https://sudonull.com/compression-of-glm-5-1-on-16-gb-vram-gqa-hacks
Breakdown of compressing 744B MoE model GLM-5.1 to 388 MB for T4. Fix GQA errors, config code, inference. For middle/senior dev — repeat on your hardware.
Search Articles — Sudonull
Search Results
In this project
https://sudonull.com/text-compression-brentwick-7-to-50-tokens
Learn how Cambridge compresses texts to prompts with 98% accuracy. Brentwick-7 method for developers: latent reduction, embeddings, markets. Test it yourself.
https://sudonull.com/turboquant-6x-compression-of-llm-kv-cache
Break down Google's TurboQuant: PolarQuant + QJL for 6x KV-cache compression without accuracy loss. Inference acceleration up to 8x. Benchmarks, principle of operation for LLM developers.
https://sudonull.com/llm-hallucinations-data-compression-artifacts
Breaking down LLM hallucinations through lossy compression: why they occur, how to minimize with RAG and fine-tuning. For developers: Shannon's theory and examples. Learn how to work with artifacts.
https://sudonull.com/turboquant-lossless-kv-cache-compression-for-ai
Learn how Google's TurboQuant compresses transformer memory down to 3 bits with PolarQuant and QJL. Benchmarks on Gemma, Mistral. Optimization for AI developers.
From other projects
https://lrivo.com/numbness-in-the-left-arm-causes-and-algorithm-of-actions
Find out what causes numbness in the left arm: from heart attack to nerve compression. Self-check algorithm and when to urgently call a doctor.
https://ymaho.com/rarity-of-heart-cancer-mechanical-force-of-contractions
Science reveals centuries-old mystery: why the heart does not get cancer. Compression suppresses proliferation through Nesprin-2. Learn about a new class of drugs.
https://ymaho.com/inflation-in-russia-5-47-why-this-is-not-a-reason-for-joy
Annual inflation slowed to 5.47%, but behind the statistics lies stagflation and demand compression. Find out what will happen to the Central Bank rate and the OFZ market. Risk analysis.
From the web
GitHub - 0xk1h0/ChatGPT_DAN: ChatGPT DAN, Jailbreaks prompt
https://github.com/0xk1h0/ChatGPT_DAN
ChatGPT DAN, Jailbreaks prompt. Contribute to 0xk1h0/ChatGPT_DAN development by creating an account on GitHub.
Download GitHub Desktop
https://desktop.github.com/download/
Simple collaboration from your desktop Download GitHub Desktop Focus on what matters instead of fighting with Git. Whether you're …
GitHub - basketikun/chatgpt2api: ChatGPT官网接口纯协议的逆向实 …
https://github.com/basketikun/chatgpt2api
Apr 19, 2026 · ChatGPT2API 主要是对 ChatGPT 官网相关能力进行逆向整理与封装，提供面向 ChatGPT 图片生成、图片编辑、多图 …
大家用 grill-me (拷问我) 这个 AI Agent Skill 的体验如何？ - 知乎
https://www.zhihu.com/question/2054005413406946147
我经常用这个工具，也推荐给了很多人。 在有这个 skill 之前，我通常会在一段 instruction 的结尾要求 agent：“ 如果我的需求或设计有 …
Writing - Reddit
https://www.reddit.com/r/writing/
Your critique submission should be a top-level comment in the thread and should include: * Title * Genre * Word count * Type of …
如何看待 Codex 与ChatGPT 合并，实际使用体验有什么变化？
https://www.zhihu.com/question/2058849029027632241
Jul 10, 2026 · 到了ChatGPT Work、Codex和ChatGPT合并之后，几乎每天增长100万，曲线立即变得陡峭。 从用户增长策略来 …
United States Air Force Reddit
https://www.reddit.com/r/AirForce/
Community for current and past members of the US Air Force.
普通人要Codex有什么用？ - 知乎
https://www.zhihu.com/question/2020702567198410280
这也算是一个能分享的小技巧。 毕竟如果一个人自己开了 20x ChatGPT Pro 订阅畅用，那至少这个人的钱包我觉得不是很普通，国内 …
GitHub - rasbt/LLMs-from-scratch: Implement a ChatGPT-like LLM in ...
https://github.com/rasbt/LLMs-from-scratch
This repository contains the code for developing, pretraining, and finetuning a GPT-like LLM and is the official code repository for the …
chatGPT在国内怎么使用？ - 知乎
https://www.zhihu.com/question/1963658042986962975
ChatGPT是由OpenAI推出的一款AI聊天对话机器人，能够进行自然语言交互，帮助用户完成问答、写作、编程等多种任务。
Trending Now
DeepWay Robotaxi 1000 km: Solid-State Batteries and AI Revolution
Chinese startup DeepWay announced a robotaxi with a range of 1000 km on solid-state batteries. Learn how this changes the electric vehicle market and why it matters.
How to Learn Python for Beginners: Step-by-Step Roadmap
Discover how to learn Python for beginners with this step-by-step roadmap. Master fundamentals, data structures, and build real projects. Start your coding journey today.
What Is a Large Language Model? Beginner's Guide
Learn what is a large language model, how AI like ChatGPT works, key capabilities, limitations, and practical uses. Understand LLMs in plain English.
MongoDB vs PostgreSQL When to Use Each: Key Differences
MongoDB vs PostgreSQL when to use each: compare ACID compliance, scalability, schema flexibility, and query power. Choose the right database for your project.
React vs Vue Which Framework to Choose in 2025 | Comparison
React vs Vue which framework to choose in 2025? Compare performance, learning curves, job markets, and costs. Make a data-driven decision for your next project.
Nvidia capitalization $5.4 trillion: surpassed Europe
Nvidia reached $5.4 trillion capitalization, outpacing the 10 largest companies in Europe. Analysis of growth reasons, overheating risks, and forecasts for the AI chip market. Read the review.
How to Design RESTful APIs for Web Applications: Best Practices
Learn how to design RESTful APIs for web applications with best practices on URIs, HTTP methods, versioning, and documentation. Build scalable, intuitive APIs that developers love.
Docker Container Networking Explained: A Practical Guide
Docker container networking explained in depth: learn bridge vs host networks, DNS, port publishing, and how to build secure, scalable multi-container apps.
Top 10 Breakthrough Technologies 2026 from WEF: Full Analysis
Analysis of trends, hidden risks, and geopolitical implications of Top 10 Emerging Technologies 2026. Who benefits from new developments and who loses. Read the analytics.
Figure 02: new humanoid robot with AI interface from Figure AI
Figure AI introduced the humanoid robot Figure 02 with physical AI and OpenAI language model. The robot passed tests at BMW factories. Learn how humanoid robots are changing manufacturing.