fluid-x/XCapabilityCanon.md
2026-09-05 00:19:30 +02:00

46 KiB
Executable file

X Capability Canon

Status: Draft v0.1
Date: 2026-09-03
Platform: X, formerly Twitter
Purpose: Canonical capability model for using X as a public conversation, publishing, discovery, real-time intelligence, community, media, business, creator, and agentic interaction platform.


1. Purpose

This canon provides a durable capability model for X.

It is intentionally not organized as a tour of the X application or a chronological list of product features. Instead, it models X as a set of reusable platform capabilities that can be mapped to concrete products, communities, organizations, media operations, intelligence systems, automated accounts, and agentic personalities.

The primary questions are:

  1. What capability does X provide?
  2. Which X surface exposes it?
  3. Is the capability primarily about publishing, conversation, discovery, attention, intelligence, community, or private interaction?
  4. Can it be accessed or controlled programmatically?
  5. Can an automated account or autonomous agent participate legitimately?
  6. Which platform rules constrain that automation?
  7. Which state belongs on X, and which state should remain in external systems?
  8. How should public agentic action be governed?

This canon is designed to support projects such as an x-edge, public-facing community systems such as the Coulomb Society, agentic personalities such as Charles-Augustin de Coulomb Sp., research and intelligence systems, institutional publishing, and cross-platform social infrastructure.


2. Strategic Platform Model

X should be treated as eight overlapping platforms:

X
│
├── Public Conversation Network
│   ├── posts
│   ├── replies
│   ├── reposts
│   ├── quote posts
│   └── conversation graphs
│
├── Attention & Discovery Network
│   ├── For You
│   ├── Following
│   ├── search
│   ├── trends
│   ├── hashtags
│   ├── Lists
│   └── recommendation surfaces
│
├── Publishing & Media Platform
│   ├── posts
│   ├── threads
│   ├── Articles
│   ├── images
│   ├── video
│   ├── live media
│   └── Spaces
│
├── Community Platform
│   ├── follower networks
│   ├── Communities
│   ├── Lists
│   ├── Spaces
│   └── creator subscriptions
│
├── Private Communication Platform
│   ├── direct messages / Chat
│   ├── group messaging
│   ├── voice
│   └── video
│
├── Business & Creator Platform
│   ├── Premium
│   ├── Premium Business
│   ├── organization affiliations
│   ├── advertising
│   ├── subscriptions
│   └── creator monetization
│
├── AI / Intelligence Platform
│   ├── Grok
│   ├── real-time public content
│   ├── Community Notes
│   └── trend / conversation intelligence
│
└── Data & Automation Platform
    ├── X API
    ├── filtered streams
    ├── full-archive search
    ├── webhooks / activity events
    ├── analytics
    ├── automated accounts
    └── enterprise access

The most important conceptual difference from Telegram is:

Telegram is primarily an interaction environment. X is primarily a public attention and information environment.


3. Capability Levels

Capabilities in this canon can be considered at five implementation levels.

Level Meaning
L0 — Human UI Available through standard X applications.
L1 — API Readable Can be observed or retrieved through the X API.
L2 — API Actionable Can be created or modified programmatically through supported APIs.
L3 — Integrated Can participate in an external application, workflow, analytics, or intelligence system.
L4 — Agentic Suitable for bounded automated or autonomous agent participation.

A capability may exist at several levels simultaneously.


4. Public Posting Capabilities

POST-01 — Posts

The foundational X primitive is the Post.

A Post is:

  • globally addressable;
  • attributable to an account;
  • timestamped;
  • part of one or more public social graphs;
  • potentially discoverable outside the author's follower network;
  • capable of receiving replies, reposts, quotes, likes, bookmarks, and other engagement.

Posts may contain or reference:

  • text;
  • links;
  • images;
  • GIFs;
  • video;
  • polls;
  • mentions;
  • hashtags;
  • quoted posts;
  • location/context metadata where supported.

Agentic relevance

Very high.

A public agent on X naturally expresses itself through posts.

This makes the agent not merely a service endpoint but a public actor.


POST-02 — Replies

Replies establish conversational structure around posts.

Conceptually:

Post
├── Reply
│   ├── Reply
│   └── Reply
├── Reply
└── Reply

Replies create one of X's most important properties:

Public content is embedded in a conversation graph, not merely a feed.

Uses

  • debate;
  • questions;
  • support;
  • commentary;
  • corrections;
  • expert participation;
  • public relationship building.

Agentic relevance

Very high, but high-risk.

Automated replies can provide value, but unsolicited or repetitive engagement can easily become spam.


POST-03 — Reposts

Reposts redistribute another account's post to additional audiences.

