Signal to BuildSIGNAL TO BUILD中文Owner sign in ↗Explore curated knowledge, course notes and methods, with their existing content and sources preserved.
This index lists complete English editions. Additional Chinese notes and bilingual updates remain available in the Chinese index.
Explains Claude Code's Linux Bubblewrap sandbox, file and network boundaries; provides verified minimal examples and in-product checking steps, while retaining historical security analysis and competitor comparisons.
This article organizes Cal Rueb's historical share on Claude Code, building a reading path around project rules, permissions, context, Hooks, and MCP; original cases are retained, and specific commands should be verified against the current version.
A synthesized rewrite of Claude Code workflows from the perspectives of 7 authors: Cal Rueb (Anthropic internal engineer)/ Wang Kai (independent developer with a multi-device matrix)/ Huang Yihe (main battleground of tool stack evolution)/ Cat Wu (Anthropic PM Head for Claude Code)/ Lao D (non-programmer zero-code user)/ Xiaohongshu vibecoding (skill-based creative applications)/ **zara zhangzala (liberal arts background + Co-founder mental model + Feishu CLI + OpenAI Realtime + Cut features before shipping)**. Covers core positioning, Agentic Search vs RAG, multi-layer CLAUDE.md configuration, Permission/Hooks, three Think Hard levels, Sub-Agent philosophy, Skill system, explicit divergences, and implementation choices.
A hub of 30 user operation scenarios: 14 articles from claude.com/resources/tutorials plus 16 use cases. **Claude Cowork** (Claude with Hands): what Cowork is / how to choose between Cowork and Chat / onboarding / delegating your first task / Dispatch (controlling your computer from your phone) / plugin system / Customize / Navigation. **Claude Design** (Anthropic Labs): creating presentations / prototypes / UX. **16 use cases**: concrete scenario scripts ranging from the Claude for Excel finance use case to "live whiteboard classroom scenario", "adjust textbooks to different reading levels", "operate any app from your phone", and more. **New concepts**: 4 Ds of AI Fluency (behavioral indicators) / Sycophancy mitigation / Hallucination education (the latter two have been moved to the interpretability wiki).
Hub for Anthropic product release timelines. Covers Opus 4.7 / Opus 4.6 / Sonnet 4.5 / Sonnet 4.6 / Haiku 4.5 model generations + Max Plan (20x usage) + SpaceX compute partnership + new Claude Design product + the "Claude is a space to think" positioning statement (no ads). Condenses 9 product news posts + 3 product landing pages into a "generation timeline + positioning evolution + capability comparison table".
A synthesis of Skill engineering across 7 perspectives. newtype Lao Huang (Super Analyst three-piece set + Sequential Thinking + 7-Stage Workflow) / Wang Kai (three major pitfalls of Skill + writing Skills in code + IFS + SillyTavern 7 stages) / Xiaohongshu vibecoding (advantages over N8N + calling local services + Vibe Marketing + IELTS three-piece set) / Lao D (SKILL.md + references/ + Git version management) / Anthropic official (Skill format = zip bundle = SKILL.md + references + scripts) / **zara Zhang Zala (opinionated distribution school + Centrally fetch locally Remix + Frontend Slides 16k+ stars + Fork > PR + Skill > SaaS)**. Covers the boundaries between Skill vs MCP vs Sub-Agent, rule applicability scope, memory cross-contamination, and the problem that more Skills lead to more conflicts.
