I built a decision protocol API. Here's why calling it is different from calling GPT-4 directly.

Zyvv — Fri, 12 Jun 2026 15:55:03 +0000

Most AI APIs return text and forget it happened.
ZYVV API does something structurally different: every call you make writes a permanent row to a choices table — which door a user picked, the doubt they raised, the refined path they took. That dataset doesn't belong to you. It belongs to the collective. And it compounds.
Here's what the call looks like:

curl -X POST https://zyvv.app/api/v1/doors \
-H "Authorization: Bearer " \
-H "Content-Type: application/json" \
-d '{"situation": "I keep starting businesses that get traction but I always burn out at month 4"}'

Response:

{
"situation_id": 4821,
"roast": "You're not burning out at month 4. You're succeeding at month 4 and panicking because success requires a different identity than hustle does.",
"doors": {
"conventional": {
"title": "Hire before you break",
"description": "The burnout is a hiring signal, not a willpower signal. Month 3 is when you bring in the first operator.",
"why_it_works": "Founders who burn out at scale are almost always doing work that a $4k/month hire could handle. The business isn't failing — you're under-leveraged."
},
"contrarian": {
"title": "Design the exit before you start",
"description": "Stop building to scale. Build to sell at month 6. The constraint forces clarity.",
"why_it_works": "A business built to be sold in 6 months has a cleaner structure, tighter metrics, and a natural handoff. Most solo founders burn out because they built something that only works if they're in it forever."
},
"alien": {
"title": "The problem isn't the business. It's the identity gap.",
"description": "You keep framing this as an endurance problem. It's a self-image problem. Month 4 is when you'd have to become a manager. You're allergic to managers. Fix that first.",
"why_it_works": "Every business you've built has hit the same wall at the same stage. The wall isn't external. You built it."
}
}
}

That situation_id carries through the full protocol. Your next call records which door they chose:

curl -X POST https://zyvv.app/api/v1/choice \
-H "Authorization: Bearer " \
-H "Content-Type: application/json" \
-d '{"situation_id": 4821, "door_type": "alien"}'

Then interrogation — the user raises a doubt, the protocol stress-tests it:

curl -X POST https://zyvv.app/api/v1/interrogate \
-H "Authorization: Bearer " \
-H "Content-Type: application/json" \
-d '{"situation_id": 4821, "door_type": "alien", "objection": "I don'\''t have money to hire a therapist or coach to fix my identity"}'

Three endpoints. The full Talmudic protocol. Every call writes to the same situations, doors, and choices tables as the consumer app at zyvv.app.
Why that matters: in 6 months, ZYVV will know exactly why people reject the Alien door. The "no money for coaching" objection will be addressed by the door itself — proactively, before the user raises it. The protocol gets smarter. Every API call you make contributes to that.
You're not just calling an AI. You're building the dataset.
API key requests: [link] | Full endpoint reference: API.md

DEV Community: Zyvv

I built a decision protocol API. Here's why calling it is different from calling GPT-4 directly.