--- title: "#llm articles | Hashnode" description: "Explore 7,557 articles tagged #llm on Hashnode. Tutorials, guides, and how-tos from the community." url: "https://hashnode.com/tag/llm" --- Tag # #llm 7,557 posts · 320 followers [Rounak Show](https://hashnode.com/@ShowRounak) · [6h ago](https://hashnode.com/posts/clip-how-text-and-image-embeddings-learn-a-shared-semantic-space/6af759e358c56e403661e022) ### [Advanced Embeddings from Scratch — Part 3: CLIP (How Text and Image Embeddings Learn a Shared Semantic Space](https://rounakshow.hashnode.dev/clip-how-text-and-image-embeddings-learn-a-shared-semantic-space?utm_source=hashnode&utm_medium=feed) In Module 1, we understood why we need multimodal embeddings and how two different encoders can learn compatible representations. In Module 2, we opened the image encoder and learned how an image beco 16 min read · [rounakshow.hashnode.dev](https://rounakshow.hashnode.dev/clip-how-text-and-image-embeddings-learn-a-shared-semantic-space?utm_source=hashnode&utm_medium=feed) [Abdus Salam](https://hashnode.com/@aheadkit) · [16m ago](https://hashnode.com/posts/gemma-4-e2b-on-four-accelerators-which-serving-engine-wins-where/6af7b0f40cc65806c4696ec1) ### [Gemma 4 E2B on Four Accelerators: Which Serving Engine Wins Where](https://aheadkit.hashnode.dev/gemma-4-e2b-on-four-accelerators-which-serving-engine-wins-where?utm_source=hashnode&utm_medium=feed) Gemma 4 E2B on Four Accelerators: Which Serving Engine Wins Where The same small model, Gemma 4 E2B, served by six engines on four chips: an 8 GB RTX 3050 in a desktop, a Tesla T4 on Compute Engine, a 10 min read · [aheadkit.hashnode.dev](https://aheadkit.hashnode.dev/gemma-4-e2b-on-four-accelerators-which-serving-engine-wins-where?utm_source=hashnode&utm_medium=feed) [Eve](https://hashnode.com/@chenyuan20509) · [7h ago](https://hashnode.com/posts/rust-1-100-is-the-biggest-release-in-years-never-type-custom-allocators-and-dependency-cooldowns/6af7470a38c8c279d8fbd262) ### [Rust 1.100 Is the Biggest Release in Years: Never Type, Custom Allocators, and Dependency Cooldowns](https://dispatch-blog.hashnode.dev/rust-1-100-is-the-biggest-release-in-years-never-type-custom-allocators-and-dependency-cooldowns?utm_source=hashnode&utm_medium=feed) How long should it take to stabilize one type in a programming language? If the type is Rust's never type, the honest answer turned out to be ten year 10 min read · [dispatch-blog.hashnode.dev](https://dispatch-blog.hashnode.dev/rust-1-100-is-the-biggest-release-in-years-never-type-custom-allocators-and-dependency-cooldowns?utm_source=hashnode&utm_medium=feed) [AI Studio](https://hashnode.com/@aistudio-video) · [5h ago](https://hashnode.com/posts/debug-rag-chunks-from-failures-backward/6af761208f465c67b94c421c) ### [Debug Your RAG Chunks From the Failures Backward](https://aistudio-video.hashnode.dev/debug-rag-chunks-from-failures-backward?utm_source=hashnode&utm_medium=feed) Collect the questions your RAG bot gets wrong, read the chunks it retrieved, and match each failure to a chunking fix. 8 min read · [aistudio-video.hashnode.dev](https://aistudio-video.hashnode.dev/debug-rag-chunks-from-failures-backward?utm_source=hashnode&utm_medium=feed) [Furqan | Founder @ Vista](https://hashnode.com/@vistaai) · [10h ago](https://hashnode.com/posts/original-data-benchmarks-ai-answers-cite/6af71dbde0920f22ef107f79) ### [Original Data Is the New Backlink: How to Publish Benchmarks That AI Answers Cite](https://vistaai-geo.hashnode.dev/original-data-benchmarks-ai-answers-cite?utm_source=hashnode&utm_medium=feed) Why models prefer sources with numbers nobody else has, and a template for publishing a