---
title: README
emoji: ⚡
colorFrom: indigo
colorTo: yellow
sdk: static
pinned: false
---

# 🔺 Project Prism

**Project Prism is an experimental AI architecture research project by SmilyAI Labs.**

We are investigating a simple question:

> **Does one neural network really need to learn knowledge, computation, and working memory in the same way?**

Our current research model, **Orion Flagship**, explores an architecture we call **Transformer 2 (T2)**.

## To Join

**Open** a new discussion in the community tab of this card and tell us why you would like to join! 


## Transformer 2

Traditional language models largely rely on a homogeneous stack of Transformer layers to learn many different functions at once.

T2 explores separating several of these roles:

**KNOW** — sparse learned memory for persistent knowledge  
**DO** — reusable neural procedures for computation and transformation  
**THINK** — persistent working state for intermediate reasoning  
**VERIFY** — additional computation for checking and refining results

The goal is to investigate whether this separation can improve parameter and compute efficiency while retaining strong general capabilities across language, code, mathematics, and reasoning.

## 🌌 Orion Flagship

**Orion Flagship** is the primary experimental model being developed under Project Prism.

The planned flagship architecture targets approximately:

- **7B total parameters**
- **~1B active parameters**
- Sparse Knowledge Vault
- Shared Procedure Banks
- Persistent Working State
- Adaptive recurrent deliberation
- General-purpose language, code, and mathematical reasoning

Before scaling to the full model, we are testing the architecture with smaller prototypes and controlled ablations.

## 🧪 Current research

Our current Mini-T2 experiments are investigating:

- whether learned memory develops useful knowledge retrieval;
- whether Procedure Banks develop reusable computational specialisation;
- whether Working State improves multi-step reasoning;
- whether recurrent deliberation improves difficult problems;
- and whether these components provide measurable advantages over compute-matched Transformer baselines.

**Results are experimental. Claims about architectural improvements will be based on controlled evaluations and ablations rather than training loss alone.**

---

### Project Prism

**KNOW · DO · THINK · VERIFY**

*Exploring what comes after the conventional Transformer.*