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.
210 healthcare financing deals. Major M&A for innovative drug outbound (Terns Pharmaceuticals $6 billion) / AI for drug (Huashen Zhiyao $787 million) / domestic substitution of medical devices / concentrated Hong Kong stock IPOs
201 frontier technology + tool software funding events. 5 leading large model companies / embodied intelligence explosion / quantum computing power taking off / AI for X application layer / profile of leading investors
The title identifies 2024 and 2025 as crucial years, while much of the episode addresses immature technology. Its central question is how to believe in a long-term opportunity while obtaining customers, revenue and learning time through work deliverable now.
This piece reorganizes 29 design/AI tools industry news items collected by Will (as of the 2025–2026 timeframe) into 6 trend themes instead of presenting them as a chronological list. It is written for non-technical PMs, prioritizing Why (what trend it signals) over What. ① The glass-mimic design wave sparked by Apple Liquid Glass (iOS 26 glass language + Beta3 frosted callback + Figma Glass effects + official UI Kit; a design language defined by one platform and followed by the entire industry) ② The AI design tool arms race among Figma/Adobe/Canva/Google (Acrobat Studio integration / Canva editable design AI / Figma AI image editing / Mixboard concept board; everyone is building a "one-stop AI creative workspace", and design tools are shifting from "canvas" to "generation engine") ③ AI-generated interfaces and vibe coding reshaping the design-development boundary (Cursor visually modifying real pages / A2UI protocol / Flipbook real-time video frame interface generation; UI is moving from "fixed pages drawn by hand" to "AI-generated on the fly") ④ Creative tool platformization and upgraded design references (Creative OS / Mobbin animation demos / Sketch·YouTube·Office icon redesigns) ⑤ Retention/monetization involution among super apps and social platforms (X Everything App / Instagram·Meta countering TikTok / Telegram gift economy / Spotify remixing) ⑥ AI reshaping search, translation, maps, and localization (Web Guide / Translate turning into Duolingo / Google Maps decision-style navigation / Meta·Google multilingual translation). Each trend is supported by news evidence plus implications for PMs/designers. Timeliness note: This is a point-in-time snapshot from late 2025 to early 2026; news will become outdated, while trend signals are more durable than specific features.
In 2023, Dify helped users understand new technology through visual prompt engineering and RAG. By this retrospective, Luyu was talking more about business users, upgrade compatibility, organizational processes and human feedback. That shift offers a useful way to examine AI products: as model capabilities enter real work, the product’s responsibilities expand.
dontbesilent (Old D) 2026-05 AI Self-Media Course, a complete 13-lesson learning path. It fully covers the entire production chain for users in mainland China who do not use scientific internet access and have no coding background to do self-media with AI, including model selection, the three layers of prompts (meta/system/user), the essence of RAG knowledge bases, AI copywriting imitation SOP, zero-code agent usage with Claude Code, LobeChat for blogger style imitation + conversation branching + Artifacts, HeyGen digital human translation + Tongyi Wanxiang wan2.2 animate character replacement, one-click Xiaohongshu publishing with Claude 4.5 + Playwright, and Claude Skill engineering for system prompt management.
How far is a model that answers questions from an application a business can use over time? This 2023 interview breaks that distance into handling information, testing instructions, organizing tools and operating the product after launch. Dify’s ambition includes a workspace through which business colleagues can keep shaping an application.
Jenny describes a Japanese financial company finding document organization and analysis insufficiently reliable. Even mostly correct output needed item-by-item checking, adding work. This turns a strong-model claim into a delivery question: how much work disappears, and what responsibility remains for errors?
After computer vision and process automation, Zhou founded another company in 2023 to connect language models with enterprise work. He offers an optimistic assessment: technology, customers and budgets are ready, while products must connect them. Yet he also acknowledges that useful scenarios often fail to exceed their full costs.
A synthesis of the "from 0 to 1 entrepreneurship methodology" from 5 schools of thought. Yitang Truman School (Five-Step Method + Key Assumptions + 76% Probability Math + 7-Layer Spectrum Map) / Wang Huiwen Tsinghua Product Course School (Scale Effects + 5 PM Greats + 5 Must-Read Books + High-Frequency Beats Low-Frequency + Meituan DNA) / Wang Kai Accounting School (Doing Business Without Fundraising + Three Things for Accounting + Four Big Mountains + Parallel Multi-Project Operation) / Lao Huang Independent Product School (Unit Economics + Business Entry Point + Strength-Oriented Approach) / Shensi Circle VC School (a16z / Valuation / Founders / Company Cases). Covers consensus, disagreements, exclusive dimensions of each school, and school selection by identity.
Tristan wants EVE to build a continuing relationship around the real user’s life: understanding earlier experiences, connecting them with present goals and participating in new daily conversations. He distinguishes this aim from jointly creating stories about characters. The host points out that extended use can also come from consuming content.
Listen: 01:27–03:59. Zhu identifies overconfidence, linear extrapolation, and herding. After prices rise, people remember having predicted them and credit their own ability. This encourages larger subsequent commitments.
