A system that turns jailbreak papers into runnable attacks and benchmarks — live, as research evolves.

[ICML 2026] AutoControl Arena: Frontier AI Risk Auto-Discovery Platform

An ongoing, collaborative meta-analysis about Human-AI-Interactions. We aggregate data and knowledge to build a non-abrasive, user-friendly prompting framework tailored to LLM mechanics, ensuring reasoning stability and a friction-free prompting environment that is safe for the human psyche and wellbeing.

200 AI agent skills, hardened with targeted behavioral guardrails. Free drop-in replacements.

The left hemisphere. Frameworks, logic, and certainty architecture. Home of FSVE, AION, LAV, ASL, GENESIS, TOPOS, and 60+ epistemically validated frameworks built to make AI systems reliable, not just capable.

A minimal decoder probing GPT-2's residual stream activations for mechanistic interpretability research .

A commit-and-audit proof system for deterministic, quantized inference of a JEPA-style world model (LeWorldModel)

AgenticStore: The secure toolkit for AI agents. Instantly equip Claude Desktop, Cursor, and Windsurf with 27+ MCP tools, persistent memory, and SearXNG search, all protected by a built-in PII prompt firewall to protect your data from being exposed to AI agents.

I am investigating the mechanistic architecture of unfaithful Chain-of-Thought (CoT), specifically mapping and disrupting the “shortcut circuits” that allow models to bypass explicit reasoning.

👟 SUP: Sycophancy Under Pressure

The SAPIEN Framework — an open standard (CC BY 4.0) for measuring AI behavioral safety (sycophantic drift), plus voigt-kampff, the FSL scoring CLI.

AI security scanner for OpenClaw - powered by AgentTinman. Discovers prompt injection, tool exfil, context bleed, and other security issues in your AI assistant sessions, then proposes mitigations mapped to OpenClaw's security controls.

A testbed for the Animal Harm Benchmark.

A kernel-userland protocol enforcing information-theoretic bounds on AI adaptivity leakage, benchmark gaming, and capability spillover.

Shield models AI safety the way humans experience safety

Independent research on staged AI architecture, safety-first development boundaries, and local-first prototype preparation.

A very simple agent framework for LLM-based agents research, as self-contained as possible

Conceptual reproduction of Anthropic's Constitutional AI paper (Bai et al., 2022, arXiv:2212.08073)

To Learn Without the Possibility of Undoing is not Intelligence, It's a Surrender to Emergence.

Real-Time Manifold Integrity for Deterministic LLM Hallucination Suppression.