A library of Agentic Skills for Claude Code and LLM agents, codified from social science standards and methodological texts for doing experimental and other kinds of work.

Agent Teams Eval: comparing Claude Code Agent Teams vs single-agent for feature implementation on LangGraph. Ceiling effect, 3.6x speedup, zero peer communication.

Agent Teams Eval: comparing Claude Code Agent Teams vs subagents for bug-fixing on Ruff. Ceiling effect — 8/8 solve rate, zero peer communication.

Exploratory compact LM research branch derived from LUMI-Arch, focused on structure-sensitive generalization under tight model constraints.

An independent study on AI interiority, documenting two experiments that probe the line between knowledge and consciousness in large language models.

Agent Teams Eval: comparing Claude Code Agent Teams vs subagents for architecture design. First significant result — Agent Teams advantage d=+0.99, p=0.014.

EIDOLON est un moteur d'inférence probabiliste qui modélise une cible comme un graphe de croyances sous contrainte bayésienne. Chaque nœud encode un trait cognitif avec prior, sensibilité et incertitude. Les arêtes typées portent une conductivité propre. Le moteur propage, calcule, et enregistre ce qui cède sous pression dirigée.

Split a single neural network into multiple smaller networks using weight splitting.

Open-source platform for designing and running online behavioural experiments