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 9
In-depth analysis and polished research
Scratchpads 27
Raw experiments and hypotheses
Projects 3
Builds and prototypes
◇ Recent Activity
Testing Needle, a Local Tool-Calling Model 2026-08-26 Abstention on No-Tool Queries (cactus-needle 2.0.5) 2026-08-17 Effect of AI-Style Phrasing Artifacts on Tool Calls (cactus-needle 2.0.5) 2026-08-17 Argument Accuracy vs Argument Complexity (cactus-needle 2.0.5) 2026-08-17 Tool-Selection Accuracy and Latency vs Catalog Size (cactus-needle 2.0.5) 2026-08-17 Confidence Calibration for Edge-Cloud Escalation (cactus-needle 2.0.5) 2026-08-17 Determinism and Paraphrase Sensitivity (cactus-needle 2.0.5) 2026-08-17 Robustness to Typos and STT-Style Noise (cactus-needle 2.0.5) 2026-08-17 Runtime Feasibility and API Surface (cactus-needle 2.0.5) 2026-08-17 Schema Violation Census Across All Series Records (cactus-needle 2.0.5) 2026-08-17