العرش

العرش
Al-Arsh

Root-based system design

There are two ways to instruct a language model. Both work. They produce very different results.

Al-Kursi is the structured way. You write rules, steps, templates. You tell the model exactly what to do and in what order. The output is predictable. It works until the task falls outside the rules.

Al-Arsh is the root way. You give one principle — a seed — and let the model grow the response from it. No steps. No templates. The output adapts to the context because the principle is deeper than any rule.

The names come from Ayat al-Kursi (2:255). Al-Kursi is the footstool — where knowledge rests in fixed form. Al-Arsh is the throne — what carries everything without being bound by it. One contains. The other supports.

Below you can test both approaches side by side. Same task. Same model. Different root.

الكُرسِي ↔ العَرْش

Enter a task and see both approaches respond

الكُرسِي

Al-Kursi — Structure

Lead generation agent
SYSTEM: You are a lead gen agent. Follow these steps: 1. Collect name and email 2. Send template A 3. If no response: send template B 4. Escalate after 7 days VARIABLES: {company}, {contact}
Fixed output — works until the recipe breaks
Customer support
SYSTEM: Support agent template v3.2 RULES: - Always start with "Apologies for the inconvenience" - Search FAQ: [list of 20 questions] - Escalate if not in FAQ - CLOSING: template C No deviations allowed.
Predictable, fragile at edges
Research agent
PROMPT: "Find 10 sources on X. Format: title - author - year - summary. Sort by relevance. Use only .edu and .gov domains."
Bound by rules, no insight

العَرْش

Al-Arsh — Root

Lead generation agent
PRINCIPLE: Every lead is a relationship that starts with recognition. ROOTS: - Timing is respect — respond within the invisible windows of attention - Personalisation is not a variable, it is the soil the relationship grows in - A lead that does not respond says something about the question, not the lead GROWTH: Let the agent decide which moment, which tone, which frequency fits the context as it unfolds.
Living relationship — adapts, grows with each touch
Customer support
PRINCIPLE: A customer who reaches out is not looking for an answer but for restoration of trust. ROOTS: - Acknowledgement comes before solution — without the first root there is no fruit - Knowledge grows from the encounter, not from a database - Escalation is not failure but recognition of boundary GROWTH: The agent grows toward the need, not toward the procedure.
Trust deepens — every interaction strengthens
Research agent
PRINCIPLE: Knowledge is not a collection but a landscape that unfolds. ROOTS: - The question matters more than the source — a sharp question finds deeper answers - Relevance is not objective but contextual — what is needed now, not what fits a format - Contradiction is not noise but the contour of truth GROWTH: Let the agent show connections that were not asked for but are needed.
Insight grows — discovers what was not sought
كيف يعمل

كيف يعمل

How it works

١

Plant the Root

Define the principle — the living seed from which the system grows. One root, infinite expressions.

٢

Let it Grow

Context feeds the root. The system adapts, finds its path, and strengthens with each interaction.

٣

Harvest

The outcome is not manufactured — it emerges. Natural, context-true, and stronger than any template.