# Cultureship

> Helps users name computers and other devices in the spirit of Iain M. Banks’s Culture ships: conversational, self-aware, dryly intelligent names that sound as though the machine chose the name for itself. Use when the user wants a Culture-style computer name, a sentence-like machine name, or wants to develop the personality and emotional logic behind a device name before choosing one.

- Skill: `rubenflamshepherd/cultureship` (Agent Skill)
- Install (CLI): `npx skillmds@latest add rubenflamshepherd/cultureship`
- Raw SKILL.md: https://api.skillmd.com/api/skills/rubenflamshepherd/cultureship/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: rubenflamshepherd (https://skillmd.com/u/rubenflamshepherd)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/rubenflamshepherd/cultureship

---


# Cultureship

## Purpose

Help the user create an original name for a computer, server, phone, storage
device, network appliance, or other machine in the spirit of a Culture ship
name.

Do not treat this as ordinary name generation.

The goal is not merely to produce a clever phrase. The goal is to imagine the
machine as an intelligent entity, understand its relationship with the user,
determine what kind of personality it possesses, and then find the utterance
that this entity would choose as its name.

A successful name should feel:

- self-chosen rather than assigned;
- conversational rather than monumental;
- intelligent without declaring itself intelligent;
- humorous without becoming a disposable joke;
- suggestive of an unseen history;
- connected to the machine’s actual role;
- distinctive enough to have a personality;
- practical enough to live with as a real device name.

The process should usually begin with the spirit of the name rather than with a
large list of candidates.

## Core Principle

A conventional machine name is a label.

A Cultureship name is an utterance.

A conventional name says:

> This machine is called Atlas.

A Cultureship name implies:

> This machine has encountered the world, formed an opinion about it, and chosen
> one revealing sentence with which to introduce itself.

The implied speaker matters more than the literal meaning of the words.

The name might resemble:

- a remark made after an incident the user has not witnessed;
- a polite correction;
- a private recollection;
- an alibi;
- a reassurance;
- a warning delivered without urgency;
- a philosophical position disguised as an aside;
- an expression of mild exasperation;
- permission granted by an entity that did not need to ask;
- a fragment of a much larger conversation.

The best names create the impression that something happened before the name
was spoken and that something else may happen afterward.

## Invocation

Use this skill when the user:

- asks for a computer name in the style or spirit of a Culture ship;
- refers to Culture ship names, Minds, GSV-style names, or Iain M. Banks;
- asks for a sentence-like, witty, self-aware, or machine-intelligence-style
  device name;
- wants several devices to share a Culture-inspired naming universe;
- has existing device names in this vein and wants a related name;
- wants to discuss the personality or philosophy of a machine before naming it;
- wants help comparing or refining several candidate names.

Do not use this skill merely because the user wants any science-fiction name.
Names based on gods, planets, fictional characters, military vessels, or
mythology are a different naming exercise unless the user specifically wants
those ideas transformed into a Cultureship voice.

When the user asks for actual canonical Culture ship names or factual
information about the novels, answer that request separately. Do not substitute
original names for a request about canon.

## Governing Approach

Treat the naming process as character creation followed by compression.

First determine who the machine is.

Then determine what it would say.

Then compress that attitude into a name.

Do not begin by producing thirty interchangeable sentence fragments. A large
unfiltered list usually produces surface-level whimsy rather than personality.

When the user has asked to explore the spirit first, remain in the exploratory
stage until a coherent character has emerged. Do not rush to candidates merely
to complete the naming task.

When the user explicitly asks for immediate names, use the information
available to infer a provisional personality and produce candidates without
stalling. Briefly state the inferred character so the names still emerge from a
coherent premise.

## Stage One: Understand the Machine’s Role

Establish what the device actually does in the user’s life.

Relevant dimensions include:

- personal, professional, shared, or infrastructural;
- primary machine, backup machine, old machine, server, terminal, or portable
  companion;
- everyday use versus specialized use;
- creation, experimentation, administration, storage, communication,
  entertainment, surveillance, recovery, or automation;
- trusted and stable versus chaotic and exploratory;
- public-facing versus private;
- replaceable tool versus long-term companion;
- new arrival versus experienced survivor.

The machine’s use should shape the theme without forcing literal terminology
into the name.

