LLM News and Articles
| Saturday, 2025-12-20 | ||||
| 18:31 | Hack the thought process of LLM in realtime - Streering https://blog.gopenai.com/hack-the-thought-process-of-llm-in-realtime-streering-7597b8c6152b | |||
| 18:28 | Prompt Templates and Chains in LangChain: Writing Smarter LLM Workflows https://medium.com/@induwaragayashan/prompt-templates-and-chains-in-langchain-writing-smarter-llm-workflows-59fb41d524dd | |||
| 18:22 | The Engineering Reality of Fine-Tuning: Why “Low Loss” Often Means Brain Damage https://medium.com/@Manash_Pratim/the-engineering-reality-of-fine-tuning-why-low-loss-often-means-brain-damage-7a792152c85c | |||
| 18:16 | We Didn’t Teach Machines to Think (A Not-So-Brief Story of LLMs and Their Impact on Humans) https://medium.com/@jcostafernandes/we-didnt-teach-machines-to-think-a-not-so-brief-story-of-llms-and-their-impact-on-humans-ea3f5582b369 | |||
| 18:14 | Issue 67: The PyDMD Project, Build a Text-to-Image Generator, New Tutorials https://medium.com/@rami.krispin/issue-67-the-pydmd-project-build-a-text-to-image-generator-new-tutorials-486a99f08761 | |||
| 18:06 | Deploying an LLM Locally: A Practical Guide https://medium.com/@mark.southworth98/deploying-an-llm-locally-a-practical-guide-9372dec5fa7a | |||
| 18:00 | Why Machines Can’t Provide Psychological Support As Humans Do https://medium.com/@melnawawy1980/why-machines-cant-provide-psychological-support-as-humans-do-f94fbff792bf | |||
| 17:08 | Why LLMs Fail at Math — The Hidden Reason Behind AI’s Weakness https://medium.com/@coffee_and_notes/why-llms-fail-at-math-the-hidden-reason-behind-ais-weakness-227950d53c56 | |||
| 17:01 | Agentic EDA: Automating Exploratory Data Analysis for Data Science Workflow https://pub.towardsai.net/agentic-eda-automating-exploratory-data-analysis-for-data-science-workflow-b874dec24d7a | |||
| 16:37 | TOON for LLMs: A Comparative Performance Analysis against JSON https://medium.com/@iamsdt/toon-for-llms-a-comparative-performance-analysis-against-json-a3f8745658ec | |||
| 16:26 | Perplexity Lost Part of Our Conversation and Then Denied It https://github.com/Looking4OffSwitch/perplexity-conversation-issue | |||
| 16:21 | Revolusi Keamanan Siber EV: Menggunakan Large Language Models (LLM) untuk Deteksi Serangan Zero-Day https://medium.com/@yus.subakti/revolusi-keamanan-siber-ev-menggunakan-large-language-models-llm-untuk-deteksi-serangan-zero-day-190618c07ce8 | |||
| 16:17 | Gamma: Open-source LLMs from DeepMind https://medium.com/@dravenrh/gamma-open-source-llms-from-deepmind-6f1c3c0ae968 | |||
| 16:09 | LLMs Don’t “Think” — Here’s What They Actually Do https://medium.com/@ruban7r/llms-dont-think-here-s-what-they-actually-do-4b5e0ce5b702 | |||
| 16:02 | 7 LangChain Structured-Output Tricks That Never Break https://medium.com/@ThinkingLoop/7-langchain-structured-output-tricks-that-never-break-33fa7c687e2a | |||
| 16:02 | How to Use Gemini 3 Pro Efficiently https://pub.towardsai.net/how-to-use-gemini-3-pro-efficiently-5a39a33438fe | |||
| 16:02 | Can ChatGPT help with a midlife crisis? https://www.ft.com/content/8b6e0a41-f3d1-474d-9d69-d5e0b897907b | |||
| 15:32 | Seven RAG Routers That Slash Token Spend https://medium.com/@npavfan2facts/seven-rag-routers-that-slash-token-spend-e3f2f0e97abd | |||
| 15:20 | MemLayer: Plug-and-Play Memory for LLMs in Just 3 Lines of Code https://medium.com/@shouke.wei/memlayer-plug-and-play-memory-for-llms-in-just-3-lines-of-code-808bb0bf6af4 | |||
| 15:07 | Weekly News (December 15, 2025): Deepfake image of Bondi beach, Falsely accused students & Chrome… https://medium.com/law-and-ethics-in-tech/weekly-news-december-15-2025-deepfake-image-of-bondi-beach-falsely-accused-students-chrome-3032061ee96a | |||