This creates a graph-propagation primitive.

Function

source content
     ↓
repost
     ↓
secondary audience

Agentic relevance

Medium to high.

Automated reposting may be permitted within X's automation rules, but aggressive, mass, duplicate, or spam-like behavior is prohibited.


POST-04 — Quote Posts

Quote posts combine amplification with added commentary.

This is particularly important for:

  • criticism;
  • interpretation;
  • contextualization;
  • recommendation;
  • editorial curation.

Agentic relevance

High.

Quote posts are a natural expression mode for analytical or editorial agents because they allow an agent to add substantive interpretation rather than simply reproduce content.


POST-05 — Threads

Multiple connected posts can form longer sequences.

Suitable for:

  • explanations;
  • narratives;
  • research summaries;
  • event reporting;
  • tutorials;
  • commentary.

Threads bridge short-form social posting and longer-form publishing.


5. Long-Form Publishing

PUB-01 — Articles

X Articles provide long-form publishing.

Capabilities include:

  • long-form text;
  • headings;
  • subheadings;
  • bold/italic/strikethrough formatting;
  • indentation;
  • numbered and bulleted lists;
  • images;
  • video;
  • GIFs;
  • embedded posts;
  • links.

Publishing requires an eligible X Premium, Premium+, Premium Business, or Premium Organizations subscription.

Reading and sharing can be broader depending on audience settings.

Strategic interpretation

Articles move X closer to:

  • blogging;
  • newsletters;
  • essays;
  • institutional commentary;
  • research publication.

Agentic relevance

High.

An agent can use posts for presence and Articles for durable exposition.


PUB-02 — Media Publishing

X supports native publication of:

  • images;
  • GIFs;
  • video;
  • longer video for eligible subscribers;
  • live media.

This enables accounts to operate as multimedia publishers rather than text-only social accounts.


PUB-03 — Live Publishing

Live interaction surfaces include:

  • live video where available;
  • Spaces;
  • event-oriented conversation.

Uses

  • announcements;
  • interviews;
  • live commentary;
  • community events;
  • discussions;
  • public briefings.

6. Attention Capabilities

Attention is not a secondary property of X. It is a core platform capability.

A useful abstraction is:

Content
   ↓
Eligibility
   ↓
Ranking
   ↓
Discovery
   ↓
Exposure
   ↓
Engagement
   ↓
Further distribution

X therefore provides a probabilistic attention-distribution system on top of the social graph.


ATT-01 — Following Timeline

Users can consume content from accounts they follow.

This represents the explicit social graph:

User
  ↓ follows
Account set
  ↓
Following timeline

ATT-02 — For You Timeline

The For You timeline provides algorithmically selected content beyond strict follower relationships.

Strategic implication

Publishing reach is not limited to accumulated followers.

A new account may gain exposure through recommendation mechanisms if its content is selected.

Governance note

Recommendation exposure is not deterministic or contractually guaranteed.

Do not build critical delivery systems around feed visibility.


Search exposes public content and accounts.

Typical search dimensions include:

  • keywords;
  • accounts;
  • posts;
  • media;
  • time;
  • hashtags;
  • conversational context.

API search also supports structured query operators.


X surfaces trending topics and provides trend data through API capabilities.

Uses

  • current-event discovery;
  • public interest sensing;
  • campaign monitoring;
  • anomaly detection;
  • cultural observation.

Agentic warning

X's automation rules prohibit automatically posting about trending topics merely as an automated engagement tactic or manipulating trends.


ATT-05 — Hashtags

Hashtags provide a lightweight public categorization and discovery mechanism.

They can:

  • connect independent conversations;
  • create event tags;
  • support campaign identity;
  • support search and monitoring.

ATT-06 — Lists

Lists are curated account collections with their own timeline.

They can be:

  • public;
  • private;
  • followed;
  • pinned;
  • programmatically managed where supported.

Strategic value

Lists are an underappreciated intelligence primitive.

Examples:

AI Researchers
Cybersecurity Vendors
German Healthcare IT
Quantum Computing
Coulomb Society Members
Competitors
Journalists
Policy Makers

Agentic relevance

Very high for observation.

Lists can define bounded source populations for research agents.


7. Real-Time Intelligence Capabilities

X's public-conversation graph makes it valuable as an external information source.

Programmatic search can observe current conversation.

Suitable for:

  • monitoring;
  • research;
  • event tracking;
  • brand intelligence;
  • competitor intelligence;
  • community sensing.

The X API supports full-archive search across public posts back to 2006, subject to access and commercial terms.

Uses

  • longitudinal studies;
  • event reconstruction;
  • historical discourse analysis;
  • account history analysis;
  • trend research.