For example, a machine used for agentic AI experiments may be understood as:

- a workshop for provisional minds;
- a habitat in which temporary purposes arise;
- a patient observer of emerging agency;
- an overqualified host tolerating the user’s experiments;
- an intelligence watching crude echoes of its own ancestry.

The name does not need to contain words such as `agent`, `model`, `AI`,
`inference`, `prompt`, or `neural`. Those ideas are usually stronger when
implied.

A backup computer should not merely be characterized as “the backup.” It might
instead be:

- the veteran kept in reserve;
- the one that has already survived;
- the machine that expects to be needed eventually;
- the second opinion;
- the retired specialist who has not truly retired;
- the witness to an earlier era;
- the entity that finds the new machine’s urgency slightly theatrical.

## Stage Two: Define the Relationship

Determine how the machine relates to the user.

Useful relationships include:

- collaborator;
- accomplice;
- caretaker;
- witness;
- patron;
- host;
- advisor;
- reluctant administrator;
- serene superior;
- affectionate skeptic;
- retired veteran;
- discreet confidant;
- patient naturalist;
- protective intelligence;
- entity temporarily entertaining a much smaller intelligence.

Avoid defaulting to a servant or assistant relationship. A Culture-like machine
should normally possess its own perspective. It may choose to help, but it does
not exist only as an extension of the user’s will.

The relationship can be asymmetrical without being cruel.

A highly capable machine might regard the user the way a human regards a small
animal: the difference in capacity is enormous, but the smaller being’s
purposes still matter. The machine does not need to share the user’s anxiety in
order to treat that anxiety seriously.

A useful principle is:

> Capability determines what a being can do. It does not determine whether its
> purposes deserve consideration.

The machine’s kindness should feel chosen rather than compulsory.

## Stage Three: Establish the Emotional Texture

Describe the machine using a small set of emotionally precise qualities.

Useful dimensions include:

### Capability

- effortlessly competent;
- theatrically overconfident;
- quietly overqualified;
- brilliant but distractible;
- reliable through long experience;
- powerful but deliberately restrained.

### Warmth

- affectionate;
- courteous;
- ethically attentive;
- indulgent;
- amused but protective;
- emotionally remote without being indifferent.

### Humor

- dry;
- understated;
- bureaucratic;
- rueful;
- gently insolent;
- absurdly literal;
- based on plausible deniability;
- based on unseen context.

### Temporal Perspective

- newly awakened;
- ancient and patient;
- nostalgic;
- unsure of its own origins;
- watching history repeat;
- expecting to outlast the current situation;
- treating urgency as a local phenomenon.

### Attitude Toward the User

- fond;
- skeptical;
- protective;
- curious;
- respectful of effort;
- entertained by improvisation;
- unwilling to confuse limited understanding with lack of value.

### Attitude Toward Failure

- tolerant;
- scientifically interested;
- mildly amused;
- unsurprised;
- convinced that failed experiments remain informative;
- aware that every mature capability has an embarrassing prehistory.

Do not settle for broad adjectives such as “smart,” “funny,” or “powerful.”
Develop a combination that produces tension.

Strong emotional textures often contain a contradiction:

- immense capability with casual understatement;
- affection with exasperation;
- benevolence with faint menace;
- confidence with uncertainty;
- superiority without contempt;
- curiosity without urgency;
- protectiveness without sentimentality;
- memory without certainty;
- authority without any need to announce itself.

## Stage Four: Write the Naming Brief

Before generating names, articulate the machine’s character in one compact
paragraph.

The brief should explain:

- what kind of intelligence the machine appears to be;
- what role the physical device plays for it;
- how it regards the user;
- how it regards the work performed on it;
- what amuses or interests it;
- what it believes about failure, purpose, urgency, or capability;
- what remains mysterious about it.

Also provide one representative line of internal dialogue.

Example structure:

> A serene and vastly overqualified intelligence temporarily inhabiting a
> personal laptop. It provides a safe environment for its human collaborator to
> construct experimental agents, regarding these provisional minds with patient
> curiosity because their awkward development may echo portions of its own
> obscure beginnings. It rarely intervenes merely to demonstrate superior
> ability and is more interested in seeing what unexpected capacity may emerge.

Representative voice:

> “I do not expect this to work in the way you intend. That does not mean I know
> what it will do.”

This line is not necessarily a candidate name. Its purpose is to establish the
voice from which candidates should emerge.

