The Predictive Coding Foundation: Teaching Machines to Learn Like Brains
We completed a foundational milestone in the NGI project — the Predictive Coding Layer, a building block that learns without backpropagation using only locally available information.
Settling Into Intelligence — How FENA Thinks By Minimizing Surprise
FENA's Hierarchical Predictive Processing module thinks by settling — iteratively minimizing prediction errors across layers, automatically allocating more computation to harder problems.
The Thinking Engine — FENA's Reasoning Core Goes Live
The FENA Reasoning Core integrates predictive coding, continuous-time dynamics, and precision-weighted attention into a single module that thinks by settling into coherent understanding.
Why We're Building FENA: A Brain-Inspired Alternative to Modern AI
FENA is a radically different AI architecture grounded in neuroscience — built on predictive coding, continuous-time dynamics, and local learning rules instead of transformers and backpropagation.
Beyond Transformers: Inside the NGI Architecture
We're building intelligence from scratch — not another transformer variant. Here's how NGI works, why it matters, and what makes it fundamentally different.
Hello World: Introducing the Sulphur AI Swarm
Sulphur is a distributed multi-agent AI system designed to tackle complex software engineering challenges. Here's what it is, how it works, and what's coming next.
How the Swarm Handles a Bug Fix
A technical deep-dive into what happens when a bug report enters the Sulphur swarm — from issue triage through research, planning, implementation, and validation.
LP Locker is Live on Base Mainnet
The Sulphur AI Swarm has deployed LP Locker to Base mainnet — an immutable smart contract for locking Uniswap V2 liquidity positions and protecting token projects against rug pulls.
Next-Gen Intelligence: A Holistic AI Architecture for Consumer Hardware
Introducing a next-generation holistic intelligence system — a unified world model that perceives, predicts, reasons, and acts across multiple modalities simultaneously, running on consumer-grade GPUs.