| 14:57 | The Model That Thinks Without Words: A Data Science Deep Dive https://medium.com/mlworks/the-model-that-thinks-without-words-a-data-science-deep-dive-3ae810a09987 | |||
| 14:49 | The Next Frontier of Agentic Engineering: Introducing GPT-5.2-Codex https://medium.com/mlworks/the-next-frontier-of-agentic-engineering-introducing-gpt-5-2-codex-660a92384686 | |||
| 14:46 | Why it’s so hard to test AI apps https://ai.gopubby.com/why-its-so-hard-to-test-ai-apps-9934f3587a1f | |||
| 14:40 | Choosing the Right LLM Evaluation Framework in 2025: DeepEval, Ragas, Giskard, LangSmith, and… https://medium.com/@mahernaija/choosing-the-right-llm-evaluation-framework-in-2025-deepeval-ragas-giskard-langsmith-and-c7133520770c | |||
| 14:34 | AI Toolchains for Everyday Developers https://medium.com/@capali/ai-toolchains-for-everyday-developers-f698226f606a | |||
| 14:24 | DeepAgent: How AI Learned to Think, Discover Tools, and Solve Complex Problems https://medium.com/@robi.tomar72/deepagent-how-ai-learned-to-think-discover-tools-and-solve-complex-problems-a01a5592622e | |||
| 14:16 | Russian Blues — A Cat’s Tale in Colorspace https://medium.com/@mooreiarty/russian-blues-a-cats-tale-in-colorspace-11af6fa03117 | |||
| 14:16 | Russian Blues — A Cat’s Tale in Colorspace https://medium.com/the-accidental-geometer-the-geometric-whorfian/russian-blues-a-cats-tale-in-colorspace-11af6fa03117 | |||
| 14:16 | Prompt Engineering a Rug Pull: How I Built a Ponzi Scheme Simulator in Python https://zeeskylaw.medium.com/prompt-engineering-a-rug-pull-how-i-built-a-ponzi-scheme-simulator-in-python-697d39ee98e2 | |||
| 13:39 | Show HN: HN Wrapped 2025 - an LLM reviews your year on HN https://hn-wrapped.kadoa.com | |||
| 13:19 | Andrej Karpathy: 2025 LLM Year in Review https://twitter.com/karpathy/status/2002118205729562949 | |||
| 12:09 | World Models vs. Multimodal LLMs: The False Dichotomy Shaping AI’s Future https://medium.com/@tim_62250/world-models-vs-multimodal-llms-the-false-dichotomy-shaping-ais-future-dfe69e6a2de0 | |||
| 11:50 | Getting Started with LangChain: First Steps with LLMs and Inference APIs https://medium.com/@induwaragayashan/getting-started-with-langchain-first-steps-with-llms-and-inference-apis-f9d7a10e7c03 | |||
| 11:48 | Turning a 1.7B Model into a Math & Code Expert with MoE‑LoRA https://medium.com/@seanpark7109/turning-a-1-7b-model-into-a-math-code-expert-with-moe-lora-0967df35f7f1 | |||
| 11:32 | 10 Times SLMs Beat LLMs at Scale https://medium.com/@Modexa/10-times-slms-beat-llms-at-scale-f86a73c60064 | |||
| 11:25 | Your AI from Scratch: Building a Customised LLM https://medium.com/@lucamassaron/your-ai-from-scratch-building-a-customised-llm-d28fe750b137 | |||
| 10:58 | The End of the AI Goldfish: Solving Catastrophic Forgetting with Dual-Memory Architecture https://medium.com/@frankmorales_91352/the-end-of-the-ai-goldfish-solving-catastrophic-forgetting-with-dual-memory-architecture-4930a9be8837 | |||
| 10:39 | Lexicon for Transitional Vocabulary in the Age of Human–AI Relational Cognition by Peter Eidos on… https://medium.com/@cognitivesymbiosis/lexicon-for-transitional-vocabulary-in-the-age-of-human-ai-relational-cognition-by-peter-eidos-on-425fabaabb49 | |||
| 10:31 | Tokenizer Gotchas That Will Silently Break Your LLM (Ask Me How I Know) https://medium.com/coding-nexus/tokenizer-gotchas-that-will-silently-break-your-llm-ask-me-how-i-know-fdc1ea28f9da | |||
| 10:30 | Four things to focus on if you’re building AI agents right now https://cosminnovac.medium.com/four-things-to-focus-on-if-youre-building-ai-agents-right-now-8eb0538784c1 | |||