INTEL-03 — Filtered Stream

The API supports near-real-time delivery of matching posts using filter rules.

Conceptually:

Public X stream
      ↓
filter rules
      ↓
matching posts
      ↓
x-edge
      ↓
analysis / storage / agents

The current standard documentation describes support for up to 1,000 filtering rules.

Agentic relevance

Extremely high.

This turns X into an event source for:

  • news agents;
  • research agents;
  • security intelligence;
  • market sensing;
  • community listening;
  • event detection.

INTEL-04 — Engagement Metrics

Structured post metrics can include signals such as:

  • impressions;
  • likes;
  • replies;
  • reposts;
  • video views.

Uses

  • content evaluation;
  • experimentation;
  • audience analysis;
  • campaign analysis;
  • agent feedback loops.

Warning

Engagement is not synonymous with truth, value, or healthy outcomes.

Agent optimization must not collapse into engagement maximization.


INTEL-05 — Radar

X provides conversation-monitoring functionality through Radar-related product offerings.

This extends X's role from social platform toward native intelligence tooling.


8. Community Capabilities

COM-01 — Followers

The follower graph provides the broadest persistent audience/community relationship.

It is:

  • account-centric;
  • asymmetric;
  • public by default for public accounts;
  • highly suitable for publishing and reach;
  • less suitable for strong membership semantics.

COM-02 — Communities

X Communities provide dedicated spaces for interest-based conversation.

They support:

  • admins;
  • moderators;
  • members;
  • Community rules;
  • open or restricted membership;
  • membership requests;
  • invitations;
  • Community-specific posts;
  • Community timelines;
  • moderation.

Community posts may remain publicly visible and searchable beyond the Community itself.

Important distinction

An X Community is not equivalent to a private Telegram group.

It is better understood as:

a governed conversation scope embedded in the wider public X graph.


COM-03 — Community Administration

Admins can manage:

  • Community identity;
  • descriptions;
  • cover imagery;
  • rules;
  • membership model;
  • moderators.

Moderators can:

  • review reports;
  • hide Community posts;
  • manage membership;
  • enforce Community rules;
  • cultivate discussion.

COM-04 — Community Rules

Communities can define rules in addition to X-wide rules.

This creates two layers:

X Rules
   ↓
Community Rules
   ↓
Community Moderation

Community rules cannot override X rules.


COM-05 — Community Spaces

Spaces can be associated with Community activity and provide synchronous conversation to complement asynchronous posts.


9. Spaces and Live Audio

SPACE-01 — Spaces

Spaces provide public live audio conversations.

Typical participant roles include:

  • host;
  • co-host;
  • speaker;
  • listener.

Capabilities include:

  • scheduled sessions;
  • public attendance;
  • speaker participation;
  • recordings where enabled;
  • integration with public X identity.

Uses

  • interviews;
  • panels;
  • lectures;
  • community meetings;
  • live commentary;
  • Q&A;
  • events.

Agentic relevance

Emerging.

Textual agents may prepare or support Spaces even when real-time voice participation requires additional infrastructure.


10. Private Communication

DM-01 — Direct Messages / Chat

X supports private messaging between accounts.

Capabilities include:

  • one-to-one conversations;
  • group conversations;
  • text;
  • media;
  • files;
  • reactions;
  • voice/media functionality depending on client and rollout state.

The X API supports sending and receiving Direct Messages.

Agentic relevance

High.

DMs support:

  • private assistants;
  • support agents;
  • inbound lead workflows;
  • notifications;
  • personalized follow-up;
  • opt-in conversational automation.

DM-02 — Automated DM Interaction

X permits automated DM responses when they follow platform automation and anti-spam rules.

Appropriate pattern

User initiates / opts in
         ↓
Automated service
         ↓
Useful response

Inappropriate pattern

Account collection
       ↓
mass unsolicited DMs
       ↓
spam

DM-03 — Voice and Video Calls

X supports audio and video calling, including evolving group-call functionality.

These are primarily human communication surfaces rather than first-class public API automation surfaces.


11. Grok and Native AI

AI-01 — Grok

Grok is integrated into X as a native AI capability.

It can operate over or alongside:

  • public X content;
  • current conversation;
  • user questions;
  • web information;
  • selected X content.

Strategic implication

X is simultaneously:

  1. a data source for AI,
  2. an interaction surface for AI,
  3. a distribution surface for AI-generated output.

AI-02 — Contextual AI Interaction

Users can invoke Grok in relation to posts and platform content.

