https://runelab.ai/en/blog/ RuneLab Blog 2026-10-15T00:00:00+00:00 https://runelab.ai/en/blog/post/model-distillation How a big model teaches a small one 2026-10-15T00:00:00+00:00 2026-10-15T00:00:00+00:00 RuneLab Model distillation trains a small model on a big one’s answers. Live figures show what the teacher knows besides its answer and let you train a tiny student. https://runelab.ai/en/blog/post/quantization How a big model fits in fewer bits 2026-10-15T00:00:00+00:00 2026-10-15T00:00:00+00:00 RuneLab Quantization stores each number of a model in fewer bits. Live figures show rounding to a grid, why one outlier ruins it, and whether a model fits your memory. https://runelab.ai/en/blog/post/multi-agent-systems One agent or many? 2026-10-14T00:00:00+00:00 2026-10-14T00:00:00+00:00 RuneLab What an AI agent is, how a team of agents shares a job, and why a second agent helps on some tasks and hurts on others, with live figures and evidence. https://runelab.ai/en/blog/post/speculative-decoding How a small model speeds up a big one 2026-10-14T00:00:00+00:00 2026-10-14T00:00:00+00:00 RuneLab A small model guesses several words ahead and the large one checks them at once. Live figures show why the text stays the same and when it is faster. https://runelab.ai/en/blog/post/ask-your-database-in-turkish Ask your database in Turkish 2026-10-13T00:00:00+00:00 2026-10-13T00:00:00+00:00 RuneLab How a text-to-SQL model turns a Turkish question into a query, why the database computes the numbers, and what an honest test of Hati Glint SQL shows. https://runelab.ai/en/blog/post/finding-the-right-clause Finding the right clause 2026-10-13T00:00:00+00:00 2026-10-13T00:00:00+00:00 RuneLab How Kenning, a Turkish legal embedding model, finds the clause you meant, not the one with your words: a live map and the model card’s scores, gaps included. https://runelab.ai/en/blog/post/how-rag-finds-answers How RAG finds answers 2026-10-10T00:00:00+00:00 2026-10-10T00:00:00+00:00 RuneLab Retrieval-augmented generation with live figures: how documents become searchable pieces, how a question finds the closest ones, how the answer cites them.