interactive-explainer

Grok Bot 头像动画引擎的逆向解剖：可交互讲解页 + 动画编辑器（程序化 SVG / 贝塞尔 clip / 弹簧 / 球面眼睛），0 依赖 0 构建 · Reverse-engineered anatomy of the GrokBot avatar animation engine — interactive explainer + editor, zero-dependency procedural SVG

生成美观自包含单文件HTML工件的技能，13种模式、五维评审、DevLoop迭代 |A skill for generating beautiful, self-contained, single-file HTML artifacts. 13 patterns, 5-dimension critique, DevLoop iteration.

LLM Architectures Explained: how language-model architectures differ, model by model, with every value sourced, a tested cost model and diagrams generated from the data.

Interactive explainer: BDH's attention has no softmax, so it is exactly a Hebbian synaptic memory — one fixed-size matrix written once per token. Run both forms, watch them agree to 1e-16, then break it with one toggle. DataForge 2026, Pathway track.

Interactive explainer of market making: 13 in-browser simulations from the limit order book to Avellaneda–Stoikov, Glosten–Milgrom, and Kyle's λ.

The complete ASD-STE100 Simplified Technical English skill for Claude Code, Codex, and LLMs. Features Andrej Karpathy's Ladder of Understanding (STE-100, diagrams, interactive HTML, 3b1b video), linter, and prompt compiler.

Multi-audience interactive explainers (JEPA & world models, orbital mechanics, quantum computing, elasticity, electromagnetism) + the Claude Code skill that generates them

Seven kinds of website, one AI agent. A WebMCP-powered interactive explainer where your agent proves what agents do to the web, room by room.

Interactive explainers on robot data: LeRobot datasets, ROS bags, MCAP, and which format to use

Agent Harnesses Explained: the loop between a model and the world (tools, permissions, context, caching, sub-agents), animated by a deterministic simulator whose Python and TypeScript halves agree token for token.

The Atlas is a knowledge base of language models with information on how they work as well as agentic loops and loop engineering!!

Agent Security Explained: how to build agents that stay safe, defences only: trust boundaries and labels, untrusted input, the lethal trifecta, least privilege and sandboxing, driven by a tested policy engine.

Language as lossy compression of thought: an interactive explainer, in a single self-contained HTML file, on which kinds of thought survive being put into words.

Numerics Explained: number formats and quantisation for AI (FP16 to FP4, MX, rounding, accumulation, quantisers), animated by a bit-exact numerics library checked against numpy and ml_dtypes.

LLM Inference Explained: an interactive guide to how real systems serve language models, from the KV cache and roofline to batching, speculative decoding and disaggregated serving, with a tested simulator in your browser.

Agent Orchestration Explained: agents as graphs, steps and checkpoints, multi-agent patterns and durable runs, on an engine checked against real LangGraph in CI.

GPU Kernels Explained: how a GPU actually executes the maths (memory hierarchy, roofline, coalescing, bank conflicts, occupancy, GEMM, FlashAttention), animated by a tested execution model.

Agent Evals Explained: how to tell whether an agent is good, and how sure you can be: benchmarks, pass@k vs pass^k, intervals, paired comparisons, graders and cost, driven by a tested engine.

Agent Context Explained: what goes into an agent's context window, retrieval and memory, with every number measured on SQuAD with real embeddings and recorded reranker scores.

AI Racks Explained: the machines underneath AI, from the package and NVLink racks to networks, power and cooling, animated by a tested model with every hardware figure sourced and dated.