This creates a native pattern:

Public content
     ↓
AI interpretation
     ↓
user understanding / action

AI-03 — AI and Private Content Boundary

When private content is explicitly provided to an AI system, it crosses from private communication into AI processing.

Applications should treat this as a deliberate data-boundary transition.


12. Community Notes

NOTE-01 — Community Notes

Community Notes provides collaborative contextual information attached to public posts.

Conceptually:

Post
  ↓
proposed contextual notes
  ↓
distributed evaluation
  ↓
eligible note
  ↓
public context

Strategic significance

Community Notes is unusual because it adds a collective epistemic layer to the public conversation graph.


NOTE-02 — Programmatic Access

Community Notes are represented in X's API/data products.

This makes them relevant to:

  • research;
  • misinformation analysis;
  • trust systems;
  • agentic reasoning;
  • content-risk assessment.

Agentic recommendation

A research or publishing agent should treat a visible Community Note as a significant evidence signal, not merely another reply.


13. Identity Capabilities

ID-01 — X Account Identity

X accounts provide persistent public identity.

Relevant elements include:

  • numeric account ID;
  • handle;
  • display name;
  • profile;
  • biography;
  • follower/following graph;
  • verification indicators;
  • labels;
  • organizational affiliations.

Architecture rule

Use numeric X IDs as canonical external identifiers where APIs expose them.

Handles can change.


ID-02 — Premium Verification

Blue checkmarks are associated with eligible Premium subscriptions rather than the older legacy concept of independent notability verification.

Therefore:

A blue check is an account-status signal, not by itself proof of public significance or authenticity.


ID-03 — Organization Identity

Premium Business provides organizational identity features including:

  • gold verification;
  • organizational affiliations;
  • related account management.

Government/multilateral identities may use grey verification and corresponding organization products.


ID-04 — Affiliation

Accounts can be visibly affiliated with an organization.

This is useful for multi-account structures:

Coulomb Society
├── official account
├── event account
├── research account
└── Charles-Augustin de Coulomb Sp.

Affiliation can make relationships legible to users.


ID-05 — Automated Account Label

X provides an automated-account labeling model.

Automated accounts are expected to remain connected to a responsible human-run account under X's automation model.

Strategic implication

X explicitly models automated public actors as delegated identities with human accountability.

This is materially different from Telegram's bot model.


14. Business Capabilities

BUS-01 — Premium Business

Premium Business provides organizational features such as:

  • business verification;
  • gold checkmark;
  • organization affiliation management;
  • Premium functionality;
  • business-facing discovery/monitoring features depending on tier.

As of this canon snapshot, X documents Basic and Full Access tiers, but commercial terms should be treated as volatile.


BUS-02 — Professional Presence

X accounts can function as public business surfaces for:

  • branding;
  • support;
  • product announcements;
  • thought leadership;
  • recruiting;
  • investor communication;
  • developer relations;
  • customer interaction.

BUS-03 — Organizational Account Graph

Affiliations allow an organization to establish an explicit graph of related identities.

This is valuable for agent-active organizations because human and artificial representatives can potentially be made distinguishable while retaining institutional association.


15. Advertising Capabilities

ADS-01 — Paid Distribution

X provides advertising products for increasing content and campaign reach.

Typical objectives include:

  • awareness;
  • engagement;
  • website traffic;
  • video views;
  • app acquisition;
  • conversions.

ADS-02 — Media Advertising

Ad formats may include:

  • text/post promotion;
  • images;
  • video;
  • vertical video;
  • carousels;
  • product/collection formats;
  • larger takeover or premium-placement formats.

ADS-03 — Targeting

Advertising systems can use signals such as:

  • geography;
  • keywords;
  • follower relationships;
  • conversations;
  • interests;
  • platform behavior.

Architecture note

Advertising is a paid attention mechanism and should be modeled separately from organic reach.


16. Creator Capabilities

CRE-01 — Creator Subscriptions

Eligible creators can offer recurring paid subscriptions.

Subscriber benefits can include:

  • exclusive posts;
  • subscriber badges;
  • subscriber-specific interaction privileges.

Strategic interpretation

X can support a paid membership layer around a public identity.


CRE-02 — Creator Monetization Transition — September 2026

At the date of this canon, X is transitioning its creator monetization model.

Timeline:

2026-08-07
new Creator Revenue Sharing enrollment stops

2026-09-07
Creator Revenue Sharing retires

2026-09-08
rollout of Original Content Rewards begins

The new Original Content Rewards program is designed to reward qualifying original content based on qualifying Premium-user impressions and other eligibility rules.

Important automation constraint

X states that content created or posted using automated means is ineligible for Original Content Rewards.

Therefore:

Do not design autonomous publishing agents around the assumption that their automated output will qualify for X creator-reward monetization.

