Proxy vs. Real

"The original framing — documents are the interface between human and AI, and between AI and AI — is still load-bearing. Edits to a spec directly change what gets planned. Edits to findings directly change what gets fixed. The file on disk is the API."

"Flow now treats every user-facing decision as a structured AskUserQuestion call: 2–4 named options, one of them marked (Recommended), with a one-line rationale per option. Free-form prose is reserved for narration. If there's a choice, it's a structured question."

"If the LLM can't name strong evidence against, the PR body says so explicitly: "I could not find evidence against this thesis, which may mean the spike didn't probe hard enough. Human reviewer should challenge this." That admission is more useful than false confidence. The failure mode for autonomous LLM work isn't producing nothing — it's producing something that looks competent and is quietly wrong, and the adversarial read is the rail that catches the second case."

"Flow is an opinionated answer to where attention should go. Documents at stage boundaries, so you can review state instead of re-deriving it. Structured questions instead of free-form prose, so decisions are clicks instead of conversations. LLM review that fixes mechanical issues silently and surfaces only judgment calls. Spike mode for thesis-validation work where you genuinely don't want to be in the loop. Reflection so the system gets sharper from your archive instead of staying static."

"I'd used 750words.com for years, but it always felt a little heavy. Too many features competing for attention, too many widgets, too much UI. I wanted something that felt closer to sitting down with a notebook and a good pen. So I built it."

"It took me a while to put my finger on it. The answer is this: Mozi is trying to help you see people you already care about. Not discover new ones. Not grow an audience. Not broadcast your life. Just — keep in closer proximity to the people who matter to you."

"Somewhere along the way, the incentive structure shifted. Engagement became the metric — and engagement is best maximized by outrage, novelty, and the variable-reward dopamine loop of infinite scroll. The thing that got you to open the app stopped being "there's something here from someone I care about" and became "I wonder what's happening.""

"The social graph didn't go anywhere. Your friends are still in there. But they got buried under brands, celebrities, algorithmic recommendations, and ads. The feed became a product, and you became the audience."

"Mozi is his bet on a different approach: strip out the media, keep the social."

"The core mechanic: Mozi knows where you are (with your permission), and when you and a friend are in the same area at the same time, it can nudge you both. "Hey, you and [friend] are both in Brooklyn right now." The rest is up to you."

"What doesn't belong in the product is almost as important as what does. There's a constant temptation in consumer tech to add features — a feed, a way to post, a way to react, a way to see who's viewed your profile. Every one of those features would make the product feel more like something users already understand. Every one of them would also pull attention toward the app and away from the person you're trying to see."

"The Harvard Study of Adult Development — one of the longest-running studies on human happiness — tracked hundreds of people for over 80 years. The finding was stark: the quality of your close relationships is the single strongest predictor of wellbeing and longevity. Not wealth, not fame, not achievement. Relationships. And specifically, close ones — not the number of acquaintances you have, but the depth of the few."

"Other research has found that loneliness is as dangerous to your health as smoking 15 cigarettes a day. The US Surgeon General declared loneliness a public health epidemic in 2023. We are, by many measures, more socially isolated than any previous generation — despite being more "connected" than any previous generation."

"In-person time with people you care about is not just nice to have. It's literally protective. The warmth of an in-person conversation activates physiological responses — oxytocin, vagal tone, the kind of nervous-system regulation that screens simply cannot replicate. Video calls are better than nothing. Text is better than silence. But neither is a substitute for actually being in the room with someone."

"The thing I find most interesting about working on Mozi is the constraint it puts on product thinking. Every feature proposal has to answer: does this help you see people in person, or does it substitute for that?"

"What passes the test tends to be coordination infrastructure — tools that make it easier to go from "we should hang out" to actually hanging out. That's a narrower product surface area, but it's the right one."

"I spent five years at Bunch building the opposite kind of social product — one that mediated social connection through a screen, through a game, at a distance. That product had real value. People genuinely felt closer to their friends after a gaming session. But it was always a proxy. Mozi is building for the thing itself."

"There's something clarifying about that. The product question isn't "how do we maximize time in app?" It's "how do we minimize time in app while maximizing time with people?" That's a different company to build. And, I think, a more honest one."

"There's a difference between ambient growth and earned growth. Ambient growth comes to you because of circumstance — a cultural moment, a pandemic, a platform shift. Earned growth comes because users have made your product part of how they live. When the circumstance changes, ambient growth leaves. Earned growth compounds. We had both for a window, and we chose not to ask which was which. The data was there every sprint. What we couldn't do was say out loud what it meant. When the wave receded and people went back to seeing friends in person, we found out what we had: we were too busy making people busy in the app. We hadn't built the thing that made them come back when they didn't have to."

"The problem was what we measured. Every sprint, we'd ship a feature and track whether that feature improved. We never organized the whole effort around a single north star — the number that would tell us whether people had a genuine reason to come back. Without that organizing principle, engineering effort drifted toward polish. Toward things we knew how to optimize. We were too busy making people busy. Solving B+ problems: well-defined, satisfying to ship, and completely beside the point."

"That constraint shaped everything. If I couldn't build a live demo around it — something I could code from scratch in front of them, something they could follow in real time and reproduce themselves — I didn't have a lesson yet. I had an outline. The demo was the lesson. The rest was scaffolding."

"The live demo was my version of the same thing: get out of the way of the content and let them watch the process. Not the polished output — the process. The typing, the mistakes, the backspace, the "oh wait, let me try this instead.""