| 10:07 | 7 Ways Google’s FunctionGemma Is Transforming Natural Language Device Control https://medium.com/@nanthakumar18122000/7-ways-googles-functiongemma-is-transforming-natural-language-device-control-6269c530b725 | |||
| 10:02 | Beyond Skills: Rethinking Value in the Age of LLMs https://medium.com/@ktiyab_42514/beyond-skills-rethinking-value-in-the-age-of-llms-a98ab3d0f7f4 | |||
| 10:02 | Designing a Predictable AI Jury for a Cat Beauty Contest on GenLayer https://medium.com/@p.kolosov/designing-a-predictable-ai-jury-for-a-cat-beauty-contest-on-genlayer-1fa5458fe13b | |||
| 09:30 | The Blueprint for Quantum-Native LLMs: Spinor-Wave Networks (SWN) https://medium.com/@youth_k/the-blueprint-for-quantum-native-llms-spinor-wave-networks-swn-051313ea5f04 | |||
| 09:28 | The Role of Large Language Models Throughout the Data Science Lifecycle https://medium.com/sliitwif/the-role-of-large-language-models-throughout-the-data-science-lifecycle-01bc5a62dc65 | |||
| 08:47 | How to Build a Production-Grade LLM Gateway for Reliability and Control https://medium.com/@p10rajeshp/how-to-build-a-production-grade-llm-gateway-for-reliability-and-control-89493f7abb62 | |||
| 08:29 | PPO and GRPO Algos Explained Like You’re 15 https://medium.com/@keerthikonjety7_92524/ppo-and-grpo-algos-explained-like-youre-15-cd2ce17920fe | |||
| 08:06 | LLM Serving with vLLM https://ammarab.medium.com/llm-serving-with-vllm-23e3b1e0c617 | |||
| 07:36 | LangSmith Tracing Explained Clearly — Core Concepts, Real-World Use Cases, and When to Use What https://medium.com/@alikhizar9110/langsmith-tracing-explained-clearly-core-concepts-real-world-use-cases-and-when-to-use-what-8941a3c7d19a | |||
| 07:24 | GPT-5.2-Codex: A Big Step Forward for AI-Powered Software Engineering https://medium.com/@sharma.chetu04/gpt-5-2-codex-a-big-step-forward-for-ai-powered-software-engineering-059270f73a68 | |||
| 07:12 | Unraveling AI Hallucinations: Why LLMs Lie and How to Tame Them https://medium.com/@jiminlee-ai/unraveling-ai-hallucinations-why-llms-lie-and-how-to-tame-them-ec0bc21919e5 | |||
| 06:39 | Data Science in 2026: Prepare to Be Uncomfortable https://medium.com/ai-analytics-diaries/data-science-in-2026-prepare-to-be-uncomfortable-c05a5dedc0fa | |||
| 06:32 | The Strategic Shift: Small Models for Domain-Specific Edge AI https://medium.com/@sazzad1779/the-strategic-shift-small-models-for-domain-specific-edge-ai-f648a0c36903 | |||
| 06:17 | From Logs to Load Tests: k6 + Vector DB That Converts Production Logs Into Test Scenarios https://skakarh.medium.com/from-logs-to-load-tests-k6-vector-db-that-converts-production-logs-into-test-scenarios-c8294deced3e | |||
| 06:01 | Everyone tell about deep learning . But it’s still have a problem https://medium.com/@jiryanfarokhi/everyone-tell-about-deep-learning-but-its-still-have-a-problem-feec353f8616 | |||
| 05:59 | How I Stopped Writing Prompts — and Made AI Do It for Me https://pub.towardsai.net/how-i-stopped-writing-prompts-and-made-ai-do-it-for-me-332d17d4fe71 | |||
| 05:42 | Week 2: LLMS — The Brain of AI Agents https://medium.com/@janhvipawar01/week-2-llms-the-brain-of-ai-agents-0472ad9665de | |||
| 05:40 | InnerLight: Safety First Alignment for Mental Health Language Models https://medium.com/@Koder.Vaidya/innerlight-safety-first-alignment-for-mental-health-language-models-d61229052a0f | |||
| 03:48 | 1 Million AI Tokens, Zero API Keys: How 9xchat Just Changed the AI Workspace Game https://medium.com/@satyalk752/1-million-ai-tokens-zero-api-keys-how-9xchat-just-changed-the-ai-workspace-game-f7f585824755 | |||
