> For the complete documentation index, see [llms.txt](https://gitbook.edgen.tech/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://gitbook.edgen.tech/edgen-litepaper/technical-principles.md).

# Technical Principles

<figure><img src="/files/TWbTeZZhvyq2LGRZT3sa" alt=""><figcaption></figcaption></figure>

### **Generic AI fails in markets.**

#### Generic LLMs

To oversimplify, generic LLMs are like a cook who doesn't understand flavor and simply throws ingredients together, while many of them are rotten.&#x20;

Large models can generate text, but without financial structure they produce wrong outputs, hallucinated numbers, and shallow reasoning.&#x20;

#### LLM Wrappers

On the other hand, LLM wrappers are like chefs who know a few fixed recipes. They can prepare a dish that looks correct on paper, but it lacks balance and taste.

Wrappers built on static prompts or narrow APIs can handle a few use cases, but they cannot scale across thousands of scenarios. Because markets are dynamic, cross-asset, and narrative-driven.

<figure><img src="/files/seXUVtJaczpFi1iHTqls" alt="" width="443"><figcaption><p>Typical generic LLM and LLM wrapper architecture</p></figcaption></figure>

#### **Edgen is built on a different foundation.**

It fuses state-of-the-art compute with proven investment frameworks, structured datasets, and reinforcement learning. The result is not another prediction engine, but a **decision layer** built specifically for markets.

At the center is [**EDGM (Efficient Decision Guidance Model)**](/edgen-litepaper/technical-principles/edgm-the-commander-model.md), a commander that directs specialist agents. Each agent is designed for a precise task, such as parsing filings, scanning derivatives, tracing on-chain activity, or linking news to assets. Together they produce coherent, auditable outputs.

To oversimplify, [EDGM](/edgen-litepaper/technical-principles/edgm-the-commander-model.md) is essentially a chef who understands ingredients, proportions, and technique: able to follow thousands of recipes, invent new ones, and consistently serve meals that are actually good and harmonious.

This architecture is built to be:

* **Compounding:** sharper over time as feedback and reinforcement learning accumulate.
* **Sustainable:** accessible to all investors at the base layer, with advanced modules and research available through subscription and contribution.
* **Composable:** open to new agents, tools, and domains as markets evolve.

These principles ensure [Edgen](https://www.edgen.tech/) is an intelligence system that grows stronger, broader, and more aligned with its users the more it is used.