[feder-cr]

Ten new wiki pages, chosen from measured US search volume (DataForSEO, 2026-10-12) plus the new Microsoft-Decision-1 category:

page	head term (searches/month)
open-source-alternative-to-microsoft-decision-1	new term, 0 so far (released 2026-10-09)
what-is-a-decision-model	what is a decision model (50)
ai-decision-making-with-probabilities	ai decision making (590)
decision-engine-vs-decision-model	decision engine (260)
openvino-for-llm-inference-on-a-cpu	openvino (2,900)
int8-weight-quantization-explained	int8 quantization (90)
multi-class-text-classification-with-a-choice-question	text classification (260)
reranking-with-yes-no-relevance-questions	reranker (320)
snake-ai-local-model-vs-cloud-api	snake ai (880)
download-a-gguf-model-from-hugging-face	hugging face gguf (40)

• Every page passes .wave/gate.py; the corpus check finds no page without an inbound prose link; build_wiki.py builds 152 pages.
• jevos facts come only from the README and the release; nothing on training, data, base model or architecture. Microsoft-Decision-1 facts come only from Microsoft's own pages, attributed and dated; no speed or accuracy comparison with it is claimed, because none was measured.
• The Snake page states its setup (GGUF on a laptop GPU via llama.cpp vs Jev over the internet), that the code words each move, and that it measures latency, not accuracy.
• Fixes: open-source-alternative-to-jev no longer describes how choice and score work inside, and has the jevos-v4 score figures; the Measurements hub notes the Snake race ran on a GPU.

🤖 Generated with Claude Code