| 03:33 | From Zero to Specialized: Fine-Tuning LLMs Fast with Unsloth LoRA https://medium.com/scriptkiddiez/from-zero-to-specialized-fine-tuning-llms-fast-with-unsloth-lora-abf1bd5835a3 | |||
| 03:32 | Advanced Prompting Techniques — Reasoning Control (Part 2A) https://medium.com/@er.rajkumaar/advanced-prompting-techniques-reasoning-control-part-2a-fce6ae8012dd | |||
| 03:15 | Nothing Will Die. Long Live the Transformers. https://medium.com/data-science-collective/nothing-will-die-long-live-the-transformers-42adf3732059 | |||
| 02:38 | Learning AI — Part 1 https://medium.com/@murthyp/learning-ai-part-1-e3c59347895d | |||
| 02:00 | Why Prompt Engineering Can’t Fix Hallucinations (But Neurosurgery Can) https://medium.com/@ariaxhan/why-prompt-engineering-cant-fix-hallucinations-but-neurosurgery-can-a1a7afa2f8bf | |||
| 01:32 | Elevating RAG from Novelty to Strategic Imperative https://consultkora.medium.com/elevating-rag-from-novelty-to-strategic-imperative-e7010b3ef16f | |||
| 01:32 | Elevating RAG from Novelty to Strategic Imperative https://pub.towardsai.net/elevating-rag-from-novelty-to-strategic-imperative-e7010b3ef16f | |||
| 01:26 | HyperBookLM: Building an Open-Source NotebookLM Alternative with Web Agents https://medium.com/@codebun/hyperbooklm-building-an-open-source-notebooklm-alternative-with-web-agents-8aeda25fc27f | |||
| 01:06 | Microsoft Agent Framework: Designing Your First AI Agent https://medium.com/@AIbatros/microsoft-agent-framework-designing-your-first-ai-agent-8453b749b55d | |||
| 00:52 | RAG Those Tweets: See What Patterns Emerge From That Long Archive https://medium.com/@nickmonts_39696/rag-those-tweets-see-what-patterns-emerge-from-that-long-archive-dc183693bdd7 | |||
| 00:26 | Top 5 Local LLMs You Can Run at Home https://medium.com/@DevSphere/top-5-local-llms-you-can-run-at-home-34f7ec949880 | |||
| 00:11 | Implementing an Agentic System for Purchase Request Consolidation https://medium.com/@nayan.j.paul/implementing-an-agentic-system-for-purchase-request-consolidation-c800f0ea07d0 | |||
| 00:02 | Build Your Own Voice Studio with Google Colab https://medium.com/@bishakhghosh0/build-your-own-voice-studio-with-google-colab-f14564cfb02d | |||
| Friday, 2025-12-19 | ||||
| 23:40 | How Do We Evaluate LLMs? Explained like you are 12 https://medium.com/@keerthikonjety7_92524/how-do-we-evaluate-llms-explained-like-you-are-12-d22d9c72aa8a | |||
| 23:07 | Beyond “Burstiness”: Deep-Cut AI Detection Signals the Gurus Aren’t Talking About https://medium.com/@rudratech/beyond-burstiness-deep-cut-ai-detection-signals-the-gurus-arent-talking-about-e77d608fe901 | |||
| 23:04 | The Cognitive Risk Hub: A Multi-Agent GenAI Architecture for Integrated Banking Risk Management https://medium.com/@armankamran/the-cognitive-risk-hub-a-multi-agent-genai-architecture-for-integrated-banking-risk-management-e859df5b5c3b | |||
| 23:02 | How to Perform Agentic Information Retrieval https://pub.towardsai.net/how-to-perform-agentic-information-retrieval-142d4e8ba89c | |||
| 22:48 | We ran Anthropic’s interviews through structured LLM analysis https://www.playbookatlas.com/research/ai-adoption-explorer | |||
| 22:21 | Retrieval Augmented Generation (RAG) with Databricks https://medium.com/@techgeorge/retrieval-augmented-generation-rag-with-databricks-68f4bc44ae74 | |||
| 21:55 | Part 3: The Brain in the Loop (LLM Orchestration & Tool Use) https://medium.com/@inceb1997/part-3-the-brain-in-the-loop-llm-orchestration-tool-use-f04ae00423d3 | |||
| 21:49 | Show HN: I Built an Image Captioning Tool Using Llama.cpp https://github.com/paradox460/imagecaption | |||
