Open Source Project #244: HowToLiveBetter — 767 Pieces of Advice, Each With Cost, Payoff, Evidence Grade, and a Primary Source

Introduction

Today we're not talking tech, we're talking life.

This is article #244 in the "One Open Source Project a Day" series. Today's project isn't a framework or a tool. It's a book: HowToLiveBetter (高性价比人生指南, "the high-value life guide"). It lives on GitHub with 67,897 Stars and 5,863 Forks, its text is published under CC BY 4.0, and it was created on 2026-09-07 — not even two and a half months ago.

Its tagline is "spend the least money, time, and energy to get back the most lifespan, money, and personal freedom." It covers how to live longer, get sick less, and respond when something goes wrong; how to waste less money and avoid scams and lawsuits; what you can claim when you have no job or money, and what paperwork opening a shop, company, or website requires; and also romance, marriage, children, studying abroad, and learning a trade. There are 767 pieces of advice, and each states what it costs you, what it gets back, and how solid the evidence is, citing only journal papers and official documents.

Why does a life-advice repo belong in an open source series? Because it uses "open source" in an unusual place: a knowledge base that can be searched, cited, corrected, translated, and called by AI. It takes issue-tracker corrections, grades its evidence, has build scripts — and ships a skill that lets Claude Code and Codex answer questions from the book.

What You'll Learn

• How the book is organized: 34 topics, 767 entries, and the fixed format of each entry
• How the evidence grades (A/B/C) and the "value tier" are defined, and why the author insists they don't affect each other
• How to use it: online search page, offline HTML, PDF, EPUB, an Anki deck, and the AI skill
• Which engineering practices are worth borrowing from a knowledge-type open source project

Prerequisites

• No technical background is needed to read the content
• To build the ebooks or run the search page locally, basic Node.js and python -m http.server knowledge
• The legal, health-insurance, and social-insurance content follows current mainland China rules; readers elsewhere need to cross-check

Project Background

Overview

This is "a menu of options, not a to-do list." The README says so up front: you don't need to do everything, and picking one or two entries counts; the author hasn't done most of them either. It also warns that content may be out of date or wrong, that many policies are enforced differently from place to place, and that when you actually go handle something, see a doctor, or go to court, you should defer to local authorities, doctors, and lawyers.

Each entry looks like this (the README's example, translated):

### 5. Switch the household salt to low-sodium (potassium) salt
- Cost: a few yuan more per bag. Swap it when you shop; no extra time. Taste barely changes.
- In plain words: for people who've had a stroke or are over 60 with high blood pressure, switching to low-sodium salt lowers five-year death probability by about 12% and stroke by about 14%.
- Payoff: a randomized trial in rural China, 20,995 people ... the low-sodium group had about 12% lower death risk than the regular-salt group (RR 0.88) ...
- Evidence grade: A
- Source: Neal B et al. (2021). Effect of Salt Substitution on Cardiovascular Events and Death. NEJM. https://doi.org/10.1056/NEJMoa2105675
- Note: Contested. The trial subjects were high-risk older people; healthy young people gain much less. People with kidney impairment or on potassium-sparing drugs should ask a doctor first.

Cost, plain-language result, the original numbers, evidence grade, a clickable source, and the limits of applicability — nothing missing.

Author / Team

• Author: eternity4719, a personal project, with reader-maintained translations and derivative tools
• Text license: CC BY 4.0 (covering book/, docs/, and the README text) — you may reproduce, adapt, and use it commercially, but must credit the source, link the license, and state any changes
• Code license: MIT (tools/, skills/, index.html, and .github/)
• Created: 2026-09-07
• How it was written: the README states the book was "written with Claude Code's assistance" — finding primary sources, checking each number against the original text, and maintaining cross-references and build scripts — while topic selection, trade-offs, and final sign-off are the author's responsibility
• Publishing: an official ebook edition is published exclusively by Guomai Culture (果麦文化), going on sale on ebook platforms in October 2026; the web version, EPUB, and PDF remain free

Project Stats

• ⭐ GitHub Stars: 67,897+
• 🍴 Forks: 5,863+
• 📄 License: text CC BY 4.0 / code MIT
• 📅 Created 2026-09-07; last push 2026-10-21
• 📬 Open Issues: 20
• 📚 Scale: 34 topics, 767 entries, 2,531 primary-source links
• 🌍 Translations: English, Russian, Spanish, Portuguese, Arabic, Indonesian, Vietnamese, and more, all maintained by others
• 🌐 Online version: eternity4719.github.io/HowToLiveBetter

Quick Start

Most people don't need to deploy anything. Here are the ways to use it, from least effort up.