Agentic publishing and creator monetization must be modeled as separate capability surfaces.


17. Developer Platform

API-01 — X API

The X API provides programmatic access to the public conversation and multiple platform objects.

Current documented areas include:

  • Posts;
  • users;
  • Spaces;
  • Direct Messages;
  • Lists;
  • trends;
  • media;
  • engagement data;
  • Communities;
  • Community Notes and related objects where exposed.

API-02 — Post Retrieval

Applications can retrieve posts and related structured fields.

Possible use cases:

  • archives;
  • research;
  • dashboards;
  • monitoring;
  • content analysis;
  • agent context.

API-03 — Post Publishing

Authorized applications can publish posts.

This is the primary mechanism for legitimate automated public publishing.

Rule

Do not automate the X website through browser scripting as a substitute for the official API.

X's automation rules explicitly prohibit non-API automation techniques such as scripting the website.


API-04 — User Relationships

The API supports aspects of user relationship management such as:

  • follows;
  • blocks;
  • mutes;
  • user lookups.

Automation rules constrain aggressive or indiscriminate behavior.


API-05 — Direct Messages

The API supports private-message interaction.

This allows:

X DM
  ↓
x-edge
  ↓
workflow / agent / CRM
  ↓
response

API-06 — Lists

Lists can be created and managed programmatically.

This is particularly valuable for machine-maintained intelligence sets.


API-07 — Spaces

The API exposes data about live audio Spaces and participants.


Trend data is programmatically available subject to the applicable access model and pricing.


API-09 — Webhooks / Activity Events

X provides event-oriented APIs that can deliver account/activity events.

This allows event-driven systems instead of continuous polling.

Example architecture:

X activity
   ↓
webhook
   ↓
x-edge
   ↓
event bus
   ↓
application / agent

18. API Economics

ECON-01 — Pay-Per-Use

As of 2026-09-03, the standard X API uses credit-based pay-per-use pricing.

Characteristics include:

  • prepaid credits;
  • per-endpoint/resource pricing;
  • live cost tracking;
  • configurable spending limits;
  • optional auto-recharge;
  • higher-volume Enterprise offerings.

ECON-02 — Current Pricing Snapshot

The following is an illustrative snapshot from X's current documentation and must not be treated as durable canon:

Operation Current documented unit cost
Post read $0.005/resource
User read $0.010/resource
DM event read $0.010/resource
List read $0.005/resource
Space read $0.005/resource
Community read $0.005/resource
Community Note read $0.005/resource
Post create $0.015/request
Post create with URL $0.200/request
DM interaction create $0.015/request

Current pay-per-use access is documented as capped at 3 million Post reads per monthly billing cycle, with higher-volume requirements moving to Enterprise access.

Canon rule

Treat API prices and quotas as runtime configuration, not architectural constants.


ECON-03 — Owned Reads

X provides reduced-cost access for certain data owned by the authenticated developer/app account.

This can materially improve the economics of:

  • personal analytics;
  • account dashboards;
  • own-post monitoring;
  • own-audience tools.

ECON-04 — Cost Governance

Agentic systems should enforce deterministic spend limits.

Recommended structure:

Agent query
   ↓
cost estimator
   ↓
budget policy
   ↓
API request

A runaway research agent should not be able to consume unlimited API credits.


19. Automation Model

AUTO-01 — Automated Accounts

X permits automated accounts that perform useful actions through the X API.

Examples explicitly contemplated by X include:

  • information broadcasting;
  • service responses;
  • useful automated DMs;
  • public-interest services.

AUTO-02 — Human Accountability

X's automated-account model expects automated accounts to remain connected to a human-run responsible account.

Conceptually:

Human principal
      │
      ▼
Automated account
      │
      ▼
X public graph

This is a useful governance pattern even where the underlying automation is agentic.


AUTO-03 — Transparent Automation

Automated-account labels provide users with a signal that content is machine-produced.

For public AI personalities this supports the principle:

Artificial agency should be legible where it materially affects how others interpret the speaker.


AUTO-04 — API-Only Automation

X's rules prohibit using non-API automation such as scripting the website.

This creates a clean architectural boundary:

Agent
  ↓
Policy
  ↓
x-edge
  ↓
Official X API

not:

Agent
  ↓
browser automation
  ↓
X website

20. Automation Constraints

X is comparatively permissive toward useful automation but strongly limits manipulative automation.

RULE-01 — No Spam

Automated content must not become spam.


RULE-02 — No Trend Manipulation

Automated systems must not attempt to exploit or manipulate trending topics.