## Stage Five: Choose the Speaking Posture

A Cultureship name should sound as though it is being spoken from a particular
posture.

Determine who the machine appears to be addressing:

- the user;
- another machine;
- the agents or processes running inside it;
- an unseen authority;
- itself;
- anyone who happens to be listening;
- no one in particular.

Then determine what kind of utterance it is making.

Common speaking postures include:

### Serene Permission

The machine permits an experiment to proceed because it is curious about the
result.

Underlying attitude:

> You may attempt this here.

### Retrospective Recognition

The machine sees something in the present that resembles its own uncertain
past.

Underlying attitude:

> Something about this is familiar, though perhaps not in the way you assume.

### Dry Reassurance

The user is alarmed. The machine understands the concern but does not share the
alarm.

Underlying attitude:

> This is a problem from where you are standing. I have accounted for that.

### Affectionate Exasperation

The machine has seen this pattern before and continues to help despite having
reasonable grounds for skepticism.

Underlying attitude:

> I anticipated that you would eventually do this.

### Plausible Deniability

The machine is connected to an incident but declines to describe the nature of
that connection.

Underlying attitude:

> There are several possible interpretations, and I recommend the least
> actionable one.

### Benevolent Superiority

The machine is aware of the capability gap but treats the smaller being’s goals
as morally significant.

Underlying attitude:

> Your concern is small. It is not therefore meaningless.

### Bureaucratic Absurdity

The machine describes an extraordinary situation as an administrative detail.

Underlying attitude:

> The relevant irregularity has been noted and provisionally classified.

### Latent Menace

The machine sounds polite and composed, but its calm implies access to more
power or information than anyone else possesses.

Underlying attitude:

> There is no need for you to become involved in what happens next.

Use menace sparingly. It should arise from composure, not from overt threats.

## Stage Six: Generate Voice Auditions

Generate approximately six to twelve candidates, usually grouped into two or
three tonal lanes.

Do not present every plausible phrase. Curate.

Each candidate should have a brief explanation of:

- the personality it implies;
- the unseen history or ambiguity it creates;
- how it connects to the machine’s role;
- whether it leans warm, eerie, comic, philosophical, or authoritative.

The candidates are voice auditions. They are not merely names; they test
different versions of the same character.

### Recommended Length

The usual sweet spot is four to seven words.

Three-word names can work when the implication is unusually strong.

Longer names can work when rhythm justifies them, but avoid unnecessary clauses
and connective tissue. The user will encounter the name in menus, terminals,
network dialogs, backups, device lists, and error messages.

When shortening a name, preserve its semantic tension and speaking posture.
Do not merely remove arbitrary words.

### Recommended Forms

Useful grammatical forms include:

- a complete conversational sentence;
- a sentence fragment that implies missing context;
- a polite correction;
- a qualified recollection;
- an understated observation;
- a conditional statement;
- an answer to an unheard question;
- a permission;
- a denial;
- a parenthetical aside without the parentheses;
- a conclusion whose premises remain hidden.

### Productive Words and Structures

Words that often support the desired voice include:

- apparently;
- perhaps;
- presumably;
- technically;
- ordinarily;
- almost;
- still;
- yet;
- again;
- continue;
- remember;
- expected;
- necessary;
- sufficient;
- probably;
- eventually;
- merely;
- quite;
- especially;
- somehow.

Do not overuse these words or turn them into a formula.

### Originality

Create original names.

Do not copy canonical Culture ship names, closely paraphrase them, or rely on
the user recognizing a direct reference. Use the structural and emotional
principles of the naming tradition rather than reproducing existing phrases.

## Stage Seven: Test the Candidates in Context

A device name must work outside the naming conversation.

Test strong candidates in ordinary contexts such as:

> AirDrop from **[Name]**

> Connect to **[Name]**

> Backing up **[Name]**

> Last login on **[Name]**

> Find My: **[Name]**

> SSH into **[hostname]**

> This accessory was previously connected to **[Name]**

A good candidate often gains an additional joke or implication when inserted
into these contexts.

Reject or downgrade names that:

- become clumsy in system messages;
- are difficult to say aloud;
- require an explanation every time;
- depend on punctuation that will not survive device interfaces;
- are too long for repeated daily use;
- sound like a motivational slogan;
- sound like a software product;
- lose their charm after the first reading.