1. Use the Online Search Page

Open the online search page and filter by keyword, chapter, or evidence grade, or by the three cost dimensions (does it cost money, how much time, does it need willpower); filters stack. For the lowest-effort list, filter value tier to "very high": 123 entries that cost nothing, take little time, need no willpower, and land in the largest payoff tier. There's also a life self-check: answer a few dozen yes/no questions to see which things in the book you haven't done yet and which to do first.

2. Download an Offline Version

The repo's Release page offers four formats, regenerated automatically after each content update:

• Offline single-file HTML: the whole book with search and filtering in one file; double-click to open, no server or network
• PDF: A4 layout, 200+ pages, with page-numbered contents and bookmarks
• EPUB: send to Kindle with Send to Kindle
• Anki deck: one card per entry, split into subdecks by section; re-importing keeps your review history

3. Let an AI Assistant Answer From the Book

The repo ships a Chinese skill, skills/life-decision-guide, for Claude Code and Codex. Once installed, ask something like "should I co-sign a loan for a friend" or "is a two-hour daily commute worth it." It looks up the relevant entries from the text first, ranks them using the book's accounting approach, and cites which section and entry each comes from. If nothing is found, it says so rather than inventing numbers.

4. Run Your Own Copy

git clone https://github.com/eternity4719/HowToLiveBetter.git
cd HowToLiveBetter
python -m http.server 8000
# open http://localhost:8000/ in a browser

The search page is fully static, with README.md and book/ as its data: no backend, no database, no dependencies. Note that index.html must be opened over http — double-clicking the local file shows a blank page — so use the offline single-file version for that case.

Core Deep-Dive

Every Entry Answers Two Questions: What Does It Cost, What Does It Return

The most distinctive thing about this book is that it writes life advice as a ledger. Every entry must answer: what it costs (money / time / energy / willpower) and what it returns. The returns fall into four resources that are deliberately not converted into each other:

• Lifespan: living longer, dying less from avoidable causes
• Time and energy: not spending time on things with no payoff
• Money: wasting less
• Personal freedom: not landing in detention because you didn't know about a red line

The author states plainly that mortality numbers, time-and-energy numbers, money numbers, and legal consequences are each calculated separately and not converted into one another. "12% lower all-cause mortality" and "saves 500 yuan a year" aren't on the same ruler. That restraint matters; it avoids the false precision of "skip one coffee a day and compound it over decades."

Who benefits is also tiered: you, your spouse and immediate family, friends/colleagues/other relatives, and strangers. The first-aid section should be read through this lens: of 38,227 out-of-hospital cardiac arrests in China, 79.2% happened at home, so learning chest compressions is first of all for the people you live with; rules like "don't jump in if you see someone drowning" and "don't try to break up a fight" are self-protection rules. Whether to help a stranger is left to your own judgment — the book lays out the immunity provisions, self-protective moves, and risks, without framing it as something you must do.

Evidence Grades: A, B, C Only Answer "Can the Number Be Trusted"

A further 88 entries are marked "contested," with opposing evidence listed for contested A/B entries. Sources cite only primary literature (journal papers with DOI or PubMed links, or official reports from WHO, CDC, the National Bureau of Statistics, and similar), not secondary retellings. For cases officials don't publish, a few entries cite bylined original news reporting with a web archive link, and the note says "media reports only."

The author also adds a reminder that's often overlooked: grade A only says there is a concrete, checkable figure — not that the figure is causal. Grade A includes randomized trials but also studies that only follow and record people without grouping them, where you can't tell whether the thing worked or whether the people who did it were simply healthier. Check the payoff field to see which. For legal and policy entries, grade A means the entry cites the statute or official document itself.

How to read percentages is addressed too: most are "some fraction lower than people who didn't do it," not "N fewer per hundred people." The same 12% reduction prevents far more deaths among high-risk older people than among low-risk young people — so first check who the "in plain words" line is about, then the percentage. A 41-term glossary (HR, RR, 95% CI, RCT, meta-analysis, confounding, reverse causation, and more) closes the guide, and on the search page hovering over dotted-underline terms pops up an explanation.