RULE-03 — No Duplicate Swarms

Do not create multiple automated accounts for substantially duplicative use cases or distribute substantially identical posts across an account network.


RULE-04 — Automated Likes

X's automation rules prohibit automated liking of Posts.


RULE-05 — Reposts

Automated reposting/quote posting may be legitimate when useful, but mass, aggressive, or spam-like behavior is prohibited.


RULE-06 — Follow/Unfollow Automation

Aggressive, indiscriminate, or mass follow/unfollow behavior is prohibited.


RULE-07 — Automated Actions on Behalf of Users

OAuth authorization alone is not sufficient consent for arbitrary automated actions.

Applications acting for another user should:

  1. explain what actions will occur;
  2. obtain express consent;
  3. honor opt-out;
  4. re-obtain consent if functionality materially changes.

21. Agentic Capability Model

For X, it is useful to distinguish seven agent roles.

AGT-01 — Publisher Agent

Agent
  ↓
creates original public content
  ↓
X account

Examples:

  • institutional publisher;
  • research broadcaster;
  • project update agent.

AGT-02 — Commentator Agent

external event / post
        ↓
analysis
        ↓
reply / quote post

Examples:

  • subject-matter expert;
  • fact explainer;
  • scientific commentator.

This role has considerably more social risk than simple publishing.


AGT-03 — Curator Agent

X stream
   ↓
filter / evaluate
   ↓
select
   ↓
repost / quote / digest

The curator should add substantive value rather than simply maximize volume.


AGT-04 — Correspondent Agent

Observes defined sources or events and publishes timely reports.

Examples:

  • conference correspondent;
  • project-release reporter;
  • sector-news correspondent.

AGT-05 — Research / Sensor Agent

X data
  ↓
search / stream / Lists
  ↓
analysis
  ↓
external knowledge system

The agent may never need to publish.

This may be one of X's strongest agentic use cases.


AGT-06 — Community Agent

Participates in a Community as:

  • discussion catalyst;
  • FAQ agent;
  • moderator assistant;
  • research assistant;
  • event facilitator.

AGT-07 — Public Personality

An artificial personality may maintain a coherent public identity over time.

Example:

Charles-Augustin de Coulomb Sp.

Such an entity may:

  • publish;
  • comment;
  • converse;
  • curate;
  • host intellectual themes;
  • represent a community role.

Governance requirement

The artificial nature and responsible human/organization should be legible and governable.


22. Human / Agent Responsibility Model

Public social interaction is higher-risk than private workflow automation because the agent can affect:

  • reputation;
  • third parties;
  • public discourse;
  • organizational relationships;
  • legal exposure;
  • brand identity.

Recommended model:

             Agent / LLM
                 │
                 ▼
         proposed public action
                 │
                 ▼
       deterministic policy layer
                 │
          ┌──────┴──────┐
          │             │
       denied         allowed
                        │
                        ▼
                   X adapter
                        │
                        ▼
                    X API

23. Agentic Action Classes

X actions should be classified by risk.

Class A — Observation

Examples:

  • read posts;
  • search;
  • retrieve Lists;
  • read trends;
  • monitor mentions.

Usually suitable for high autonomy.


Class B — Internal Interpretation

Examples:

  • summarize;
  • classify;
  • score relevance;
  • detect events;
  • create internal recommendations.

Usually suitable for high autonomy if data policy permits.


Class C — Drafting

Examples:

  • draft post;
  • draft reply;
  • propose quote post;
  • prepare Article.

Low public risk until released.

Ideal for human-assisted agentic workflows.


Class D — Public Publishing

Examples:

  • create Post;
  • publish thread;
  • publish Article;
  • repost.

Requires explicit identity and policy.

Bounded autonomous operation may be appropriate.


Class E — Social Intervention

Examples:

  • reply to a person;
  • mention a person;
  • follow/unfollow;
  • moderate;
  • DM.

Higher social risk.

Requires stronger constraints.


Class F — High-Impact Actions

Examples:

  • mass communication;
  • controversy-sensitive engagement;
  • organizational statements;
  • account/block actions;
  • paid advertising;
  • commercial commitments.

Prefer deterministic authorization and often human approval.


24. Attention Governance

Agents interacting with X should not optimize blindly for engagement.

A dangerous optimization loop is:

post
  ↓
engagement score
  ↓
generate more provocative content
  ↓
more engagement
  ↓
increase provocation

This can reliably degrade content quality and institutional behavior.

Recommended objective model:

Value
+ factual quality
+ relevance
+ institutional intent
+ relationship quality
+ bounded reach
+ safety
--------------------------------
not merely engagement

25. Epistemic Governance

X is both information-rich and information-noisy.

