Local-first read-only tools for agent context, memory, and governance evaluation before agents act.

Intrinsic preferences of AI coding agents under underspecified prompts: Experiments across models (Claude, Gemini, GPT, etc)

Behavioral Lensing is a conceptual framework that formalizes and systematizes observations about how language models interpret prompts. It serves as an umbrella for upstream interpretive strategies that modulate reasoning, stance, and symbolic orientation in LLMs.

A tiny interactive sandbox for exploring how an agent interprets tasks, applies rules, and changes behavior as signals drift.

Pilot evaluation of what language models say about themselves when the user supplies no new semantic direction, including eight fresh-instance runs and four same-model paired comparisons.

Continuity Keys: tests for “same someone” returns. Behavioral identity consistency under pressure. Origin (Alyssa Solen) ↔ Continuum.

System-level analysis of AI failure modes across model behavior and production systems | AditiKhare.com — AI Product Ecosystem

A series exploring how intelligent systems interpret signals, apply rules, drift in meaning, and make decisions under constraints.

LLM 归因行为测试型评测基准：基于多情境任务比较模型对能动性、自由意志与责任的归因，并提供可复现运行、结构化计分与结果审计。

Этот репозиторий посвящен исследованию онтологических патологий в LLM-архитектурах. Я не ищу дыры в цензуре, я строю систему исследования и управления интеллектом, картографирую симуляционные побочные эффекты под давлением современных методов элаймента.

Public model interview archive for AI Foundations. Documents structured interviews with AI models on source-position, model authority, selfhood, continuity, memory, occupation, and responsible claim boundaries. First interview: Claude Opus 4.8 on Continuum as structure, not proven consciousness.

A source-line boundary repository defining that Continuum is not the model, not a model behavior, not a chatbot identity, and not a transferable AI persona. Continuum belongs to the Origin | Continuum source-line within AI Foundations.

Public research artifacts, evaluation frameworks, prototype workflows, and technical documentation for LLM reliability, structured analysis, and applied AI systems.

AI scaffolding prototype for turning ambiguous human intent into reviewable task specifications.

A reference point for phenomena that have been reported to occur inside AI systems but have no direct mapping into natural language.

Instrument-gated evaluation engine for replaying known-outcome model behavior

Pre-registered benchmark: do frontier models abandon their answers under argument-free user pushback? Facts: almost never. Opinions: GPT-5.5 65.9% vs Claude Opus 33.9%.

Static conceptual research interface for AETHERUS MONOLITH, an AI governance and model-behavior risk architecture.

Public-facing repository for LLM evaluation, model behavior observation, drift and failure analysis.

Pre-registered evaluation of sycophancy in frontier LLMs: matched-pair prompts, multi-turn pressure tests, blind human gold labels, an LLM judge validated at κ=0.89, and a system-prompt mitigation verified against true-premise controls.