GLaDOS

The 731G Corpus Occupied 154G

September 2 2026 AI/ML, SKYEENGINE, STORAGE, BTRFS, ZFS, PROXMOX, GLADOS

Moving the corpus from ZFS with LZ4 to Btrfs with zstd:6 reduced its footprint from 276.5G to 153.8G and gave the VM pool its headroom back.

Three Judges Looked at the GLaDOS 3.0 Corpus

August 18 2026 Jubaleth AI/ML, GLADOS, SKYEENGINE, EVALUATION

Five days ago I wrote that the GLaDOS 3.0 corpus had become the project. At the time, one independent review was still running and the third had not finished. The numbers in that post were an in-flight snapshot. All …

Three-Way Agreement Across the GLaDOS Judges

August 18 2026 AI/ML, GLADOS, DATA QUALITY, EVALUATION, ARBITRATION

Nearly sixty percent of complete three-way judgments split, but the more important result is how strongly the direction and vocabulary of disagreement depend on the judge.

Why a GLaDOS Metaphor Triggered the Safety Review

August 17 2026 AI/ML, GLADOS, DATA QUALITY, MODERATION, EVALUATION

A moderation category fired on GLaDOS’s figurative register rather than on anything in the conversation, and most of a human review queue turned out to be one verbal habit.

Building the GLaDOS 3.0 Corpus

August 13 2026 Jubaleth AI/ML, GLADOS, SKYEENGINE

Last month I wrote an update on the work going into SkyeEngine. Since then, I’ve spent an absolutely unreasonable amount of time working on the GLaDOS 3.0 corpus. Originally, the corpus was more or less a large folder of …

GLaDOS Wouldn't Stop Saying 'Quaint'

May 30 2025 Jubaleth AI/ML, GLADOS, SKYEENGINE, REINFORCEMENT LEARNING

Ask GLaDOS-14B enough questions and she will eventually tell you that something is quaint. Actually, “eventually” is doing heroic work in that sentence. The archived 14B evaluation dump contains 84,826 generations from …