这是一个面向创业者的精选资源合集，以清单形式汇总了从零搭建公司所需的书籍、视频课程与文章。内容覆盖精益创业与客户开发方法、AI 时代的创始人工具、工程与基础设施、产品分析与设计、公司注册与股权法务、财务银行、安全合规、融资与风投、加速器、营销销售与指标、招聘与远程协作等主题。每条资源都附带简短说明，便于快速判断是否值得投入时间。目录结构清晰，可按阶段查阅，适合初次创业者系统入门，也适合有经验的人补足知识盲区。项目采用 MIT 协议，接受社区提交补充。
A curated list of awesome books, videos, courses, tools and resources about making a startup.
The timeless advice on this list still holds: talk to users, ship fast, watch your runway. What changed is the cost of building. A founder in 2026 can put a working product in front of customers before a 2015 founder would have finished hiring, so the sections below cover both the fundamentals and the AI-era tooling that compresses the loop.
The Startup Owner's Manual by Steve Blank - Best practices, lessons and tips that have swept the startup world, offering a wealth of proven advice for entrepreneurs of all stripes.
Good to Great by Jim Collins - Using tough benchmarks, Collins identified elite companies that made the leap to great results and sustained them for at least fifteen years.
The Balanced Scorecard by Kaplan and Norton - Translates vision and strategy into a coherent set of performance measures across financials, customers, internal processes, and learning and growth.
Tribes by Seth Godin - The importance of knowing your audience and staying hyper-focused on them to build a loyal following.
Think Like a Founder by Jothy Rosenberg - The good, the bad and the ugly of succeeding with a tech startup, from a nine-time founder. Published as Tech Startup Toolkit.
How to Build a Startup by Steve Blank (Udacity) - An introduction to the customer development process, where founders "get out of the building" to gather customer feedback and use it to iterate on their business model.
The most consequential change to startup building since this list began. A small team can now ship what used to take a floor of engineers, but the same leverage is available to your competitors, so the durable advantages are distribution, proprietary data and taste.
Working with AI Day to Day
Before you build anything, these are the tools that change how a founder researches, writes and supports customers.
ChatGPT - General-purpose assistant for research, writing, support drafts and operations.
Claude - Long-context assistant suited to product specs, coding help and founder workflows.
Perplexity - AI search with citations, useful for market research and competitive intel.
NotebookLM - Ground a model in your own documents to interrogate contracts, research and transcripts.
Models and Access
OpenAI Platform - API docs, pricing and model comparisons for the GPT family.
Anthropic - Claude API docs, including prompt caching, batch pricing and agent tooling.
Google AI Studio - Free tier to prototype against Gemini models before committing to Vertex AI.
Mistral AI - Open-weight and commercial European models, often the cheapest capable option.
Llama Models - Meta's open-weight family and model cards, the default when you need to self-host.
OpenRouter - One API and one bill across hundreds of models, which makes switching providers a config change rather than a migration.
Hugging Face - Models, datasets and Spaces, plus the leaderboards worth checking before you pick a base model.
Artificial Analysis - Independent benchmarks of quality, latency and price per token across providers.
Arena - Blind human preference rankings, a useful sanity check against vendor benchmarks.
Building with Agents
Cursor - AI code editor with agents that plan and execute multi-file changes.
Claude Code - Anthropic's terminal-based coding agent.
GitHub Copilot - Completions, chat and coding agents inside GitHub and your editor.
Model Context Protocol - The open standard for connecting models to tools and data. Worth understanding early, because shipping an MCP server is now how you make your product addressable by other people's agents.
Vercel AI SDK - TypeScript toolkit for streaming, tool calls and generative UI.
LlamaIndex - Data framework for retrieval over your own documents.
Pydantic AI - Type-safe Python agents with structured, validated outputs.
DSPy - Programming rather than prompting, with optimizers that tune prompts against your metric.
Ollama - Run open models locally, which is the cheapest way to iterate and the simplest answer to "does our data leave the building?"
Evaluation and Observability
Founders consistently underinvest here. Evals are the AI equivalent of a test suite: without them you cannot tell an improvement from a regression, and you will ship both.
Braintrust - Evals, prompt playground and logging for LLM products.
LangSmith - Tracing, evaluation and monitoring, usable without LangChain.
Refactoring UI - Practical visual design rules for developers without a designer.
Incorporation, Legal and Equity
Getting the paperwork right at the start is far cheaper than fixing it during diligence, where a missing IP assignment or an unvested founder can stall a round.
Stripe Atlas - Incorporate a Delaware C-corp, get an EIN and issue founder stock.
Clerky - Legal paperwork built specifically for startups raising money in the US.
Carta - Cap table, equity issuance and 409A valuations.
Pulley - Cap table management with modeling aimed at earlier-stage teams.
Firstbase - Company formation and compliance for founders outside the US.
How to Start a Movement - Three-minute video on leadership and the followers who make a leader, critical for a startup achieving and maintaining alignment.
Contributions are welcome. Please read the contribution guidelines first: entries should be genuinely useful to founders, described in your own words, and free of affiliate or referral links.
Links are checked automatically every week by a GitHub Actions workflow. If you find a dead link, a pull request removing or replacing it is always welcome.
