HOME-Overview
◇ About
Investigating LLM creativity through controlled experiments. Small, falsifiable tests exploring how prompts and constraints shape creative outputs. From absorption points to seeding strategies, each experiment isolates a single variable.
Articles 7
In-depth analysis and polished research
Scratchpads 16
Raw experiments and hypotheses
Projects 3
Builds and prototypes
◇ Recent Activity
GEPA Prompt Optimization on Tiny Local Models 2026-05-10 CLIR: A Fixed-Vocabulary Intermediate Representation for LLM Code Generation 2026-05-07 Replicant: Giving AI Agents GPU Access Through Google Colab 2026-04-08 Output Diversity vs Generation Parameters (gpt-oss-120b) 2026-03-05 Passphrase Seeding — Random Noise (gpt-oss-120b) 2026-03-05 Passphrase Seeding — Themed Words (gpt-oss-120b) 2026-03-05 How Well Do LLMs Follow Length Instructions? 2026-02-20 Character Count Range Compliance (gpt-oss-120b) 2026-02-10 Compound Length Constraints — Words + Characters (gpt-oss-120b) 2026-02-10 Word Count Range Compliance (gpt-oss-120b) 2026-02-10