KNOWLEDGE NOTESKnowledge Notes
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.
350knowledge notesThe essence of 19 Nielsen Norman Group UX Slogans, which compress decades of accumulated usability common sense into pithy lines you can stick at your workstation. They are clustered into 5 major themes: ① Don’t treat yourself as the user — You Are Not the User / Design for Them, Not for You / UX Is People / Users Aren't Lazy, They're Efficient. ② Less is more (restraint) — Keep It Simple (guard against Featuritis) / Less Is More (focus on core tasks and set a value threshold) / More Choices, More Trouble (choice overload) / Make It Easy (proactively iterate rather than passively exercise restraint, contrasting the “keep” vs “make” in Keep It Simple). ③ Communication is design — Communication, Not Decoration / You Can't Impose Joy (counterexamples: splash screens/animations) / Brevity is Brilliance (cut web copy in half) / This Is Too Easy to Understand, Said No One Ever (even experts prefer clarity) / Clarify, Don't Mystify (Information Scent · anti-clickbait). ④ Understand the problem first — Why Beats What (use Think-Aloud to find root causes) / Stop Solutioneering (list unknowns first, then do research). ⑤ Replace guesswork with evidence — If You're Not Checking, You're Guessing (100% vs 25% empirical evidence from forums) / Show Me the Data (quantitative A/B testing + qualitative interviews) / UX Researchers, We Like to Watch (behavior outperforms self-report · Margaret Mead/Descartes) / It Depends (the spectrum between best practices and context). Target readers = product managers; jargon is explained inline; each cluster comes with a Why explanation + counterexamples.
A complete reference after the major regulatory changes in May–June 2026 — the CSRC imposed heavy penalties on Futu/Tiger/Longbridge; starting 12 June, mainland investors can only sell/withdraw but cannot buy/deposit; HKMA requires investment account holders to sign a declaration that "funds come from outside mainland China", while savings accounts are explicitly unaffected; practical account opening guidance (traditional banks as primary choice + ZA Bank as backup/ID card accepted as proof of address/interview talking points/three key figures for fund compliance) + why credit card applications fail + QDII as the only legal channel for small capital to invest overseas + a list of debunked circulating claims. Highly time-sensitive; policies may change at any time during the two-year rectification period
Complete selection reference for anti-blocking egress in H1 2026 — protocol comparison (Reality is most anti-blocking / protocol does not matter for dedicated line scenarios) + buying off-the-shelf airports (line type IPLC > relay > direct connection / objective judgment signals / limbopro runaway blacklist / candidate list) + self-built VPS (line IPLC > CN2 GIA > GT > direct connection / reality-ezpz) + mihomo url-test automatic failover + list of marketing numbers rejected by adversarial verification. Highly time-sensitive; airport names expire on a weekly basis, methodology remains valid long-term
Reference for residential proxy + anti-detect fingerprint browser selection and China deployment for multi-account farming — Datacenter IPs (including self-built VPS) cannot be used for account farming; iron rule of one static residential IP per profile; AdsPower (smoothest payment via Alipay) + IPRoyal/Decodo (Alipay/USDT); two-layer topology of GFW-bypass tunnel chained with residential proxy (no proxy chain inside profiles); underlying GFW-bypass must have obfuscation (Reality qualifies) to pass GFW entropy detection; WebRTC/DNS leak prevention verification; list of rejected marketing exaggerations. Highly time-sensitive, market conditions change rapidly
A full review and structural analysis of the 25-day conflict between the US-Israel coalition and Iran, codenamed "Operation Epic Fury / Roaring Lion", from 2026-02-28 to 03-24. Coverage includes the war progression, the Strait of Hormuz blockade game, Iran's "barefoot outsider outside the system" game framework, RMB settlement challenging US dollar hegemony, Trump's three-layer strategic objectives, the 31st MEU island-seizure plan, the SPR + floating storage crude oil buffer timeline, Polymarket odds and strategy framework (4/30 No + 6/30 Yes positions), Goldman Sachs structural oil price analysis, macroeconomic cascades (Fed no rate cuts / debt spiral / impossible trinity), US energy sector allocation between USO and XLE, four post-war exit scenarios, latest negotiation developments, and the rift between Iranian moderates and the IRGC. This is a reference-class resource / decision deduction template.
SPA reverse API + cookie warmup + bucketed rate-limit bypass + captcha subdomain differences + Playwright/curl hybrid, complete SOP for scraping 1135 2026 funding events + 2530 articles + 9448 images
Listen: 01:08–03:16. Qu first rejects equating the event with the failure of conversational companionship. In his description, model researchers and people continuing the product receive different arrangements. A technical team's decision to enter a large company does not establish that users no longer want character conversations.
Systematized notes from Truman's business pre-judgment course. 15-character core mantra / 7-layer spectrum chart / 76% probability math / 4-stage deliverables / 9-character mindset. When to use it, how to use it, and extracted lessons from pitfalls.
