A machine that reads the common conscience of online crowds, starting with Twitch chat.
Video games, esports and online communities have been my biggest obsession since I was a kid. I have spent more hours in Twitch chat than in any meeting room, and I have founded two companies in spaces I cared about far less. This is the first thing I am building in the space I actually love, and I want to say plainly what it is and how far I intend to take it.
Take a stream with thirty thousand people watching. The chat scrolls faster than anyone can read. Every message is short, half of them are emotes, and most of them are reactions to the same thirty seconds of video. Nobody moderates it into sense, nobody summarizes it, and the person it is addressed to, the streamer, cannot read it either. They glance at it between deaths and catch one line in a hundred.
And yet something forms in there. A mood turns. A joke catches and the whole room repeats it. A decision on screen gets judged in real time by a crowd that agrees with itself within seconds and then forgets it agreed. It is a common conscience, assembling and dissolving many times an hour, in public, in text, with timestamps. I do not know of another place where a crowd thinks out loud like this, at this speed, where anyone can read the transcript.
The first product reads the room for the streamer. Which message is worth responding to right now. Whether the mood is turning, and on what. What the chat wants that the streamer cannot see because they are busy playing. That is a real job today, done badly by glancing, and the people who do it for a living are the first people I am talking to.
I am running discovery calls with streamers now, starting with the ones I have watched for years, and I am asking them about their days, not about my idea. If the first version is right, it will be small: a second screen that makes a streamer better at the thing they already do. Small is fine. DataPrysm ran close to a hundred discovery calls wide and one signed contract deep, and the lesson I kept is that the wedge has to be something one specific person wants on a Tuesday.
Twitch is the training ground because it is the fastest and most public crowd there is. The same machine, once it reads one room well, can read others: a subreddit, a Discord, a comment section, and eventually the platforms where large groups of people react to the news at once.
The far end is the part I hold loosely and say anyway. I want to predict, with high confidence, what a large group of people is about to feel, intend and eventually do, early enough for someone to act on it. Crowds produce the worst outcomes we know: panics, wars, the slow slide toward both. A machine that reads them early could prevent some of that, and I would like to have built it.
Isaac Asimov called the finished version of this psychohistory, a mathematics of crowds that could see centuries ahead. That is fiction. Foundation Zero is the name for the ancient ancestor of it: the first machine that reads one crowd in one room in real time, and gets it right often enough to matter.
I have the obsession, two companies’ worth of scars, and a fleet of agents that lets one person build faster than the two of us could two years ago. I do not have evidence yet. I have not shipped a line of it, and the streamers I am talking to have not told me they want it. That is the correct order, and I am writing this down before the first call so the record shows what I believed at the start.
If you stream, run a community, or build in the gaming space and any of this sounds like your problem, I would love to talk. Skin in the game and boots on the ground beat a great idea every time.