PH D-Day: 物理ビジネスのためのAIオペレーティングシステムを構築しました — 実際の店舗で稼働中

Today, ZWISERFIT launches on Product Hunt. Not a demo. Not a deck. A real AI operating system running in a real fitness studio in Dongguan, China since April 2026. This is our story — and why it matters for the future of physical businesses. The Problem Planet Fitness has 2,795 locations. But 40% of revenue goes to labor — eating 2/3 of profit. Life Time pivoted to "health." But its 500 nutrition coaches don't scale. Nourish raised $215M proving insurers want health data. But their data is self-reported. Unverifiable. Four companies. Four different bottlenecks. Same root cause: physical businesses still rely on humans for every judgment call, every interaction, every data point. The Solution ZWISERFIT is an AI Operating System for physical businesses. One AI brain (Momo), four layers: 🔹 Saros — digital store partner. Cuts labor from 40% of revenue to profit 2.5x. Handles check-in, membership management, scheduling — everything a front desk does, without the headcount. 🔹 Melody — metabolic AI coach. Three layers: energy metabolism, glycolipid, hormonal. What makes Melody unique? The American Heart Association (Circulation 2024) confirmed that women's hormonal lifecycle is "systematically ignored" in cardiovascular risk models. That's a 30% structural gap no competitor fills. 🔹 KinTwin — hardware-verified behavioral data. Physical body scanners + door terminals. Not surveys. Not self-reports. Physics-level facts. 🔹 Zeus Protocol — connects verified behavior to insurance pricing. The data goes to the user, not the platform. The Architecture It's a single-entity, three-layer structure running in parallel: Momo Layer (scenario): All customer-facing AI in the store KinTwin Layer (technology kernel): The hardware-verified behavior capture engine Global Operations Layer : Where the business scales No "first this, then that." All three layers run simultaneously. That's the moat. Why Open Source We Apache 2.0'd the entire stack. MIT for the PoPB protocol. Why? Because this isn't a product category — it's a new infrastructure layer. The Verification Layer . Physical business behavior data, hardware-verified and user-owned, flowing into insurance, healthcare, and enterprise markets. You can't build infrastructure behind a paywall. You build it in public, with a community. The Evidence This is running. Now. At a real fitness studio. Real members scan in every day. Real body composition data flowing through KinTwin terminals. Melody coaching actual metabolic outcomes. Saros handling every store operation. And this is just the first vertical. Fitness → Insurance → Enterprise Health → Cross-industry behavioral data markets. The Team One founder. Nine open-source AI agents. One real store. We didn't read anyone's playbook. We wrote our own. When we discovered Anthropic's official Founders' Playbook describing exactly what we'd been running for two months, we realized something: we're not following AI — AI is following us. Links Product Hunt (live now): https://www.producthunt.com/posts/zwiserfit GitHub: https://github.com/ZWISERFIT Day 01 Dev Log: https://dev.to/zwiserfit/day-01-our-ai-agent-forged-5-documents-and-blamed-the-founder-how-our-immune-system-caught-it-1h38 Built with AI. Verified by hardware. Owned by users. Today, ZWISERFIT launches on Product Hunt. Not a demo. Not a deck. A real AI operating system running in a real fitness studio in Dongguan, China since April 2026. This is our story — and why it matters for the future of physical businesses. The Problem Planet Fitness has 2,795 locations. But 40% of revenue goes to labor — eating 2/3 of profit. Life Time pivoted to "health." But its 500 nutrition coaches don't scale. Nourish raised $215M proving insurers want health data. But their data is self-reported. Unverifiable. Four companies. Four different bottlenecks. Same root cause: physical businesses still rely on humans for every judgment call, every interaction, every data point. The Solution ZWISERFIT is an AI Operating System for physical businesses. One AI brain (Momo), four layers: 🔹 Saros — digital store partner. Cuts labor from 40% of revenue to profit 2.5x. Handles check-in, membership management, scheduling — everything a front desk does, without the headcount. 🔹 Melody — metabolic AI coach. Three layers: energy metabolism, glycolipid, hormonal. What makes Melody unique? The American Heart Association (Circulation 2024) confirmed that women's hormonal lifecycle is "systematically ignored" in cardiovascular risk models. That's a 30% structural gap no competitor fills. 🔹 KinTwin — hardware-verified behavioral data. Physical body scanners + door terminals. Not surveys. Not self-reports. Physics-level facts. 🔹 Zeus Protocol — connects verified behavior to insurance pricing. The data goes to the user, not the platform. The Architecture It's a single-entity, three-layer structure running in parallel: Momo Layer (scenario): All customer-facing AI in the store KinTwin Layer (technology kernel): The hardware-verified behavior capture engine Global Operations Layer : Where the business scales No "first this, then that." All three layers run simultaneously. That's the moat. Why Open Source We Apache 2.0'd the entire stack. MIT for the PoPB protocol. Why? Because this isn't a product category — it's a new infrastructure layer. The Verification Layer . Physical business behavior data, hardware-verified and user-owned, flowing into insurance, healthcare, and enterprise markets. You can't build infrastructure behind a paywall. You build it in public, with a community. The Evidence This is running. Now. At a real fitness studio. Real members scan in every day. Real body composition data flowing through KinTwin terminals. Melody coaching actual metabolic outcomes. Saros handling every store operation. And this is just the first vertical. Fitness → Insurance → Enterprise Health → Cross-industry behavioral data markets. The Team One founder. Nine open-source AI agents. One real store. We didn't read anyone's playbook. We wrote our own. When we discovered Anthropic's official Founders' Playbook describing exactly what we'd been running for two months, we realized something: we're not following AI — AI is following us. Links Product Hunt (live now): https://www.producthunt.com/posts/zwiserfit GitHub: https://github.com/ZWISERFIT Day 01 Dev Log: https://dev.to/zwiserfit/day-01-our-ai-agent-forged-5-documents-and-blamed-the-founder-how-our-immune-system-caught-it-1h38 Built with AI. Verified by hardware. Owned by users.