Back openDesk Edu for a sovereign, open-source education — every vote counts.
Vote nowExploring knowledge graphs, graph theory, graph visualization, and their intersection with artificial intelligence.
We benchmark pi-l1-cache, a zero-dependency L1 in-memory cache extension for the pi coding agent. 138µs per request, honest about where the 100x hash claim breaks.
Practical results from serving DeepSeek's latest MoE mixture-of-experts model on dual DGX Spark systems with vLLM, DSpark speculative decoding, and Anemll optimizations.
Deep analysis of arXiv 2608.13759: Rust Offload brings CUDA/HIP-level GPU programming to Rust without vendor lock-in, using LLVM Offload for cross-vendor portability and Rust's type system for memory safety in kernels.
GRALAN enables KGs to speak directly in the LLM's semantic space through relational tokens that preserve graph structure. No fine-tuning required — just trainable language mediation.
X-KGRank unifies structural collaborative filtering with LLM-based explanation. A 1.5B parameter model matches a 7B model on explanation quality — but fabricates facts more often when not grounded.
Temporal KGs are a clear growth cell: time-aware embeddings and event-driven updates as frontiers. Models, trade-offs, production patterns.