// Field notes from the founders
AI is single-player but your D2C business is multiplayer
"Power corrupts. PowerPoint corrupts absolutely," Edward Tufte claimed back in 2003. We've all been there: a bazillion-page slide deck in font 9, probably from some expensive consultants who need to justify their spendy day rate. Another small fragment of your soul crumbles off like an aging bath bomb and melts away, never to return. Amazon got the memo a long time ago and (now famously) banned slides outright — every meeting there opens with everyone sitting in silence, reading a written document. They'd worked out something many companies feel but never grasp: the format and medium of the information shape the nature and quality of the interaction. With PowerPoint always one click away and the seductive power to add whizzy diagrams, the fires of PowerPoint hell were still burning brightly until recently — when they were arguably overshadowed and outshone by a newcomer… the AI dump in Slack.You know the message. Someone asks ChatGPT a question. It's probably interesting, and there's a fair chance it's a solid answer, and they paste the entire 600-word wall straight into the channel. At first I tried to read them — I don't like being behind on context in important Slack channels. Even then the success rate was mixed. I'd have other things to get on with, a meeting or some code I was writing, and I didn't want to break my flow. So I'd bookmark the messages, or leave them unread to come back to later. About six months ago I gave up and just stopped trying.And I bet I'm not the only one. Life is too short. The AI dumps keep coming, adding cognitive load and noise to the channel instead of removing it. Multiply that across a team and your Slack becomes a graveyard of brilliant, context-free monologues — the slide deck nobody asked to sit through, reborn as a wall of words nobody reads. The only possible benefit is that the messages get slurped back up into your company brain, and so this depressing loop is closed and we've gone full circle.Both the format and the medium play a part here. A dense block of impenetrable, verbose text stifles readership and engagement wherever it's presented — the same way a mega slide deck overwhelms its viewers and switches their neurons into hibernation mode — but the medium plays its part too.The slide deck informs; recommendations are often crammed in at the end and discussed with two minutes to spare before the meeting closes.The AI dump in Slack festers. Even if someone reads it, it doesn't stimulate the additional interaction required to refine the thinking, align on actions, and set priorities. It doesn't end up on anyone's todo list, and as a result nothing happens.The real issue is that AI today is single-player in a company that is a team sport. It's the second of the three fragments I set out in AI has made your silos worse, and it's going to change, rapidly. My bet is that in as little as six months we'll look back and think "what on earth were we doing".On YC's Lightcone podcast, Pete Koomen named it exactly: we're still in "the single-player era of agents — the harnesses that have got really popular are all designed to be used by a single human." Each person runs their own thread, in their own tool, and the context dies the second they close the tab — or gets dragged into the next tool by hand.Multiplayer AI tools like Viktor and Dust have been around for a while, but adoption is still low. With Anthropic recently launching Claude Tag and entering this space, that's set to change fast. Anthropic say it's now one of the main ways things get done internally, and no doubt we'll soon look back on the single-player era of AI in companies the way parents watch their kids announce exactly where they're about to hide in a game of hide and seek. Cute, but not effective.One of the operators we work with described his actual daily reality: seven different models and tools on the go, and a meaningful chunk of his day spent copy-pasting context between them so nothing got lost in the gaps. He wasn't complaining about any one tool — each was excellent. He was describing the absence of the thing between them: a shared place where context lives and compounds.Garry Tan put the consequence of leaving that unsolved about as bluntly as it can be put: scale single-player AI across an organisation and "without a persistent, shared source of truth, you don't get a cohesive virtual organisation — you get a thousand disconnected, hallucinating micro-teams stepping on each other's work."That's not a distant risk. That's most D2C teams' Slack today. If it doesn't resonate, look again.The fix isn't a better prompt or a better paste. It's making AI multiplayer: shared, persistent sessions where context accumulates across the team instead of dying with each individual thread.In that world, "have you seen this?" means handing someone the whole working session — the question, the data it pulled, the reasoning, the decision — not a screenshot of an answer with all its context stripped away. The session works like any other conversation in Slack, with people and the agent popping up as required. The model remembers what the team decided last week, and why. It's the Amazon memo versus the slide deck: a shared, structured thread that invites interaction between the team and the agent, versus the context-free AI dump.This is the part of the "company brain" that the operators we talk to get excited about — far more than any dashboard. Not "the AI gave me an answer," but "the AI gave my whole team the same context, and it got smarter every time we used it."This space is developing rapidly and will grow exponentially over the next year. Our view, though, is that to really solve problems like this you have to go deep — and we're going deep to build an intelligence layer for D2C that is multiplayer from the ground up.Check out Signal over Noise to apply for the Pilot!