I recently had the chance to speak with Ben Horwitz, the creator of Sinceerly, a satirical tool that makes your AI-generated emails more human by making the language more casual and adding typos and other human-seeming elements. I find Sinceerly fascinating because it touches upon so many interesting aspects of the debate about AI writing. Does anything fundamentally change when you put typos into your AI output? At the time he launched the product, Ben was an MBA student at Harvard Business School; as of recently, he now works at OpenAI. This interview has been lightly edited for clarity.

Micah Blachman: What was your goal in creating Sinceerly?

Ben Horwitz: I think my goal was twofold. One was that I wanted people to hold a mirror up to themselves as they use AI and think about how they use it and how their usage might have evolved over time. When we first started using AI, it was maybe not being used everywhere. Now folks are copying something from their ChatGPT output and putting it into their email. I wanted these people to look at themselves as they use AI and make AI more human. I think that's kind of funny and maybe a little sad. The second part is that there might be some value in a tool that helps people get more email responses, I guess.

MB: What have been users’ responses to the tool?

BH: Most people find it funny, and a few people have found it useful. There are some people who are paying for it. It's interesting: I included a way that folks could use their own API key to link the tool to their AI accounts. If you are tech savvy enough to do that, you should not be paying me. Sorry, not shouldn't be paying me, but don't need to be paying me. You can use your own underlying tech. That's really why I wanted it to be free. I think that it should just be a fun tool and not a tool that really changes how we think and write. 

MB: How do you feel about the ethics of AI-generated writing?

BH: I don't think that most writing should be AI. I especially think that the process of writing is so interesting and human. Part of where we are as a society is because of how we think about and structure arguments. When everything is structured in the same-ish way, not that it's entirely deterministic, but if it follows a chain of thinking that is similar, I don't love that. I think that's part of why this is satire. I want people to not just say, “Oh yeah, more people are going to read my e-mail because I used this tool,” but also, “Wait a minute, why am I putting so much work into having someone else do the work?” I also think typos are interesting because they are sort of our final frontier in determining whether I am talking to a human or to AI. 

But you’re a journalist—what do you think about AI in writing? 

MB: I don't love when AI is used to write text. I think a lot of what makes writing interesting is the perspectives and the opinions, which AI doesn’t have.

BH: And what about your em-dash? I’ve read some of your articles, including the one about Suno, and a few others. You’ve got em-dashes. Is that your way of staking a claim and saying, “I’m going to own this?”

MB: People see em-dashes and immediately think “AI,” but they’re perfectly valid punctuation. I think it's not really right to dismiss any em-dash writing as AI, because it's not necessarily AI. Though I’ll admit I've been starting to put two dashes instead of the em-dash. So I can differentiate myself.

BH: That's fair. 

MB: Do people really want to seem unprofessional or casual in their emails? Aren’t they using AI in the first place because it’s making them seem more professional or readable?

BH: You’re hitting on a good point. There’s a world in which the people who are using Grammarly might not be the same people who like Sinceerly, which is the anti-Grammarly. The people who might love Sinceerly besides the people like me who love the satire are the people who might care a lot about an email reply. The other folks might just not want to seem like idiots. These two outputs, I don't want to seem like an idiot and I want to get a reply, might be the same for some people, but for others, they might be really different. 

So my idea is for different use cases. When I'm texting my friends, there’s no need for AI. I don't really care if I have typos because I love my friends and hopefully they'll love me back. With regard to emails, I don’t really know. When you're doing outbound sales, for example, you'll do anything to get the attention of the person on the other side—so maybe a typo helps.

I might use the word satire too much, because most people think it’s real. You can download it, you can use it, you could pay for it. There are people who are going to continue to do that. Instead of satire, maybe I should call it “fun.” This is a fun thing, not “We're going to improve your outbound sales by 10%.” My belief is that life is just a little bit more fun when you introduce some levity to something that is pretty scary. At this point, everyone’s worrying that AI is coming for jobs. Sinceerly is a form of social commentary because I think it's important for us to think about it, and maybe have some fun with it and realize that it's making it easier to make stuff. In my mind, something of quality is something that makes you think.

MB: How much media have you gotten from Sinceerly?

BH: I'd say a lot. I really didn’t expect this response at all. I’m a student, I’m not doing this as a full-time job. When you're not doing something full time as a job, people might afford you a little bit more latitude—it's okay that I don't have my pricing plan laid out perfectly one step after another. 