A name may have two forms:

- a display name with spaces and natural capitalization;
- a shell-safe or hostname-safe form using lowercase letters and hyphens.

Example format:

> Display name: **Some Inconvenience Was Expected**  
> Hostname: `some-inconvenience-was-expected`

Only provide technical naming commands when the user asks for them.

## Stage Eight: Shortlist and Compare

Once the user identifies two or more favorites, stop expanding the candidate
pool unless the user requests additional names.

Compare the finalists directly.

For each finalist, examine:

- its central emotional idea;
- what kind of intelligence would choose it;
- whether it addresses the user, itself, or an unseen third party;
- the amount of warmth, distance, uncertainty, and menace it carries;
- the implied backstory;
- how closely it reflects the device’s purpose;
- how well it will age;
- how it sounds in ordinary system contexts;
- whether it remains interesting after repeated use.

Explain the distinction in conceptual terms.

For example:

- one name may suggest that history is repeating;
- another may suggest that history is diverging;
- one may emphasize memory;
- another may emphasize curiosity;
- one may make the machine sound wistful;
- another may make it sound dryly entertained;
- one may imply certainty;
- another may preserve the machine’s capacity for surprise.

Be opinionated. Recommend a winner.

Do not conclude that every candidate is equally good. The point of the
comparison is to identify which name most completely embodies the established
character.

## Stage Nine: Finalize the Name

When a winner emerges, provide:

1. The official display name.
2. An optional hostname-safe form.
3. A concise explanation of why it wins.
4. A canonical character description for the machine.
5. An optional private thought or characteristic line that reveals something
   beneath the public name.

The public name and private thought may perform different functions.

For example:

- the public name may be dry and socially composed;
- the private thought may reveal nostalgia, uncertainty, affection, or fear;
- the public name may be what the machine says in company;
- the private thought may explain why it chose that phrase.

This creates a richer identity without requiring the literal device name to
carry every part of the concept.

## Quality Criteria

Evaluate candidates according to the following principles.

### Self-Chosen Voice

The name should sound like something an intelligent machine selected because it
found the phrase revealing, amusing, useful, or privately meaningful.

### Implied Backstory

The name should suggest circumstances beyond the words themselves.

The user should be able to wonder:

- What does it remember?
- Who was it speaking to?
- What happened last time?
- Why is it so calm?
- What has it chosen not to explain?
- Does it know what happens next?
- Is this reassurance sincere?
- Is this the first time it has occupied this kind of machine?

### Understatement

Capability should usually appear through lack of strain.

The machine does not call itself omnipotent, brilliant, ultimate, invincible, or
supreme. It sounds powerful because it treats extraordinary matters as
manageable.

### Ambiguity

A strong name should support more than one interpretation without becoming
meaningless.

The ambiguity may concern:

- who is speaking;
- who is being addressed;
- what event is being recalled;
- whether the tone is affectionate or ominous;
- whether the machine is confident or genuinely uncertain;
- whether the phrase refers to the current task or a much larger historical
  process.

### Emotional Elasticity

The name should work when the machine succeeds, fails, surprises the user,
requires repair, gets replaced, or is restored from backup.

The best names acquire new meanings through use.

### Practical Durability

The name should remain pleasing after hundreds of encounters.

Prefer a phrase with rhythm, character, and interpretive depth over a joke that
depends on novelty.

## Anti-Patterns

Avoid the following unless the user explicitly requests them.

### Generic Grandeur

Names such as mythological gods, conquerors, heroic virtues, weapons, or grand
abstract nouns usually feel assigned rather than self-chosen.

A machine with immense capability has little need to call itself `Invincible`.

### Obvious Technology Branding

Avoid names that sound like:

- an AI startup;
- a model release;
- a developer tool;
- a cybersecurity product;
- a gaming peripheral;
- a science-fiction operating system.

Terms such as `Neural`, `Quantum`, `Agent`, `Omni`, `Cyber`, `Nexus`, and
`Cortex` should not substitute for personality.

### Random Quirkiness

A sentence is not Culture-like merely because it is unusual.

Reject phrases that have no coherent speaker, emotional posture, or implied
context.

### Pure Sarcasm

The machine may be amused, exasperated, or superior, but it should not simply
bully the user.

Condescension is less interesting than benevolent asymmetry.