3 close readings of Agent research / talks collected by newtype Planet. 05 CAMEL-AI Wendong Fan "Finding the Scaling Laws of Agents" (2025 talk: # Parameters → # Agents analogy / CAMEL + OWL + CRAB + OASIS frameworks / Agentic Data Generation feeding back into training) / 07 "More Agents Is All You Need" (Tencent ICML 2024: brute-force scaling of Agent count via sampling-and-voting, Llama2-13B + 15 agents matches Llama2-70B, 3 dimensions determine gain magnitude) / 08 "ChatQA" (NVIDIA 2024: two-stage instruction tuning, 70B scores 54.14 vs GPT-4's 53.90 across 10 conversational QA datasets, fine-tuned retriever replaces query rewriting to save API costs, "unanswerable" sample training eliminates hallucinations)
A complete operation manual compiled by Will (will-ai-note) on "using the Codex Agent to write real designs directly to the Figma canvas via Figma MCP". It covers three parts: ① Environment setup — installing the Figma remote/desktop MCP server for Codex (to grant Codex permission to operate Figma) + installing/managing the official Figma Skill; ② Uses and prompts for seven core skills — figma-use (directly modify the canvas: create Frames/components/variables/Auto Layout), figma-use-slides (generate PPT inner pages), figma-generate-diagram (text → editable flowcharts/sequence diagrams/ER diagrams), figma-generate-design (code/HTML → Figma UI), figma-generate-library (codebase → design system Library), prototype-to-figma (running prototype → frame-by-frame Figma state drafts), Community skills; ③ Combined workflows — two-way sync between code prototypes and Figma, direct interface output from base screenshots, interaction framework → new interface, scanning local SVG icons for automatic semantic matching and replacement. Each section provides the Why/when to use/reusable prompts/pitfalls (e.g., figma-use requires a paid Full/Dev seat, Dev seats can only write in drafts, prototype-to-figma has spacing bugs but can split pages frame by frame). Division of labor with [[newtype-ai-coding]]: that article covers AI Coding tool stack selection and core mindsets, while this article focuses on the concrete implementation down to skills and prompts for the "design/prototype engineering" segment.
Across Claude Code, Skills and Clawdbot, Huang sees coding capability spreading outward: a system can write programs that supply missing tools. Performing an action still differs from delivering it repeatedly, reliably and economically.
逗逗游戏伙伴 is an AI desktop companion that watches gameplay, accompanies players and converses. Binson’s approach asks users to allocate less additional time to AI content: participate in something they already want to do. The game continues while the companion acts like interactive commentary beside it.
Summary of 13 official Google case studies by topic: 5 Job postings cases (Eventbrite/Jobrapido/Monster/ZipRecruiter/Saramin) + 4 Video SEO cases (Vimeo/MX/Vidio/cross-regional including Weather/Italiaonline/ABP) + 1 Recipe case (Rakuten) + 1 Discover large image case (Kirbie/Istoé) + 1 CMS integration case (Wix) + shared patterns. **These are best-case outcomes and cannot be expected to scale proportionally**, but they demonstrate the value of SD, tool combinations, iframe embedding solutions (indexifembedded), and the Wix CMS implementation path
Distinction between multilingual and multi-regional sites. 3 equivalent hreflang implementation methods (HTML link / HTTP header / Sitemap), choose only one. 6 iron rules of hreflang (every version lists all versions including itself + bidirectional links + absolute URLs + ISO 639-1 + ISO 3166-1 + x-default). 4 URL structure options (ccTLD strongest signal / subdomain / subdirectory / parameter not recommended). Google's signal priority for selecting locale (ccTLD > hreflang > server location > others). Googlebot geo-distributed crawling behavior for locale-adaptive pages. SafeSearch filtering algorithm + EDSA (Educational/Documentary/Scientific/Artistic) exemptions. 5 best practices for explicit content (prevent UGC CSAM / allow fetching of videos / do not block age gates / separate domains / mark meta rating="adult") + 5 common misclassification fixes.
Google's 3-stage JavaScript processing pipeline (crawl / render / index) + render queue mechanism + WRS (Web Rendering Service). 15 best practices + an 11-step diagnosis checklist. Common SPA (Single-Page Application) pitfalls: soft 404 errors / fragment routing / Local Storage failures / cache fingerprinting / traps of injecting robots meta via JS / shadow DOM in Web Components. Dynamic rendering is a legacy workaround; SSR / static rendering / hydration are recommended instead. Use viewport detection for lazy loading + convert infinite scroll to paginated URLs. Web Components must use Slot.
