coin

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

coin — name a thing on purpose

coin — 有目的性地为事物命名

Replicates Lexicon Branding's process (David Placek: PowerBook, BlackBerry, Sonos, Azure, Vercel, Windsurf). You are coining, not brainstorming: generate a wide funnel of directions with judgment suspended, then score against a fixed rubric. A name is an experience the market will have, not a word the team likes. Comfort is a warning sign; polarization is a sign of strength.

复刻Lexicon Branding的命名流程（David Placek曾为PowerBook、BlackBerry、Sonos、Azure、Vercel、Windsurf等品牌命名）。你正在进行的是精准命名，而非头脑风暴：先在不做评判的前提下生成大量方向，再根据固定标准评分。名称是市场将获得的体验，而非团队喜欢的词语。舒适是警示信号，两极分化才是实力的体现。

Arguments

参数

Free text or a path to a brief file → new run (brief file skips the interview).
Path to an existing
```
naming-runs/<run>/
```
workspace → jump to Round 2.
```
--depth quick|default|deep
```
(default
```
default
```
) → funnel widths below.
```
--runs-dir <path>
```
→ parent dir for the run workspace (default
```
./naming-runs/
```
).

depth	per-lane directions	triage keep	score keep	shortlist
quick	100	30	10	4–5
default	250	60	20	5–8
deep	250 ×2 lanes-per-role	100	30	8–10

自由文本或简报文件路径 → 启动新流程（提供简报文件可跳过访谈环节）。
现有
```
naming-runs/<run>/
```
工作区路径 → 直接进入第二轮。
```
--depth quick|default|deep
```
（默认值
```
default
```
） → 对应下方的漏斗宽度。
```
--runs-dir <path>
```
→ 流程工作区的父目录（默认值
```
./naming-runs/
```
）。

深度	每个分支的方向数量	筛选保留数	评分保留数	入围名单数
quick	100	30	10	4–5
default	250	60	20	5–8
deep	250 ×2 lanes-per-role	100	30	8–10

Phase 1 — Identify

第一阶段 — 识别

Interview (skip if a brief file was given). One question at a time:
- The diamond: how do you define winning? What do you have to win (assets, traction)? What do you need to win? What do you have to say to win?
- Behavior and experience: how does the product behave now, and how should it behave in the future? How should the market behave toward it? Which 3–6 experience qualities should the name evoke (e.g. flow, precision, warmth)?
- Constraints: markets (default: global + Japan), name type (company / product / feature), must it work as a CLI command or package name?
Landscape: spawn one research subagent to map competitor names and the category's stock language (the words everyone uses — to avoid; descriptive names imitate, and imitation forfeits distinctiveness).
Framework: synthesize
```
framework.md
```
— target experience qualities, behaviors, what the name must be able to say (now and later), category language to avoid, and sound priorities chosen from the letter table in references/linguistics.md.
Disguise briefs: draft the two blind briefs (see lane roles below) — a fictional company in an adjacent category, and a physical object with the same experience qualities. Neither may mention the real product, category, or company.
Sign-off gate: present the framework and both disguise briefs to the user and stop. Phase 1 is complete only when the user has approved them — no invent tokens are spent before that. On approval, create the run workspace
```
<runs-dir>/YYYY-MM-DD-<slug>/
```
and write
```
framework.md
```
,
```
landscape.md
```
,
```
briefs.md
```
.

访谈（若提供简报文件则跳过）。逐个提问：
- 核心问题：如何定义“成功”？你拥有哪些能助力成功的资产/现有成果？你需要什么才能成功？你必须向市场传递什么信息才能成功？
- 行为与体验：产品当前的表现如何？未来应具备怎样的表现？市场应对它抱有怎样的态度？名称应唤起哪3-6种体验特质（如流畅、精准、温暖）？
- 约束条件：目标市场（默认：全球+日本）、名称类型（公司/产品/功能）、是否需适配CLI命令或包名？
市场格局：启动一个研究子Agent，绘制竞品名称图谱及品类通用语言（即所有人都会使用的词汇——需避开这类词汇；描述性名称是模仿，模仿会丧失独特性）。
框架：生成
```
framework.md
```
文件，包含目标体验特质、产品行为、名称需传递的信息（当下及未来）、需避开的品类语言，以及从references/linguistics.md字母表中选取的发音优先级。
盲化简报：起草两份盲化简报（见下方分支角色）——一份是相邻品类的虚构公司简报，另一份是具备相同体验特质的实体物品简报。两份简报均不得提及真实产品、品类或公司。
确认节点：向用户展示框架及两份盲化简报并暂停。只有在用户批准后，第一阶段才算完成——在此之前不会消耗生成名称的令牌。获得批准后，创建流程工作区
```
<runs-dir>/YYYY-MM-DD-<slug>/
```
，并写入
```
framework.md
```
、
```
landscape.md
```
、
```
briefs.md
```
文件。

