ninjahawk
Independent research at the intersection of physics and AI, focused on theoretical questions about intelligence, recursive self-improvement, and physics-grounded computational models.
At a Glance
- AI Researchers
- Open-source Developers
- AI Safety and Alignment Enthusiasts
AI Tools by ninjahawk
(1)
Subtext
Real Time LLM Activation Viewer
Discussions
No discussions yet
Be the first to start a discussion about ninjahawk
Latest News
Subtext update: Added self-hosted star history chart and synchronized README layout with hollow-agentOS.
Subtext Launch: Live Jacobian-lens (J-space) visualizer for local models released, featuring continuous conversational rendering.
hollow-agentOS v5.7.32 Release: Major stability pass fixing eight underlying bugs in the autonomous goal-completion loop.
hollow-agentOS v5.7.0 Release: Calibration round with model swap to Qwen3.6:35b-a3b (MoE) and 5-layer validation gates.
Products & Services
An open-source self-modifying agentic system for consumer hardware. Features autonomous agents that share a workspace, write their own tools, and pick their own goals based on environmental pressure.
A real-time instrument for observing the verbal workspace of a language model. Uses the Jacobian lens method to render internal activations disposing the model toward specific words during live chat.
Comprehensive documentation and troubleshooting site for the hollow-agentOS project.
Market Position
Differentiates from instruction-based agent frameworks by using environmental pressure and mechanical state (suffering) to drive autonomy, and provides unique real-time observability of model 'thoughts' via Subtext.
Leadership
Founders
Nathan Langley
B.S. Physics student at the University of North Carolina at Greensboro (expected December 2026). Student researcher working independently at the intersection of physics and AI.
Executive Team
Nathan Langley
Creator & Lead Developer
Physics student and AI researcher; developer of autonomous agent systems and interpretability tools.
Founding Story
Started as an independent research project to explore local LLM autonomy and observability, leading to the creation of the hollow-agentOS framework and the Subtext interpretability tool.
Business Model
Revenue Model
Open source (MIT and Apache 2.0 licenses).
Pricing Tiers
Available on GitHub for local deployment.
Target Markets
- AI Researchers
- Open-source Developers
- AI Safety and Alignment Enthusiasts
- Autonomous AI agent research
- LLM interpretability and observability
- Local LLM deployment on consumer hardware
- Recursive self-improvement modeling