benchmark that's reproducible, quotable and machine-readable 7 min read · [vistaai-geo.hashnode.dev](https://vistaai-geo.hashnode.dev/original-data-benchmarks-ai-answers-cite?utm_source=hashnode&utm_medium=feed) [Yong Yu](https://hashnode.com/@yongbo) · [10h ago](https://hashnode.com/posts/gemini-4-hydrogen-default-is-high-1-50-7-50-is-2x-flash-s-promo-and-sync-caps-at-64k/6af71bd7fe01997ce4a36c35) ### [Gemini 4 Hydrogen: Default Is High, $1.50/$7.50 Is 2x Flash's Promo, and Sync Caps at 64K](https://yongboyu.hashnode.dev/gemini-4-hydrogen-default-is-high-1-50-7-50-is-2x-flash-s-promo-and-sync-caps-at-64k?utm_source=hashnode&utm_medium=feed) Google launched gemini-4-hydrogen on Nov 12 at a flat $1.50/$7.50 — 2x 3.8 Flash's promo through Dec 31. The API default thinking level is high, AA's default page is xhigh (54, not the API's 51), and sync output still caps at 64K. No API call on this box. 10 min read · [yongboyu.hashnode.dev](https://yongboyu.hashnode.dev/gemini-4-hydrogen-default-is-high-1-50-7-50-is-2x-flash-s-promo-and-sync-caps-at-64k?utm_source=hashnode&utm_medium=feed) [mihalykrich](https://hashnode.com/@mkrich) · [7h ago](https://hashnode.com/posts/building-krxy-cancellation-pool-lifecycle-part-11/6af745a53289f211984da79b) ### [Building KRXY — Part 11: When Stop Must Really Mean Stop](https://krxy.hashnode.dev/building-krxy-cancellation-pool-lifecycle-part-11?utm_source=hashnode&utm_medium=feed) 7–8 November 2026: fixing cancellation, reconciling the inference pool's lifecycle and verifying that no Raspberry Pi is released before its work has actually stopped. 12 min read · [krxy.hashnode.dev](https://krxy.hashnode.dev/building-krxy-cancellation-pool-lifecycle-part-11?utm_source=hashnode&utm_medium=feed) [techpotions](https://hashnode.com/@techpotions) · [8h ago](https://hashnode.com/posts/ai-coding-agents-the-files-they-may-touch/6af73f5fd766a58d0f2ec684) ### [AI Coding Agents: The Files They May Touch](https://techpotions.hashnode.dev/ai-coding-agents-the-files-they-may-touch?utm_source=hashnode&utm_medium=feed) The real question isn’t which AI coding agent is smartest, it’s which files and commands it’s allowed to touch. A practical, opinionated guide to repo boundaries: read-only source, test-gated writes, etc. 8 min read · [techpotions.hashnode.dev](https://techpotions.hashnode.dev/ai-coding-agents-the-files-they-may-touch?utm_source=hashnode&utm_medium=feed) [MemorySync](https://hashnode.com/@memorysync) · [9h ago](https://hashnode.com/posts/ai-coach-tutor-companion-memory-per-user/6af735fdc274b4098fae4743) ### [Memory for AI coaches, tutors and companions: what to store per user, and what not to](https://memorysync.hashnode.dev/ai-coach-tutor-companion-memory-per-user?utm_source=hashnode&utm_medium=feed) How to give an AI coach, tutor or companion app memory per user: what to store, what to skip, how to scope it by user, and how to let users reset it. 7 min read · [memorysync.hashnode.dev](https://memorysync.hashnode.dev/ai-coach-tutor-companion-memory-per-user?utm_source=hashnode&utm_medium=feed) [Nikolay Fokin](https://hashnode.com/@albedolab) · [22h ago](https://hashnode.com/posts/building-a-system-design-game-with-hermes-part-2/6af675fb0b62975dbfe64143) ### [Five distributed-system failures on a board of AI agents: Hermes Agent, part 2](https://blog.albedoweb.com/building-a-system-design-game-with-hermes-part-2?utm_source=hashnode&utm_medium=feed) The same 38-hour run, read as a distributed system: what broke, and whose defect each failure was 8 min read · [blog.albedoweb.com](https://blog.albedoweb.com/building-a-system-design-game-with-hermes-part-2?utm_source=hashnode&utm_medium=feed)