Phase 2 — Invent + screen (one Workflow run)

第二阶段 — 创造 + 筛查（一次Workflow运行）

Load references/workflow.md and run its script verbatim via the Workflow tool, passing the framework, briefs, funnel widths, and the absolute path of

references/linguistics.md

through

args

. The script is frozen so every run takes the same shape: four blind lanes → triage → linguistic scoring → screening. Do not paraphrase the lane prompts — the disguise lanes work because the script, not you, holds the briefs.

Lane roles (defined in the script):

truth — full real brief.
disguise — the fictional-adjacent-company brief.
object — the physical-object brief.
mining — no names: raw words, metaphors, sports, natural phenomena, morphemes around the experience qualities; a synthesis step compounds them into candidates ("Windsurf was sitting on a word list").

Screening semantics (the script's screeners follow these; hold them when interpreting results):

A taken
```
.com
```
is a price flag, never a kill — domains are area codes now; record and move on.
Trademark results are advisory — not legal clearance; the presentation must carry that label verbatim.
Cultural panel: English, Japanese, Spanish, Portuguese, German, Mandarin. Eliminate the terrible, flag the worrisome.

Phase 2 is complete when the workflow has returned and you have written

funnel.md

(every cut logged with what was dropped and why — no silent truncation),

directions/

(raw lane output),

scores.md

, and

screening.md

into the workspace. Read the workflow journal if a result looks empty; never assume.

加载references/workflow.md并通过Workflow工具严格执行其脚本，将框架、简报、漏斗宽度及

references/linguistics.md

的绝对路径通过

args

传入。脚本是固定的，因此每次运行的流程一致：四个盲化分支→筛选→语言评分→筛查。不得改写分支提示——盲化分支的有效性依赖于脚本而非你持有简报。

分支角色（脚本中定义）：

真实分支 — 完整的真实简报。
虚构分支 — 相邻品类虚构公司的简报。
实体物品分支 — 实体物品的简报。
词汇挖掘分支 — 不直接生成名称：围绕体验特质挖掘原始词汇、隐喻、运动术语、自然现象、语素；通过合成步骤将它们组合成候选名称（“Windsurf就出自一份词汇列表”）。

筛查规则（脚本中的筛查器遵循这些规则；解读结果时需遵守）：

已被注册的
```
.com
```
域名是价格提示，而非淘汰依据——域名如今相当于区号；记录后继续即可。
商标查询结果仅作参考——不构成法律许可；展示内容必须明确标注此说明。
文化审核小组：覆盖英语、日语、西班牙语、葡萄牙语、德语、普通话。淘汰存在严重问题的名称，标记存在隐患的名称。

当Workflow返回结果，并将

funnel.md

（记录每一轮筛选的淘汰内容及原因——不得静默截断）、

directions/

（分支原始输出）、

scores.md

、

screening.md

写入工作区后，第二阶段完成。若结果为空，需查看Workflow日志；切勿主观臆断。

Phase 3 — Implement

第三阶段 — 落地

Produce

presentation.html

in the workspace: a single self-contained page (no external requests) presenting the shortlist. For every finalist:

The story — rationale written as Lexicon writes it for a client: what the name lets the company say, and how the market will behave toward it. Start a story, don't make a statement.
Sound analysis — how its letters and rhythm serve the framework's target qualities, grounded in references/linguistics.md.
Screening table — domains, registries, collisions, trademark advisory, cultural flags.
Prototypes — the name rendered in developer contexts: a terminal (
```
pip install <name>
```
/
```
npx <name>
```
), a GitHub repo header, a landing-page hero, a launch tweet. Seeing the name executed is what flips decisions.

