The site is designed to connect two kinds of evidence. The first is editorial: short AI news briefs, longform explainers, visible source trails, multilingual editions, and a consistent taxonomy of topics and vendors. The second is engineering: public projects, automation systems, dashboards, and local tools that show how the workflow is built.
TG-NEWS is the production news pipeline behind the public channel. It discovers sources, maps stories, prepares publication output, and gives mlllm.io the raw material for a better long-term archive. The website should make that archive readable for people, crawlable for search engines, and understandable for user-directed LLM agents.
My current focus is practical AI infrastructure: agent workflows with reviewable receipts, Telegram automation, open-source tools, and knowledge systems that preserve context instead of letting it disappear into chat history.
The site should make that work easier to evaluate: readers can follow the news feed, inspect explainers, read project notes, and reach the public repositories or channel from a stable author profile.