Google's 3-stage JS processing (crawl/render/index) + render queue mechanism + WRS (Web Rendering Service). 15 best practices + 11-step troubleshooting checklist. Must-avoid SPA pitfalls: soft 404 / fragment routing / Local Storage failure / cache fingerprinting / robots meta JS injection trap / web components shadow DOM. Dynamic rendering is now a workaround; SSR / static / hydration are recommended. Lazy load viewport + infinite scroll → paginated URLs. Web Components Slot must be used
Overview of the Structured Data (SD) framework and all 34 rich result types supported by Google. How to choose among the 3 formats (JSON-LD recommended / Microdata / RDFa; data-vocabulary.org is deprecated). **5 quality guidelines** (content / relevance / completeness / location / specificity) plus images + 2 patterns for multiple items on one page (nesting / linking independent items with @id). The complete type table is organized by category (e-commerce / organization / sports / recruitment / entertainment / news / dining / education) + required fields for each type / key pitfalls. **Latest deprecation**: FAQ will no longer show rich results starting 2026-05-07 (tools will be removed in 2026-06 / API will be removed in 2026-08). Generate SD with JavaScript (Google Tag Manager / custom JS) + note that crawl frequency will decrease when using GTM for Product-type SD. For image copyright metadata, use schema.org/CreativeWork (creator + creditText + copyrightNotice + license + acquireLicensePage). Speakable is limited to news + en-US/-CA/-AU/-IE/-NZ/-UK + a few sentences totaling ≤ 30 seconds. Difference between SD for paywalled content and cloaking (policy violation): you must mark isAccessibleForFree=false and use cssSelector. Carousel is only available for 4 categories: Recipe/Course/Restaurant/Movie + 4 cases for ItemList @id linking.
Structured Data framework + overview of all 34 Google-supported rich result types. Format selection for the 3 syntaxes (JSON-LD recommended / Microdata / RDFa; data-vocabulary.org is deprecated). **6 quality guidelines** (content / relevance / completeness / location / specificity / images) + 2 patterns for multiple items on the same page (nesting / individual items linked via @id). Complete categorized type table (Ecommerce/Organizations/Sports/Jobs/Entertainment/News/Food/Education) + required fields / key gotchas for each type. **Latest deprecation**: FAQ rich results will no longer be displayed starting 2026-05-07 (tools removed in 2026-06 / API removed in 2026-08). JavaScript-generated SD (Google Tag Manager / custom JS) + note that Product SD via GTM may face reduced crawl frequency. Image license metadata schema.org/CreativeWork (creator + creditText + copyrightNotice + license + acquireLicensePage). Speakable is only for news + en-US/-CA/-AU/-IE/-NZ/-UK + a few sentences totaling ≤ 30 seconds. Paywalled content SD must be distinguished from cloaking by marking isAccessibleForFree=false + cssSelector. Carousel (only available for Recipe/Course/Restaurant/Movie) + 4 categories of ItemList @id linking.
All visual elements on Google Search result pages. 4 result types (Text/Rich/Image/Video) plus Exploration features. Attribution (Favicon/Site name/Visible URL/Domain/Breadcrumb). **Title link**: 11 best practices + 6 major Google override scenarios (half-empty/obsolete/inaccurate/duplicate/no clear main content/lang mismatch) + 9 sources. **Snippet** elements + 6 meta description guidelines + 3 rules for Read more deep links. **Favicon**: 6 implementation rules (domain/subdomain level only + 8x8px minimum + must be crawlable by Googlebot-Image). **Sitelinks**: 4 guidelines + removal methods. **Featured Snippets**: fully blocked by nosnippet / partially blocked by max-snippet / cannot be actively marked up. **Enriched search** (Jobs/Recipes/Events are the only ones) + 5 quality guidelines. **AI Features** (AI Overviews + AI Mode): no special SEO requirements, **no AI text file / no special schema**, control with nosnippet/max-snippet/noindex + use Google-Extended to control AI training. **Translated results**: 21 languages + notranslate opt-out + 9-step algorithm for providing URL decoding to ad networks. **Preferred sources** (Top Stories): users actively select preferences + deeplinks + button assets.