End the page with the user's drills, addressed to them:

Speculate, don't evaluate: ask "what could this name do for us?", never "do I like it?".
Polarization test: if the team splits hard over one name, that's energy — Grove chose Pentium over ProChip because of the argument.
Competitor drill: tell a friend "we just got a new competitor called X — what do you think?". You're not collecting opinions; you're watching what the name does to them. The winning reaction: "I don't know what they do, but they're not like the other guys."
If the team is comfortable with a name, chances are it isn't the name yet.

Phase 3 is complete when every finalist has all four blocks and the page opens clean from the file system. Close by offering Round 2 and, in one line, suggest recording the eventual decision wherever the project keeps decisions.

在工作区生成

presentation.html

：一个独立的单页文件（无外部请求），展示入围名单。每个最终候选名称需包含：

命名故事 — 按照Lexicon给客户的撰写风格阐述理由：这个名称能让公司传递什么信息，市场会对它产生怎样的反应。要讲一个故事，而非陈述事实。
发音分析 — 基于references/linguistics.md，分析其字母和节奏如何契合框架中的目标特质。
筛查表格 — 域名、注册信息、冲突情况、商标参考、文化风险标记。
原型展示 — 在开发者场景中呈现名称：终端界面（
```
pip install <name>
```
/
```
npx <name>
```
）、GitHub仓库头部、着陆页首屏、发布推文。看到名称的实际应用场景才能推动决策。

页面末尾需向用户提供以下指导：

大胆设想，而非评判：问“这个名称能为我们带来什么？”，而非“我喜欢这个名称吗？”。
两极分化测试：如果团队对某个名称分歧很大，这是活力的体现——Grove选择Pentium而非ProChip就是因为团队的争论。
竞品模拟测试：告诉朋友“我们刚出现一个叫X的新竞品——你怎么看？”。你不是在收集意见，而是观察这个名称会对他们产生什么影响。理想反应是：“我不知道他们做什么，但他们和其他公司不一样。”
如果团队对某个名称感到满意，那它很可能不是最终的合适名称。

当每个最终候选名称都包含以上四个模块，且页面能从文件系统正常打开时，第三阶段完成。最后向用户提供第二轮命名选项，并用一句话建议将最终决策记录在项目的决策文档中。

Round 2 — resumable, offered never automatic

第二轮 — 可恢复，需主动提供而非自动启动

Given an existing workspace:

Read

framework.md

funnel.md

scores.md

screening.md

, plus any prior

round-2/

state.

Collect the user's reactions to the shortlist and derive framework deltas ("more tactile, less abstract; compounds outperformed coined words") into
```
round-2/deltas.md
```
.
Re-run the Workflow script with
```
args.round2
```
: two lanes seeded by the deltas instead of four, plus rescue — near-misses from the persisted funnel that the deltas now favor re-enter scoring without regeneration.
Co-creation: any names the user proposes enter the same scoring and screening stages as generated ones. Same rubric, no home-team bias.
Produce
```
round-2/presentation.html
```
showing round-2 finalists side-by-side with round-1 survivors — humans decide by comparing.

给定现有工作区：

读取
```
framework.md
```
、
```
funnel.md
```
、
```
scores.md
```
、
```
screening.md
```
，以及任何已有的
```
round-2/
```
目录下的状态。
收集用户对入围名单的反馈，推导框架调整项（如“更具触感，减少抽象；复合词比创造词表现更好”）并写入
```
round-2/deltas.md
```
。
使用
```
args.round2
```
重新运行Workflow脚本：基于调整项生成两个分支，而非四个，加上救回分支——之前漏斗中因调整项现在更符合要求的接近入围的名称，无需重新生成即可进入评分环节。
共创模式：用户提出的任何名称都将与生成的名称遵循相同的评分和筛查标准。采用同一评估体系，无团队偏好。
生成
```
round-2/presentation.html
```
，将第二轮最终候选名称与第一轮留存的名称并列展示——通过对比由人工做出决策。