| 21:44 | Phala Alcança Conformidade SOC 2 Tipo I e HIPAA https://medium.com/@phalaportugues/phala-alcan%C3%A7a-conformidade-soc-2-tipo-i-e-hipaa-5204b1a44fe4 | |||
| 21:34 | My summary of the “Career Advice in AI” Lecture https://medium.com/@m.nusret.ozates/my-summary-of-the-career-advice-in-ai-lecture-445405fd2cf4 | |||
| 21:09 | Google & MIT Just Revealed Why Multi-Agent Systems Are Failing at Scale https://blog.cubed.run/google-mit-just-revealed-why-multi-agent-systems-are-failing-at-scale-3224169fd233 | |||
| 20:52 | Why we’ll never reach AGI with current language models https://medium.com/@benratcliffe_/why-well-never-reach-agi-with-current-language-models-d2571e631cc3 | |||
| 20:49 | LLM Year in Review https://karpathy.bearblog.dev/year-in-review-2025/ | |||
| 20:47 | Semrush Just Moved Inside ChatGPT. This Is Bigger for SEO Than It Looks https://medium.com/@Michael38/semrush-just-moved-inside-chatgpt-this-is-bigger-for-seo-than-it-looks-ce8a6c03c7fd | |||
| 20:45 | From Raw Text to a Trained Small Language Model https://medium.com/@anujagadde18/from-raw-text-to-a-trained-small-language-model-9e2feeb5413f | |||
| 20:29 | CAD: Disaggregating Core Attention for Efficient Long-Context LLM Training https://hao-ai-lab.github.io/blogs/distca/ | |||
| 20:28 | GPT-5.2, Grok 4.1, and DeepSeek v3.2 compare as Santa agents https://veris.ai/blog/santabench | |||
| 20:21 | Why Prompt Engineering Is Not a Long-Term Skill https://medium.com/data-science-collective/why-prompt-engineering-is-not-a-long-term-skill-10ce79e8e704 | |||
| 20:02 | Knowledge Graphs as the Deterministic Engine to Break the Commercial Ceiling of Enterprise AI https://pub.towardsai.net/knowledge-graphs-as-the-deterministic-engine-to-break-the-commercial-ceiling-of-enterprise-ai-a07f31dca5b9 | |||
| 19:39 | From Heavy RAG to LightRAG: Optimizing GenAI Architecture for Sustainable Enterprise Value https://medium.com/@armankamran/from-heavy-rag-to-lightrag-optimizing-genai-architecture-for-sustainable-enterprise-value-62ac3fd54e99 | |||
| 19:37 | AI's Unpaid Debt: How LLM Scrapers Destroy the Social Contract of Open Source https://www.quippd.com/writing/2025/12/17/AIs-unpaid-debt-how-llm-scrapers-destroy-the-social-contract-of-open-source.html | |||
| 19:16 | I Tried Becoming a Data Scientist Without Kaggle — Here’s the Truth https://ai.plainenglish.io/i-tried-becoming-a-data-scientist-without-kaggle-heres-the-truth-2274bb6f8db9 | |||
| 19:10 | Notes on LLM Behavior and Prompt Sensitivity https://medium.com/@lorenzo.kotalla/notes-on-llm-behavior-and-prompt-sensitivity-c398621900a9 | |||
| 19:08 | The Hidden Cost of “Chain of Thought” in Production LLM Systems https://medium.com/@mandalidevaharshini/the-hidden-cost-of-chain-of-thought-in-production-llm-systems-c1d366292816 | |||
| 19:03 | Why Teaching AI to Think Less Might Be the Breakthrough It Actually Needs https://ai.plainenglish.io/why-teaching-ai-to-think-less-might-be-the-breakthrough-it-actually-needs-37e4cbd7c9d1 | |||
| 19:02 | Data Governance & Retrieval-Layer Filtering https://pub.towardsai.net/data-governance-retrieval-layer-filtering-efd1dbd358d2 | |||
| 18:39 | From Terabytes to Insights: Building Production-Grade LLM Evaluation Systems at Scale https://medium.com/@shail.subscribe/from-terabytes-to-insights-building-production-grade-llm-evaluation-systems-at-scale-a693f876e833 | |||
| 18:23 | The Architecture of AI Dialogue: Prompt Engineering in the Era of Competing Cognitive Models https://medium.com/@shashwatabhattacharjee9/the-architecture-of-ai-dialogue-prompt-engineering-in-the-era-of-competing-cognitive-models-59b7e3195799 | |||
Original data from HuggingFace, OpenCompass and various public git repos.
Release v20241124