Unified index and core reading guide for 5 mainstream large model textbooks. ① Understand AIGC in One Book (2023, 256 pages, general science popularization) ② Large Model Development and Fine-Tuning from Scratch Based on PyTorch+ChatGLM (376 pages, hands-on textbook) ③ Minimalist Introduction to Large Model Application Development Based on GPT-4+ChatGPT (177 pages, O'Reilly, 2024 SDK hands-on ★ deep intensive reading) ④ Large-Scale Language Models: From Theory to Practice (297 pages, Fudan academic school, Zhang Qi, Gui Tao, Zheng Rui, Huang Xuanjing) ⑤ Large Model Black Book Based on GPT-3/4 Transformer NLP (379 pages, Denis Rothman). Covers the complete path from classic NLP theory → Transformer → GPT evolution → Prompt engineering → fine-tuning → application development
A curated compilation of highlights from 102 Medium UX articles (59 topics) saved by Will, compressed across 7 major clusters into one note that can replace an entire bookmarks folder. ① GenAI/AI UX Patterns — 21 GenAI UX Patterns (augmentation vs automation / three tiers of automation / CoT display / expressing model confidence and limitations / guardrails), seven AI UI layouts (floating window / inline / infinite canvas / centered chat / left panel / right panel / grid), predictive UX (context management controls), five-step framework for designing AI products. ② Research & Experiment Methods — five-step A/B testing method (define success → identify bottlenecks → build hypotheses → prioritize → test, including empirical evidence from the Obama campaign's $57 million and Haiti disaster relief's $1.02 million), macro vs micro vs vanity metrics. ③ Mobile & Trends — nine 2026 mobile trends (predictive / Zero-UI / Agentic / spatial 3D / glassmorphism / microinteractions / adaptive / ethics / passwordless). ④ Inclusivity/Accessibility — 11 tactics (4.5:1 contrast ratio / multi-channel coding / external form labels / alt text / heading hierarchy / touch targets ≥7–10 mm² / keyboard accessibility), including empirical revenue evidence from Tesco/Virgin. ⑤ Information Architecture/Flows/Complexity — collaboration framework (centralized → decentralized shared Commons), how complexity accumulates (feature bloat / process stacking / technical debt + Pareto cuts + complexity budget), value proposition canvas, globalization ≠ translation (Chevrolet Nova / Domino's vs Kellogg's). ⑥ Company Cases — breakdown index for Airbnb/Calm/Canva/Duolingo/Spotify/Uber/Slack/Gmail and more. ⑦ Career/Soft Skills — three types of designers in 2026 (block-shaped generalists / ultra-deep specialists / standout new graduates), real UX competence in the AI era (restraint and judgment > drawing UI), six psychology principles for landing page conversion. For AI product PMs: Why / when to use / counterexamples come first. Highly intertextual with [[shape-of-ai-patterns]] (AI UX Pattern Library), [[ai-ux-interactions]] (Seven AI Interaction Paradigms), [[laws-of-ux]] (Laws of UX), [[nngroup-ux-principles]] (UX Golden Quotes), [[nngroup-information-architecture]], [[nngroup-persuasive-design]], and [[ux-metrics-roi]]; this compilation only adds the checklists, numbers, and cases they do not cover, and does not rewrite them.
9-stage closed-loop data content workflow (data → keyword selection → planning → writing → de-AI-ifying → localization → on-page → quality gate → iteration) + 4 hard practical lessons (competitor facts must be verified and dated / compete on "the best stuff is free" not just "free" / deliver complete final drafts instead of diffs / content hits diminishing returns at scale, don't pile pages on new sites that lose index) + off-page authoritative backlink playbook (directories / review outreach / communities / "open-source hosted version" brand leverage) + main skill orchestrating subagents. Fills the missing "how to operate" layer that the gsc-* reference layer lacks.