someone hits publish: an open door at the end of each story, roll-ups that connect contributors and niches where sharing feels like talking to trusted colleagues.
The thought arrived while I was packing the dishwasher. Nina had asked what would ease the friction for someone who is busy but has something to say, and my honest answer was that the button isn't the hard part. We're making it one press. The harder questions are whether people are comfortable sharing with Nina, whether that trust has been built and whether they have the time in that particular minute. I happened to have the time right then.
Bonzai, which I'm building with my co-founder, Jeff, has a concept we call your AI journalist: Nina interviews you, and a story comes out of the conversation. It would be easy to describe that as an AI writing tool. But this isn't just a matter of making an AI writing tool. What I've been thinking about is the network beyond publishing: the ideas and people that could gather around what gets written.
It starts with the post itself. Seeing a lovely post is really the first magical moment. In that first conversation, I'd want someone to get the feeling that a good article is being written. It should feel like something they would write themselves. I know there are professional writers out there, but it should sound like them, cover the topics they discussed accurately and feel really well written. It should also have a beautiful image and great formatting, and be almost ready to share.
Seeing a lovely post is really the first magical moment.
To some degree, that's less about voice than it sounds. We're looking for a published, polished, journalistic voice. What has to be theirs is the subject matter and the content of what they discussed, along with any instructions they've given about voice. If someone asks for their next post as a poem, it should be a poem.
Trust comes down to a quick read or scan to see whether it sounds like them. We do have a listen button, so they can hear the content, and I find listening is often a better way to understand whether content is any good. When I use it, I'm screening for inaccuracies and for whether it sounds good overall. I'm not really overthinking it more than that. It's AI, so you're a little skeptical about what it might have written.
The next step is for somebody other than the author to discover it. Early on, we would imagine people sharing their posts by text, email or social media, and that might draw others in to look at the content. For people already on the network, we would have outbound emails and a news feed they can look through. That would be our own attempt to engage people they wouldn't otherwise have met.
Then there's what a reader does with a post. They've read it, and they sort out the ways their own experience of the world aligns with it or disagrees with it. An AI interviewer can pull on the thread the article started. So I'm thinking we can have calls to action at the end of articles inviting people to add their own take on the topic. I want that to feel like an open door, not a nudge.
The journalist would let them take it in whatever direction they want. Maybe they give a quick reaction; for a comment, I imagine it could be incredibly quick. Hopefully that leads to more commentary, so we'd ask whether they have more they want to say. If Nina is quick to set up, she should be able to pull on that thread, so whether someone takes the longer step probably has more to do with their own time and inclination to engage with her.
I suspect this would mostly be for people on their second, third or fourth post: power users who've shared a few articles and are leaning into the service in a deeper way, people who have already established some trust with Nina. I haven't thought through whether someone's first use might be responding to someone else's post rather than creating a new one.
What would make someone a repeat contributor? A density of content around their chosen niche. Maybe it's steam trains. Maybe it's diffusion models instead of large language models, or some other niche like that. And the feeling that they're sharing their thoughts with colleagues, really: trusted colleagues, or trusted comrades in their chosen niche.
It would tap into people's need for belonging. We have an idea for roll-up articles that pull in content from multiple contributors. Once a few diverse opinions on a topic have emerged from a few different authors, and maybe different posts from the same author over time, we would gather them into a roll-up post, publish it and somehow let the contributors know they were part of it. That would let them feel connected to one another, I'm guessing.
The smallest signal a quietly curious reader might give is perhaps subscribing to other people. That's a quiet signal; the person could receive a notification that they have a new subscriber. But ultimately we want to move people to publish, and the nudge is really to get them into a conversation with the voice system we have.
If someone does step through that door, what I'd hope they feel is that they have found their people. When does that happen? Probably, actually, when somebody responds to their post.