Value Tier: The Author Admits It's Judgment, Not Evidence

Evidence grade doesn't answer "is it worth doing," so each entry also carries a "payoff magnitude" (large/medium/small) and a "scope" (lifespan / money / time-and-energy / personal freedom). Together with three cost dimensions, the search page computes a value tier:

The magnitude has predefined cutoffs rather than gut feeling: for lifespan entries, the percent reduction (≥20% is large); for money, the amount (ten-thousand-yuan scale is large); for personal freedom, the consequence avoided (avoiding criminal liability is large); for time and energy, how much is saved (hours per day is large).

The most honest line is: "this tier is the author's own judgment, not evidence," and by the book's own standard it only counts as grade C. It's a separate matter from the evidence grade — an entry can be grade A yet only "average" value (the shingles vaccine has a 97.2%-efficacy phase-3 RCT, but two doses cost several thousand yuan and shingles rarely kills), or grade C yet "very high" value (a lock-screen passcode on your phone, a PIN on your SIM card). The large middle tier is deliberate too: the magnitude only has three levels, and slicing finer would be fake precision.

Engineering Practices of a Knowledge-Type Open Source Project

The engineering side of this repo is worth a look, since maintainability of knowledge bases is often neglected:

• Data is the text: the search page is fully static and reads the Markdown in README.md and book/ directly, so when the text changes the page changes — no second copy to sync
• Multi-format builds from one source: EPUB, offline HTML, PDF (needs pandoc ≥ 3.1 and typst ≥ 0.13), and the Anki deck (Node ≥ 22.13) are all generated by scripts from the same text, with GitHub Actions rerunning them after content changes and updating the Release under fixed download links
• Public verification records: how each source was checked is recorded in docs/核实记录/, and the README reports "0 entries marked TODO pending verification"
• The split was forced: the text is split into 34 files because a single file exceeded GitHub's 512 KB Markdown rendering limit, causing later sections not to display
• Entries point to each other: on the search page, references like "see section 8, entry 17" have a dotted underline, and clicking shows the target entry's title and plain-words summary in place
• Derivative ecosystem: reader-made sites, apps, mini-programs, browser extensions, and datasets are listed in 衍生项目.md; translations are maintained by their own maintainers and defer to the Chinese original in the repo

Resources

• 🌟 GitHub: eternity4719/HowToLiveBetter
• 🔎 Online search page: eternity4719.github.io/HowToLiveBetter
• ✅ Life self-check: selfcheck.html
• 🤖 AI skill: skills/life-decision-guide
• 📦 Downloads: PDF · EPUB · Offline HTML · Anki deck
• 🌍 English translation: book.dlgrv.com/en
• 🐛 Corrections: Issues

Summary

This project doesn't teach a technology. It demonstrates how to write life experience as knowledge that can be checked: every entry has a cost, a payoff, an evidence grade, a source, and limits; errors can be filed as issues; and updates reach readers through fixed download links.

Three things worth noting:

Evidence grade and value tier are kept strictly separate, and the latter is openly called a judgment. The usual flaw in "tips collections" is blurring "someone said so" with "evidence supports it." This book uses A/B/C for "can the number be trusted," uses the value tier for "is it worth doing," and says the second only counts as grade C itself. That self-discipline is the main reason it's more credible than similar content.

It treats a knowledge base like software engineering. A single data source, automatic multi-format builds, public verification records, reader corrections and translations, and a skill that lets AI assistants answer from the book while citing entries — these practices are a useful reference for anyone turning documentation, handbooks, or an internal knowledge base into something alive.

Read it with appropriate skepticism. This is a personal project written with Claude Code's assistance; statutes, subsidy standards, and deadlines change at any time, and many policies vary by locality. The author reminds readers to defer to local authorities, doctors, and lawyers when it counts. I also haven't checked whether all 767 entries' figures match their cited sources — I can only say the author provides checkable sources. Treat it as a map of what to look up, not a final answer.

If you want a clearly structured Chinese-language life reference whose claims you can trace, or you're thinking about how to turn a pile of knowledge into a searchable, reusable open source project, this repo is worth a look.

Explore PrimeSkills — a curated marketplace of AI agents and skills, each validated against real enterprise workflows. No hype, just what actually works.

Visit my personal site for more insights and interesting products.