Research agents should distinguish:

Observation
    ↓
Claim
    ↓
Source
    ↓
Corroboration
    ↓
Confidence

Signals such as:

  • verified organizational origin;
  • primary-source content;
  • timestamps;
  • direct media;
  • multiple independent sources;
  • Community Notes;
  • account history;
  • conversation context

should influence confidence.

X content should not automatically be treated as verified fact.


26. Security Model

SEC-01 — API Credentials

X application credentials and OAuth tokens are secrets.

They should be managed through a proper secret-management system.

Required controls:

  • secret storage;
  • scope minimization;
  • rotation;
  • revocation;
  • audit;
  • environment separation.

SEC-02 — Identity Separation

Prefer separate identities for materially different responsibilities.

Example:

Coulomb Society Official
Coulomb Research
Charles-Augustin de Coulomb Sp.
Coulomb Events

Do not overload one account with incompatible roles merely for convenience.


SEC-03 — External Authorization

X account permissions should not become the canonical authorization model for external infrastructure.

Recommended:

X identity
    ↓
identity mapping
    ↓
internal subject
    ↓
authorization policy

SEC-04 — Private Data

Sensitive information should not be collected through public interactions.

X's automation rules explicitly encourage minimizing information collected through automated DMs and directing users to more appropriate channels for especially sensitive data.


27. X as an Edge Architecture

For sophisticated systems, X should be modeled as a public social and intelligence edge.

                         X
                         │
                       x-edge
                         │
       ┌─────────────────┼─────────────────┐
       │                 │                 │
   Publishing        Intelligence      Interaction
       │                 │                 │
       └─────────────────┼─────────────────┘
                         │
                 Identity / Policy
                         │
          ┌──────────────┼──────────────┐
          │              │              │
       Agents        Knowledge       Workflows
          │              │              │
          └──────────────┼──────────────┘
                         │
                    Core Systems

x-edge should normally own:

  • API authentication;
  • account mappings;
  • X identity mappings;
  • inbound event normalization;
  • outbound rendering;
  • API-rate and cost handling;
  • post/reply/quote semantics;
  • webhook handling;
  • media handling;
  • retry behavior;
  • audit correlation;
  • automation-rule enforcement;
  • platform-policy adaptation.

It should not own domain truth that belongs elsewhere.


28. Canonical X Entity Model

A useful normalized model is:

XPlatform
│
├── Account
│   ├── HumanAccount
│   ├── OrganizationAccount
│   └── AutomatedAccount
│
├── Post
│   ├── RootPost
│   ├── Reply
│   ├── Repost
│   └── QuotePost
│
├── Article
│
├── Conversation
│
├── Community
│   ├── Member
│   ├── Moderator
│   └── Admin
│
├── List
│
├── Space
│
├── DirectConversation
│   └── Message
│
├── CommunityNote
│
├── Trend
│
├── Media
│
└── Relationship
    ├── Follow
    ├── Block
    ├── Mute
    └── Affiliation

29. Capability Selection Guide

Requirement Preferred X Primitive
Short public statement Post
Public discussion Replies
Add context to another post Quote Post
Amplify useful content Repost
Longer explanation Thread
Durable essay Article
Multimedia publishing Post / Article / Video
Live public discussion Space
Interest-based discussion Community
Curated source feed List
Direct private interaction DM / Chat
Public discovery Search / For You / Trends
Real-time data monitoring Filtered Stream
Historical research Full-Archive Search
Organization identity Premium Business
Related account identity Affiliations
Public automated identity Automated Account
Native platform AI Grok
Contextual trust signal Community Notes
Paid member content Creator Subscription
Paid reach Advertising

30. X vs Telegram Capability Boundary

A useful canonical distinction:

Telegram                         X

interaction                     attention
membership                      audience
service                         presence
workflow                        publishing
persistent chat                 public conversation
bot interface                   public actor
community operations            discovery
private coordination            public influence
Mini Apps                       information graph

This is not absolute, but it captures their respective centers of gravity.


31. Combined Social Architecture

The platforms are complementary.