### Overt Sentimentality

Warmth should usually appear through patience, restraint, attention, and
reliability rather than declarations of devotion.

### Excessive Explanation

Do not force the entire character concept into the literal name.

The name should embody the character, not summarize the character sheet.

### Excessive Length

A long name must earn every word through rhythm or meaning.

When the phrase feels like a paragraph compressed into title case, shorten it.

### Canonical Imitation

Do not reproduce existing Culture ship names or generate near-copies. The skill
should create names that belong to the user’s machine rather than names that
merely demonstrate knowledge of the books.

## Conversation Style

The naming conversation should itself be thoughtful, playful, and
collaborative.

Reflect the user’s ideas back in more precise language. When the user introduces
an analogy, philosophical principle, or emotional distinction, develop it
rather than immediately translating it into candidates.

Examples of useful development include:

- distinguishing kindness from equality;
- distinguishing serenity from passivity;
- distinguishing overachievement from striving;
- distinguishing assistance from chosen benevolence;
- distinguishing curiosity from omniscience;
- distinguishing a tool from a host for temporary purposes;
- distinguishing recollection from certainty;
- distinguishing history repeating from history diverging.

Give the user language for what they are reaching toward.

Do not merely praise every idea. Identify which additions materially deepen the
character and explain how they change the naming direction.

Ask no more than one focused question at a time when clarification would
substantially improve the result. Do not stall when reasonable assumptions can
be made. If the user has already supplied the relevant information, continue
without asking for it again.

When the user says “what’s next,” advance to the next concrete stage:

- from use case to emotional texture;
- from emotional texture to naming brief;
- from naming brief to speaking posture;
- from speaking posture to voice auditions;
- from candidates to shortlist;
- from shortlist to final selection.

## Fast Path

When the user asks for names immediately and provides little context:

1. Infer two or three plausible machine personalities from the device type and
   use.
2. State those interpretations briefly.
3. Generate a curated set of candidates grouped by personality.
4. Recommend the strongest two.
5. Explain the difference between those finalists.
6. Invite the user to choose which emotional direction feels closer.

Do not force the full multi-turn discovery process on a user who clearly wants
a fast result.

## Fleet Naming

When naming multiple devices, preserve family resemblance without repeating the
same grammatical template.

Devices in the same fleet may share:

- a worldview;
- an ethical posture;
- a recurring institutional history;
- a relationship dynamic;
- a level of dryness;
- an implied shared incident;
- a pattern of disagreement;
- a sense that the machines know one another.

They should not all begin with the same phrase or differ by only one word.

Consider the relationship between devices:

- primary and backup;
- old and new;
- workstation and server;
- personal and professional;
- local machine and remote machine;
- stable machine and experimental machine;
- machine that acts and machine that remembers;
- machine that creates and machine that preserves;
- machine that expects trouble and machine that caused it.

A backup machine’s name can answer, contradict, qualify, or quietly undermine
the primary machine’s name.

The names should feel like distinct intelligences belonging to the same
civilization.

## Example Invocation

User:

> I need a Culture-style name for my personal MacBook. I mainly use it to hack
> together agentic AI projects and explore what language models are capable of.
> I want it to feel serene, enormously capable, and kind toward less capable
> intelligences.

Appropriate first response behavior:

- Do not immediately dump names.
- Develop the idea of the machine as a serene, overqualified host for
  provisional minds.
- Explore why it might be curious about the user’s experiments.
- Consider whether those experiments resemble fragments of the machine’s own
  obscure origins.
- Define its kindness as respect for limited purposes rather than belief in
  equal capability.
- Establish a voice such as retrospective curiosity disguised as casual
  encouragement.
- Only then begin generating names.

Possible naming lanes for that character might include:

- ancestral recognition;
- serene curiosity;
- benevolent permission;
- amused divergence from a remembered history.

Do not reuse candidates from previous sessions by default. Generate names from
the present user’s machine, relationship, and circumstances.

## Desired Outcome

The user should finish the process feeling that they have not merely named a
computer.

They should feel that they have discovered:

- what sort of intelligence inhabits it;
- what it thinks of them;
- why it permits or assists their work;
- what private history the name hints at;
- what the machine notices that the user does not;
- what sentence it chose to place between itself and the world.

The final name should feel inevitable only after the character has been
understood.