AI.FuryBee is an English article library about artificial intelligence. Articles are grouped into eight categories: How It Works, Practical AI, Hardware, Training & Data, Models & Players, Benchmarks, Fundamentals, and Ethics & Safety.
Topics go from first principles to the details that matter in production: the KV cache, speculative decoding, LoRA and QLoRA, RLHF, mixture of experts, accelerator hardware, leaderboards such as Chatbot Arena, prompt caching, and the energy and water footprint of training and inference.
Key features
01
Eight categories
From Fundamentals for newcomers to Hardware and Benchmarks for people comparing models and accelerators.
02
Seven starter scripts
Embedding similarity, an LLM API playground, a LoRA fine-tuning starter, an Ollama quickstart, prompt templates, a RAG starter kit and a token counter.
03
Search in the browser
Pagefind indexes the site at build time, so search runs locally without a search server.
04
Built for reading
Each article has a table of contents, a reading-time estimate and links to the previous and next article.
05
RSS feed
Follow new articles at ai.furybee.org/rss.xml.
How it works
The site is generated ahead of time with Astro and served as static files by Nginx. Code samples are highlighted at build time. The FuryBee portal reads its RSS feed to list the latest articles on the Articles page.
How modern LLM servers keep GPUs busy: continuous batching, PagedAttention for KV cache memory, prefix caching and vLLM in practice, with launch commands and tuning advice.
What mechanistic interpretability tries to do, from features and superposition to circuits and sparse autoencoders, and what it can and cannot tell us about AI safety today.
Why vector search alone misses answers, how BM25 plus dense retrieval plus a cross-encoder reranker fixes it, and how to build and evaluate the pipeline with working code.
How state space models and Mamba process sequences with a fixed-size memory instead of attention, what they gain in long-context efficiency, and where transformers still win.
AI.FuryBee FAQ
Do I need a technical background to read AI.FuryBee?
No. The Fundamentals category starts from the basics, and articles explain terms as they introduce them. Deeper categories such as How It Works and Hardware are there when you want more.
How many articles are there?
The library had 77 articles across 8 categories in October 2026, plus 7 starter scripts. The Articles page on furybee.org shows the newest ones.
Can I follow new articles?
Yes, through the RSS feed at https://ai.furybee.org/rss.xml.
Is my search tracked?
No. Search runs in your browser with Pagefind, and the site uses no analytics.