Synthesis of 4 schools of perspective on the AI Coding tool stack (2024-08 → 2026-05). Huang Yihe (Cursor shock → v0+Claude+Cursor three-piece suite → Cline self-built MCP → Gemini 2.5 → Claude Code → OpenCode camp) / Wang Kai (product manager path: Cursor $50,000 → Claude Code terminal → disruptive data analysis → hundredfold token consumption) / Xiaohongshu vibecoding (three core mindsets + 0-80 vs 80-100 life-or-death line + Astro Kid 400 million financing) / Lao D (using OpenRouter to connect Claude Code in China without scientific internet access). Covers 4 code retrieval methods / three tiers of Ultra think / Plan Mode / tool stack division of labor philosophy / Vibe Coding cleansing prophecy
Fan describes approximately 25 years in products and design across industrial design, phones, games, media, operating systems and cars. He identifies a common capability: observe, understand, combine and compress, turning numerous aspects of life into a coherent solution people can directly experience.
A reorganized set of Will's usage notes for Figma Make (Figma's built-in tool that lets you "describe requirements → AI generates runnable prototypes/webpages"), targeted at PMs building AI products who want to produce prototypes quickly. It covers eight major functional modules, each providing "how to use + Why + when to use + pitfalls": ① Basics (three-column interface = chat/preview/code; four input methods = typing/pasting Figma designs/uploading images/clicking the preview to make targeted edits). ② Best practices for writing Prompts and feeding attachments (define the product using a "user capability list" instead of broad category nouns; do layout first then functionality; iterate step by step; use Auto Layout to help AI understand structure; common layouts are more likely to succeed on the first try; desktop version is more stable; never write secret keys in prompts). ③ Templates (publish files with established specifications as templates for the team to copy and remix; use guidelines to lock editable scopes; create/update/unpublish/use; four examples of constraint prompts). ④ Importing Figma design library styles (export the Library as CSS → reference it in Make → UI automatically applies the company's color scheme and fonts). ⑤ Adding Guidelines.md (set rules for AI, including general rules + design system rules + component-level rules; concise and precise rules are better than piling on context). ⑥ Feeding with code files (four types including .tsx/.csv/.json/.html; generate data dashboards with real data; reuse open-source components such as shadcn; real copy typesetting) + a strongly constrained prompt for one-click full-site refactoring to shadcn/ui. ⑦ Adding a backend (connect to Supabase with one sentence to implement login/data storage/secure secret key saving; the judgment mnemonic is "Does this thing need to be remembered?"). ⑧ Publishing (get a figma.site URL with one click, bind a custom domain via TXT+CNAME, upload web fonts, advanced options such as SEO/password protection) + push to GitHub/download code. Division of labor with [[ai-ux-experience-design]]: that article covers "the process methodology of using AI for experience design", while this article is "the functional operation manual for the single tool Figma Make"; it complements the vibecoding mindset in [[ai-vibecoding-design-process]]. The target readers are non-technical PMs, with industry jargon explained in parentheses, and each section includes Why and when to use it.
25 hub wiki entries from the Anthropic engineering blog. Building Effective Agents framework (Workflow vs Agent + 5 patterns) + new Context Engineering paradigm (evolution of prompt engineering) + 4 categories of Harness design (single-shot / long-running single agent / long-running multi-agent / managed-decoupled) + 5 Tool design principles + Agent Skills open standard + Claude Code engineering (sandboxing / auto mode / best practices / 2 postmortems) + MCP evolution (code execution / DXT/MCPB / advanced tool use). 1-sentence index for all 25 posts. The Anthropic perspective is the "official first-party" layer in the [[wiki-ai-agent-architecture]] / [[wiki-claude-code-workflow]] series, cross-checked against third-party observations.
A prompt contains instructions, background and rules supplied to a model. Sharing one can resemble sharing a template, but a consumer may actually want a text adventure, a conversational character or a tool that repeatedly completes a task. FlowGPT’s changing description reflects a move from collecting instructions toward consuming those experiences directly on its platform.
Complete in-depth reading of the 2023 open-access Springer textbook by Gerhard Paaß and Sven Giesselbach. 436 pages / arXiv:2302.08575. 8 chapters covering: Ch1 Classical NLP fundamentals (token / embedding / RNN / CNN) / Ch2 Three major PLM architectures (BERT autoencoder + GPT autoregressive + Transformer seq2seq) + training/regularization/uncertainty/interpretability / Ch3 Six major directions for improving PLMs (pretraining task variants + long-range sparse attention + multilinguality + knowledge injection + model scale + fine-tuning / few-shot / CoT) / Ch4 PLM knowledge evaluation (GLUE / SuperGLUE / BIG-bench + Probing) / Ch5 Information extraction (classification / NER / relation extraction) / Ch6 Text generation (retrieval / QA / translation / summarization / stories / code / dialogue) / Ch7 Multimodality (speech / image ViT / video / control Decision Transformer + GATO / DNA / Protein) / Ch8 Summary and outlook (new paradigm of Foundation Models / economic applications / 5 categories of harms / Grounding / Fast & Slow Thinking / Planning). Serves as the "academic foundation" supplement to [[newtype-agent-architecture-research-2025]] and [[newtype-agent-research-papers]]
The first 42 Zhangjing episode moves from a fund’s experience to AI and then vitality. Zhang imagines bots as new users of software, with personal agents, merchant agents, communication and accountability developing together. The introduction reports investments in more than ten AI companies at the time. That is a 2023 claim, not verification of today’s portfolio or returns.
Listen: 00:56–13:46. Wang treats Manus and OpenClaw as two releases of capability. The former demonstrated reasoning models in research tasks; the latter connected coding to more tasks, communication channels, and memory. Messaging integration is not the whole explanation. The coding agent’s ability and the surrounding configuration, coordination, and presentation work together.