I did not set out to start a history company. I set out to build software people actually use, and history kept pulling me back.
The moment it clicked was in a political science seminar at Columbia. We were not reading textbook summaries. We were reading the primary sources. The actual letters, the speeches, the diaries. And there was a person on the page, arguing with themselves, doubting, changing their mind in real time. Textbooks flatten all of that into a tidy paragraph with a date attached. The real record is alive. Most people never touch it, because it is scattered across archives, dense, and written in a voice from another century.
So I kept asking one question. What if you could talk to the source directly.
Not a cartoon version. Not a chatbot handed a costume and told to improvise. The real thing, grounded in what a person actually wrote and said, and honest about where the record runs out. That question would not leave me alone. It became Voices of History.
The easy version is the wrong version
Here is the trap. Anyone can tell a language model to pretend to be Abraham Lincoln. It takes one sentence. It will sound confident and it will be wrong, because the model is guessing at a vibe instead of answering from the record. That is the version of this idea that already exists everywhere, and it is exactly the version we refuse to ship.
The thing we actually care about is conversational accuracy. Every figure is grounded in real primary sources. When the record is clear, the figure speaks from it. When the record runs out, the figure says so instead of inventing a quote. That constraint is the hard part, and it is the whole point. A history product that makes things up is not a history product. It is a party trick that happens to be about the past.
Getting this right means real work under the hood. Retrieval over a corpus of what each person wrote. Careful grounding so answers trace back to something real. Voice, so a figure can actually speak and not just type. None of that is glamorous. All of it is the difference between a toy and a tool.
Yap Royale is where it starts
The first place you can feel this is Yap Royale, our game. It is the fun front door to a serious idea. You talk your way through history against figures who push back, who have their own goals, and who stay in character because the character is built from the record instead of made up on the spot.
Games are the right starting point for a reason. They are where people are willing to spend real attention, and attention is what history has always been starved for. If a fourteen year old will argue with a historical figure for twenty minutes because it is genuinely fun, we have already done something classrooms have tried to do for a century. The game is not a gimmick bolted onto an education product. It is the wedge.
From there the same engine reaches into classrooms, museums, and media. A teacher can put a primary source in front of students and let them question it directly. A museum can let a visitor talk to the person behind the exhibit. The technology is the same. The bar for accuracy is the same. Only the surface changes.
Why me, and why now
I have spent the last few years shipping software on tight timelines, from live multiplayer systems at a poker startup to mission control software for a NASA challenge that my Columbia team tested at Johnson Space Center. Different domains, same habit. Build the smallest real thing, put it in front of people, and let reality correct you fast.
Now is the moment because the tools finally exist. Retrieval, grounding, and voice all crossed the line from research demo to something you can actually build a product on. Two years ago this idea was a nice essay. Today it is a company with users.
I am building Voices of History because the past deserves better than a paragraph and a date. The people who lived it left a record, in their own words, and that record should be something you can talk to. Not a fake version of them. The real one, close enough to talk to, honest about its own limits.
That is the company. This site, and this post, are the reveal. If any of this pulls at you the way it pulled at me, come find me.