Project Overview
Experimental AI Research, Documented End-to-End
This project investigates how training order, dataset structure, and controlled novelty influence learning dynamics and downstream behavior in language models. All experiments are versioned, reproducible, and publicly documented.
Curriculum-first designExplicit control conditionsFailure cases published
What makes this different
- Curriculum-first experiment design
- Explicit control conditions
- Dataset provenance and versioning
- Behavioral analysis beyond loss curves
- Failure cases published, not hidden
What you can explore
- Research milestones and experiment history
- Dataset composition and evolution
- Training run metrics and comparisons
- Observed behavioral shifts across phases
Explicit non-goals
- Product demos
- Chatbot showcases
- Claims of sentience
- AGI hype
Active Milestone
Milestone queue
Milestones will populate from the research filesystem once ingested.
Open milestone index →Latest Experiment
Gutenberg Curriculum Baseline
Training on curriculum-aligned Gutenberg narrative text will reduce loss steadily without early collapse into repetition, and will improve short prompt continuation coherence relative to a minimal smoke dataset.
View experiment →Experiment Snapshots
View all →Gutenberg Curriculum Baseline
Training on curriculum-aligned Gutenberg narrative text will reduce loss steadily without early collapse into repetition, and will improve short prompt continuation coherence relative to a minimal smoke dataset.
Status: draft
Curriculum Ordering Baseline (Dataset Plan Locked)
A curriculum-ordered training schedule will alter training stability and/or downstream behavior relative to a shuffled control, even when total token count and model architecture are held constant.
Status: draft
EXP-0004 — Guided Choice / Preference Signal (Adaptive Sampling)
If we allow the organism to bias its dataset sampling using a simple adaptive rule, then training and behavior will diverge measurably from fixed mixing weights, because the organism’s learning dynamics will express a stable preference sig…
Status: draft
EXP-0003 — “Play” via Controlled Novelty (Small Perturbations)
If we introduce a small, controlled novelty pressure (e.g., lightly perturbed text during sampling, or prompt mutation during evaluation), then the organism will become more robust to minor corruption without destabilizing training, becaus…
Status: draft