Synthesis of aesthetics / democratization of creation from 3 perspectives. Xiaohongshu vibecoding aesthetic philosophy (models don't matter, aesthetics do + Stanford aesthetics course 5 units + 3 Kimi 2.6 tuning tricks for 1:1 replication + three terms Hero/Shader/Three.js + Pretext 21 million views + 80→100 make-or-break line + 3 tiers of realm)/ xhs-vibecoding-ai-web (3D interaction + AI web pages)/ newtype Dan Koe creator story ($3K → $4.2M) + Trinity Visual Learning. Covers aesthetics as the primary productive force, AI democratizing creation, 3 tiers of realm, and the creator economy
Asked whether AI has a bubble, the opening answer is yes. The ensuing conversation repeatedly distinguishes technical value, expectations inside prices, infrastructure payback, who ultimately receives money and whether founders survive until another financing round.
The official claim concerns Hugging Face’s Model Trending popularity list, not first place in overall performance. Lu post-trained an existing model rather than pretraining a general model from scratch. Improvement on a task does not establish stronger capability everywhere.
Listen: 00:59–09:27. The guest recalls monthly token consumption growing more than 20% over the previous month, without supplying the statistical institution, sample, or consistent measurement definition. The number explains his starting observation; it is not a verified industry growth rate.
A complete cross-author synthesis of agent architectures, orchestration patterns, memory, evaluation and operational responsibility, including the published historical cases and additions through October 21, 2026.
A systematic framework for agent evaluation released by the Anthropic engineering team in 2026. It covers an 8-entry concept glossary, a 3-type grader matrix, the dual-track Capability vs Regression evaluation design, 4 sets of evaluation playbooks for different agent categories (coding / conversational / research / computer use), an 8-step roadmap for building evaluations from scratch, the mathematical intuition behind pass@k vs pass^k, the evaluation framework ecosystem, and common anti-patterns. Phase A anchors the wiki, and sets the density and bilingual style baseline for the anthropic-* series of content.
A research report on AI Agent market investment trends compiled by newtype in January 2026. It contains full 2025 data on VC / sovereign wealth funds / unicorns / application layer / foundation model financing / sub-sectors (coding/healthcare/legal/enterprise/prediction markets). It covers OpenAI's valuation rising from $500 billion to $830 billion / Anthropic from $183 billion to $350 billion / xAI's cumulative funding of $42.1 billion / $87 billion in global GenAI financing / 191 new unicorns / $46 billion from sovereign funds / the "kingmaking strategy" phenomenon of valuations doubling within 27–60 days / single-point data on Cursor / Cognition / Harvey / Mercor and others / analyst views from Gartner / McKinsey / BCG / IDC / 25x revenue multiple / 5x cash burn multiple / Sequoia's three-stage "Agent Economy" thesis / bubble warnings.
A research report on the evolution of AI Agent technology architecture compiled by newtype in January 2026. It is based on papers published from October 2024 to January 2026, official blogs from Anthropic/OpenAI/Google/Microsoft, open-source projects, and real-world deployment cases. It covers 6 major evolution trends / 4 multi-Agent orchestration patterns / Google's 8 foundational patterns / memory management (MemoryOS 3-layer architecture + vector database selection) / tool calling + MCP standardization / reasoning frameworks (ReAct → ToT → GoT → WoT evolution) / Large Reasoning Models (LRM) / 7-layer production architecture / comparison of mainstream frameworks (LangGraph/CrewAI/MAF/AutoGen/LlamaIndex/ADK) / real deployment cases from Morgan Stanley + Klarna + BBVA + Rakuten / 2026 trend forecasts / compliance and governance / production checklist. It serves as a "research report-level" supplement to [[newtype-agent-architecture]] (Huang Yihe's post-level judgments).
Connecting AI to the physical world — self-driving labs for materials and biology (Periodic, Ginkgo), Waymo, AI chips and inference compute (Cerebras, Baseten), space data centers (Starcloud), AI smart stoves — and the cross-domain playbook that actually makes "AI for X" work in the world of atoms
A compilation of themes from 13 podcast episodes — whether foundation models have moats (Evans: no), whether SaaS is dead (Linear pushes back), one-person billion-dollar companies, GTM in the AI era (SEO → agent-led growth + AI research), human-AI division of labor (AI sandwich), full-stack/hyper-leveraged builders, tech geopolitics (Pax Silica), and whether creative AI (Suno/ElevenLabs) can be a real business.
13 episodes of frontier research podcasts reorganized by "debates": LeCun bashing LLMs (dead end vs. the scaling camp) / three camps on world models: JEPA vs. linguistic symbols vs. renderers / whether continual learning is overhyped / whether RL can succeed outside the coding domain / recursive self-improvement via AI building AI / frontier risks and alignment / Hassabis's endgame for AGI / three non-mainstream architectures: EBM, Mistral, Leanstral. Speakers include LeCun, Pachocki, Dubois, Kolter, Manning, Dehghani, Vinyals, Hassabis, Brockman
A collection of long-term views from investors and AI ecosystem observers including garrytan, nikunj, mattturck, swyx, and adityaag on X — the shifting nature of moats (retention > momentum), the thin harness vs. fat skills debate, agent labs / businesses that get stronger as models improve, default-fundable companies, and taste as a moat
