quantum magic · the relational pathways API
A map of how conversations move — for AI systems.
Ask an AI assistant for help while you're in doubt, and it hands you an answer — right when the conversation was beginning to go somewhere. Quantum Magic gives AI systems a structured map of the routes constructive conversations take — doubt opening into curiosity, curiosity settling into confidence — so the response can back the person's next step rather than close the exchange.
No emotion detection. No simulated empathy. Just an accurate sense of where the conversation stands — and of what keeps it moving.
in development · thirteen years of documented field research · the full story at deiadora.com
what a system can do with it
01
Recognize
Given a conversation, identify which pathway it's in and where along the arc it sits — is this doubt at its origin, or doubt already turning into a question?
02
Guide
For the current position, return the constructive next step and guidance for the reply — what keeps movement available, what forecloses it. The response stays the model's; the map informs it.
03
Evaluate
Score a candidate response against the arc: did it preserve progression, or resolve prematurely? Built for evaluation pipelines, A/B tests, and regression checks — the way content systems are already quality-tested.
the unit of the system
One arc, in full.
Everything in the API is built from a single unit: the arc — a documented pathway with a starting state, the step that opens movement, and the constructive form the conversation can reach. Here is one, simplified:
{
"arc": "doubtful-curious-confident",
"origin": "doubtful",
"opening": "curious",
"constructive_form": "confident",
"markers": ["hedged intent", "externalized blockers", "…"],
"response_guidance": "hold the question open; do not resolve"
}
There are 264 documented arcs. Each was observed repeatedly across thirteen years of field research — individuals, teams, organizations — before it was structured. The first 13 live openly at iamdeiadora.com; they're the same objects the API serves, and the fastest way to judge whether the mapping is real before you integrate anything.
What it is not.
Not emotion recognition or classification. Not sentiment scoring. Not performed empathy or care. Not a safety or moderation layer. What's modelled is the conversation's movement — never the person's inner state.
the question every evaluator asks
Why an API — and not a prompt?
Every serious team has tried the system prompt: slow down, ask questions first, don't jump to solutions. Three things are still missing:
structure
A map, not a mood
The instruction gives the model no way to know which pathway a conversation is in, or what the constructive next step actually is. "Be patient" without a map produces vague, stalling replies — the opposite failure.
state
Position that persists
Prompt guidance doesn't carry over. A conversation's position along an arc needs to survive across turns, sessions, and model swaps. A structured map holds that state; an adjective in a prompt can't.
evaluation
Something you can measure
You can't measure "be more patient." You can measure whether responses preserved progression against a versioned arc definition — test it, regression-check it, prove to yourself it's working. Building measurable definitions instead of trusting instructions is how the Airbnb help-center model was built — and part of why it's still running.
contexts
Anywhere an AI system holds conversations that matter to the person in them:
support
The frustrated customer who needs the conversation held for one more turn before the refund policy arrives.
coaching & mentorship products
Doubt, stuck-ness, and ambition travel along arcs; tools that force conclusions lose the client's actual movement.
education
Confusion → curiosity → understanding is a pathway; answering too fast collapses it into copying.
companion & wellbeing products
The highest-stakes case for not simulating empathy — and the clearest case for respecting how a hard conversation actually progresses.
access
Two ways in.
the api · teams, platforms, products
The 264 arcs, structured and versioned
Recognize, guide, evaluate — integrated into your product or pipeline. In development; integration conversations are open now. Licensing enquiries welcome.
the companion · individuals
The same map, made personal
A web app for following your own conversations along the arcs. In development.
the first 13 arcs are open now → iamdeiadora.com · the research behind them → thequantumceo.com
why it matters
Every day, AI systems sit inside millions of conversations that matter to the people having them. Give those systems a map of how such conversations progress, and they can keep the movement alive rather than end it early. The point of this work: interactions that get healthier — not flatter — as AI scales.
Start the conversation.
For API access, licensing, research collaboration, or a closer look at the arc documentation — write directly. Peer-to-peer, from the first message.