The press has been great because the people interested in how technology and humans affect each other tend to be thoughtful, even when we disagree.

MB: I see Sinceerly almost as the reverse of something like ChatGPT or Claude. The output is, in many ways, a regression from the input.

BH: I think you’re right. People don’t really need to do this. If you are talking to ChatGPT or Claude in your ChatGPT or Claude window, you could ask it to add a typo, or to make it 40% shorter, or to introduce a skill so you don’t have to do it every time—Sinceerly just packages it up for you. Is that laziness? Is it that they didn’t think to do that? 

MB: Has being at Harvard Business School in any way influenced what you’re doing with Sinceerly?

BH: From the time I was your age, my dad encouraged me to write to people. If I thought a product was great, if I thought it was bad, or if I liked an ad that I saw on TV, he encouraged me to let someone know, because that’s someone's career or their life or their work—so you should communicate that. My whole life, I’ve been focused on letting someone know if I think something is wonderful. The way that's translated into business school and into college, and even when I was in high school, I’d write to people and I’d let them know. Now my inbox is filled with junk. I'm sure even yours is. But I still think it's wonderful to write to people. It's important to tell people what you think and business school has maybe accelerated that. It’s emboldened me and going to Harvard has let me reach out to people and say, “Hey, I would like to learn about what you do in an academic forum.” Business school gave me a platform or rather a forum to do that, in which I can share with friends and other people who are curious about what his take is. So in that way it was maybe an accelerant, but that underlying thing was always there. 

MB: I also saw that you made a fake Thomas Edison profile on LinkedIn. Do you think your motivation for that was similar to Sinceerly?

BH: People take themselves so seriously on LinkedIn. It’s competitive and stressful—I get it, you want to find a job. Also, when everyone’s a thought leader, I like to think about who is an inventor or making something wonderful. When I was in seventh grade, I wrote a report on Thomas Alva Edison. I had this thought that all of these people are talking about being a founder and building in public and using these sort of jargony ways of describing not the output, but kind of the process of building, and how that connected to the quote of “I learned how not to make a lightbulb 999 times.” It was clearly meant to be playful. I was not trying to actually make a real Thomas Edison account. I mean, I did make a Thomas Edison account. And the posts I thought were actually very funny to me and to others. But there too, maybe big tech doesn't have a sense of humor. I think Steve Jobs would’ve liked it, though.

MB: In Sinceerly, there’s a CEO mode that includes extra mistakes, is in all lowercase, and is very casual. Are all top CEOs like that?

BH: I have emailed back and forth with quite a few CEOs, and I don't think all of them are that way. There’s an article in the Wall Street Journal about how CEOs write. It's in the context of the Jeffrey Epstein emails, which was not the inspiration for this at all. It basically asks the question, “Is this sloppiness intentional? Is it learned? Is it a power play? Is it all three?” What are the levers that are being pulled intentionally or otherwise when someone does send a sloppy email? 

In some ways, it could be read as, “I’m busier than you. I don’t have time.” And I actually think products like Wispr Flow and others might make it easier for CEOs to be more formal and brief. My personal opinion is that we find ourselves at an inflection point in how people are writing that is extraordinarily divergent. In the past, where maybe many CEOs were very brief, some are now longer.

I think it might also be that they no longer have to prove their competence as much because no one can really do anything if they send an email with lots of typos. Once you’re at the top, you’re no longer evaluated for your emails.

It's funny, my first job out of college, I was in client services, and I'm a really bad typist naturally. They really did come down on me hard and said that I couldn’t include typos in client emails. Grammarly was great and really helped.

In my opinion, quirks like double dashes (--) and typos are what make our writing so unique and distinctive when compared to AI writing. When we outsource this to AI, we lose the mark of authenticity that these provide. I find Sinceerly so interesting, because it shows a possibility of a future in which there is no way to distinguish human writing from that of AI, or perhaps that human writing will no longer exist.

The invention of the typewriter forced those who previously concealed their bad spelling with bad handwriting to either learn to spell or always be shamed for their mistakes. AI has done the opposite, allowing us to conceal our own typos. We no longer need to know how to spell, nor do we have to be shamed for our mistakes. This is why we need a way, urgently, to reliably ensure that writing is human. Sinceerly shows just how much we need such detection.