Raft
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications

In RAG, Classic BM25 Found the Answer in 72% of Queries; Neural Search Found It in 49.7%
An engineering breakdown for RAG: on EntityQuestions, BM25 keyword search found the answer 72% of the time, neural search 49.7%. Where BM25 breaks and how to blend it with embeddings.

Codex agent built a personal knowledge base in Obsidian from 4 Habr articles instead of RAG
Codex in Obsidian turned four Habr articles into a linked knowledge base: cards, concept pages, and duplicate checks. Karpathy's idea works for personal research but gets expensive at 50–100 sources.

OmniVoice Voices Five Minutes of Russian Text in 44 Seconds, But Gets the Stress Wrong
The free OmniVoice model reads Russian aloud: five minutes of audio in 44 seconds, but with clipped phrases and wrong stress. Here is what Raft's test showed.

GLiNER Guard Replaces the LLM Filter Stack with One Model: 54 Checks per Second vs. 1.3
A single 307M-parameter model replaces the chatbot filter stack, checking safety and detecting personal data in one pass at 54 requests per second, versus 1.3 for WildGuard.

Tencent's 1B-parameter HunyuanOCR reads 70% of scans without edits, but confuses tables
A Data Scientist tested Tencent's compact OCR model on Russian documents: about 70% of pages come out usable without manual fixes, but it mixes up table cells and barely reads handwriting.

China's AI Security Market: A Sovereign Stack, Party Standards, and the Same Threats as Everywhere Else
Raft dissected China's AI security market: opaque from the outside but mature inside, with guardrails, agent protection, and Party-defined standards for "safe AI."

Raft's AI4BI: An AI Analyst for Coaches and Scouts in Soccer, Hockey and Basketball
Raft explains how AI4BI, its sports analytics platform, grew from a 2017 cricket MVP into a product for clubs across five sports and for fitness organizations.