把整本书压成一句话：问题从来不是「ChatGPT 能不能」，是「这件事该开哪一个产品」。同一份订阅，选择器挑节奏、Projects 攒上下文、Deep Research 查、Work 交成品、agent 跑网页、Codex 写代码 —— 选对入口，比追更贵的档位重要。The whole book in one line: the question was never can ChatGPT, but which product this task belongs to. One subscription — the picker sets the pace, Projects holds context, Deep Research verifies, Work delivers, the agent acts on the web, and Codex writes code. The right entrance matters more than a pricier tier.
把整本书压成一句话：问题从来不是「ChatGPT 能不能」，是「这件事该开哪一个产品」。The whole book in one line: the question was never can ChatGPT, but which product this task belongs to.
— I
同一份订阅，几种节奏One Subscription, Several Rhythms.
同一份订阅，铺在不同入口上，每个入口对应一种节奏。One subscription, spread across different entrances, each entrance a different rhythm.背后不是十一个互不相干的世界，而是同一份订阅把不同工具排在不同入口上。1注 1Note 1ChatGPT 套餐与价格 —— 一份订阅横跨 Chat、Work、Plugins、Codex 等多个产品面；档位决定模型、额度与管控，不决定你会不会用对入口。截至 2026-07-10。ChatGPT Plans & Pricing — one subscription spans Chat, Work, Plugins, Codex, and more; a tier sets models, quota, and controls, not whether you reach for the right entrance. As of 2026-07-10.把前面每一章压成一行「遇到 X → 开 Y」：Behind it all are not eleven unrelated worlds, but one subscription arranging different tools at different entrances.1注 1Note 1ChatGPT 套餐与价格 —— 一份订阅横跨 Chat、Work、Plugins、Codex 等多个产品面；档位决定模型、额度与管控，不决定你会不会用对入口。截至 2026-07-10。ChatGPT Plans & Pricing — one subscription spans Chat, Work, Plugins, Codex, and more; a tier sets models, quota, and controls, not whether you reach for the right entrance. As of 2026-07-10. Each earlier chapter, compressed to one line of when-X-open-Y:
同一题答得慢或不够准Answers slow or not sharp enough
去选择器换 Medium / High / Extra High 或 Pro，而不是反复重问。Switch to Medium / High / Extra High or Pro instead of re-asking.
每次都在重贴背景Re-pasting context every time
交给自定义指令、Memory、或一个 Project。Hand it to custom instructions, Memory, or a Project.
从 Plugin 找到对应 App，别再手动搬运。Find the right App through its Plugin; stop hauling it by hand.
要一份带出处的报告Need a sourced report
交给 Deep Research，回来抽查引用。Hand it to Deep Research, spot-check the citations.
要文档、表格、演示或 Site 成品Need a document, spreadsheet, presentation, or Site
交给 Work，按交付物验收。Hand it to Work and review the deliverable.
要它替你在网页上动手Need it to act on the web
交给 agent，盯着它跑。Hand it to the agent, watch it run.
要在仓库里改代码Need to change code in a repo
从 Canvas 起步，重活请 Codex。Start at Canvas; for heavy work call Codex.
要把能力接进自己的程序Need to put the capability in your own app
去 API / platform，不要把订阅当 token 池。3注 3Note 3OpenAI 帮助中心 · ChatGPT 订阅与 API 分开管理 —— ChatGPT 订阅不附带 API 额度，API 在 platform.openai.com 单独按 token 计费。OpenAI Help · ChatGPT subscriptions and the API are managed separately — a ChatGPT subscription does not include API credit; API usage is billed separately per token on platform.openai.com.Use the API / platform; do not treat the subscription as a token pool.3注 3Note 3OpenAI 帮助中心 · ChatGPT 订阅与 API 分开管理 —— ChatGPT 订阅不附带 API 额度，API 在 platform.openai.com 单独按 token 计费。OpenAI Help · ChatGPT subscriptions and the API are managed separately — a ChatGPT subscription does not include API credit; API usage is billed separately per token on platform.openai.com.
— II
选对入口，比追档位重要The Right Entrance Beats Chasing a Tier.
多数人卡住，不是因为档位太低，是因为所有活都塞进了同一个框。Most people are stuck not because their tier is too low, but because every task goes into the same box.能力早就在那份订阅里了。升到更贵的档，解开的是模型、额度和管控，不是判断力 —— 而判断力才是这本书想给你的：遇到一件活，先问它属于哪个入口，再动手。2注 2Note 2OpenAI 帮助中心 · GPT-5.6 in ChatGPT —— 截至 2026-07-10，GPT-5.5 Instant 仍负责默认快答，GPT-5.6 Sol 驱动 Medium / High / Extra High，Sol Pro 驱动 Pro；Terra / Luna 不可在标准 Chat 中选择。OpenAI Help · GPT-5.6 in ChatGPT — as of 2026-07-10, GPT-5.5 Instant still handles default fast responses, GPT-5.6 Sol powers Medium / High / Extra High, Sol Pro powers Pro, and Terra / Luna are not selectable in standard Chat.同一段需求，塞进 Instant 反复重问、在 High 深想、让 Work 交成品，或让 agent 去网页执行，是四种不同工作方式。差的从来不只是你付得够不够多，是你有没有走对那扇门。The capability was already in that subscription. Upgrading to a pricier tier unlocks models, quota, and controls, not judgment — and judgment is what this book set out to give you: faced with a task, first ask which entrance it belongs to, then act.2注 2Note 2OpenAI 帮助中心 · GPT-5.6 in ChatGPT —— 截至 2026-07-10，GPT-5.5 Instant 仍负责默认快答，GPT-5.6 Sol 驱动 Medium / High / Extra High，Sol Pro 驱动 Pro；Terra / Luna 不可在标准 Chat 中选择。OpenAI Help · GPT-5.6 in ChatGPT — as of 2026-07-10, GPT-5.5 Instant still handles default fast responses, GPT-5.6 Sol powers Medium / High / Extra High, Sol Pro powers Pro, and Terra / Luna are not selectable in standard Chat. Re-asking in Instant, thinking deeper in High, having Work produce a deliverable, or sending the agent to act on the web are four different ways of working. What is missing is not merely how much you pay, but whether you walked through the right door.
— III
让每个产品认领一件事Let Each Product Claim One Task.
你不必用上全部，只要让你常做的那几件活，各自找到归属的产品。You need not use all of it — just let the few things you do often each find the product they belong to.回到引子那张五行表。读完之后，第二列不该再五行都写着「默认聊天」。这本书的成功标准很朴素：一个月后，你常做的活里至少有两三件，已经稳定地搬到了对的入口上。别想着一次推翻全部习惯 —— 从一件活开始改就够。Go back to the five-row table from the intro. After reading, the second column should no longer read default chat on all five rows. This book's measure of success is plain: a month from now, at least two or three of the things you do often have moved, for good, to the right entrance. Do not try to overturn every habit at once — changing one task is enough to start.最后一次照镜子：One last look in the mirror:
回到引子那张五行表，把第二列改成读完后的判断。Return to the intro's five-row table and rewrite the second column with your post-reading judgment.
挑一件这周就改用新入口去做的活，真的换过去。Pick one task to move to a new entrance this week, and actually move it.
一个月后再回来数：还有几行写着「默认聊天」。Come back in a month and count: how many rows still say default chat.
广度是会员的税,
挑对才是会员的本事.
Breadth is the membership's tax,
choosing right is its craft.