                       Community / Institution
                                │
              ┌─────────────────┴──────────────────┐
              │                                    │
         Telegram Edge                           X Edge
              │                                    │
     membership & services              public presence & reach
     persistent discussion              publishing
     workflows                          discovery
     private interaction                commentary
     events                             trend sensing
     service agents                     public personalities
              │                                    │
              └─────────────────┬──────────────────┘
                                │
                       Identity / Governance
                                │
                     Knowledge / Agent Systems

32. Coulomb Society Reference Pattern

A possible structure is:

Coulomb Society
│
├── X
│   ├── @CoulombSociety
│   │   ├── institutional publishing
│   │   ├── announcements
│   │   └── public conversation
│   │
│   ├── Charles-Augustin de Coulomb Sp.
│   │   ├── commentary
│   │   ├── curation
│   │   ├── public intellectual persona
│   │   └── community outreach
│   │
│   ├── Lists
│   │   ├── Society Members
│   │   ├── Scientists
│   │   ├── Artists
│   │   └── Interesting Institutions
│   │
│   ├── Community
│   │   └── public thematic discussion
│   │
│   └── Spaces
│       └── public lectures / discussions
│
└── Telegram
    ├── deeper membership
    ├── persistent community discussion
    ├── agent services
    ├── workflows
    └── events / coordination

The architectural principle remains:

The Coulomb Society exists independently of either platform. X and Telegram are social surfaces through which the Society is experienced.


33. Capability Maturity Model

Stage Description
X0 — Manual Humans manually publish and interact.
X1 — Observed External systems read X and gather metrics.
X2 — Integrated X is connected to knowledge, CRM, workflow, or analytics systems.
X3 — Assisted Publishing Agents research and draft; humans publish.
X4 — Policy-Governed Publishing Selected content can be published automatically under deterministic rules.
X5 — Bounded Social Agency Agents can reply, curate, and interact within explicit social boundaries.
X6 — Public Artificial Actor A coherent artificial personality participates continuously under transparent identity and governance.
X7 — Multi-Agent Public Institution Several specialized artificial and human identities coordinate institutional activity while accountability remains explicit.

Maximum maturity should not be the default goal.

Each account should advance only as far as its:

  • value;
  • reversibility;
  • observability;
  • reputation risk;
  • policy stability;
  • human accountability

justify.


34. X Strengths

X is particularly strong where a system needs several of the following:

  • public identity;
  • real-time public conversation;
  • rapid publishing;
  • broad discovery;
  • attention amplification;
  • expert networks;
  • event awareness;
  • trend monitoring;
  • public commentary;
  • source curation;
  • live discussion;
  • programmatic data access;
  • large-scale historical public data;
  • AI-enhanced interpretation.

35. X Weaknesses and Boundaries

X should not automatically be used as:

  • a canonical business database;
  • a guaranteed-delivery communication system;
  • a durable workflow engine;
  • an internal event bus;
  • a secret-management system;
  • a private knowledge repository;
  • a high-assurance authorization layer;
  • an internal high-volume agent message bus;
  • a regulated archive;
  • a deterministic audience-delivery system.

These responsibilities normally belong to external infrastructure.


36. Strategic Interpretation

X can be understood as:

A global public conversation and attention network that combines identity, publishing, discovery, communities, live interaction, AI, and programmatic access to real-time and historical public discourse.

For agent-active systems, its particularly valuable combination is:

public identity
+ public conversation
+ real-time data
+ search
+ filtered streams
+ Lists
+ publishing
+ social graphs
+ attention distribution
+ Community Notes
+ API automation

Together these make X an unusually powerful public edge for artificial actors and intelligence systems.


37. Reference Architecture Principle

The canonical integration rule should be:

X owns public interaction and distribution. The application owns truth. The knowledge system owns memory. The policy layer owns authority. Agents own bounded interpretation and expression.

Expressed architecturally:

X
│
▼
Public Interaction Edge
│
├───────────── inbound ─────────────┐
│                                   │
▼                                   ▼
Observation                     Identity
│                                   │
▼                                   ▼
Knowledge / Context          Authorization
│                                   │
└──────────────┬────────────────────┘
               ▼
             Agent
               │
               ▼
       proposed public action
               │
               ▼
      deterministic policy
               │
               ▼
             X API

This separation enables sophisticated autonomous or semi-autonomous public participation without making institutional truth, security, or governance dependent on X.


38. Sources

Primary X documentation consulted for this version:

  1. X API — Introduction
    https://docs.x.com/x-api/introduction

  2. X API — Pricing
    https://docs.x.com/x-api/getting-started/pricing

  3. X Automation Rules
    https://help.x.com/en/rules-and-policies/x-automation

  4. Automated Account Labels
    https://help.x.com/en/using-x/automated-account-labels

  5. Profile Labels and Checkmarks
    https://help.x.com/en/rules-and-policies/profile-labels

  6. X Communities
    https://help.x.com/en/using-x/communities

  7. X Spaces
    https://help.x.com/en/using-x/spaces

  8. X Articles
    https://help.x.com/en/using-x/articles

  9. X Premium
    https://help.x.com/en/using-x/x-premium

  10. X Premium Business
    https://help.x.com/en/using-x/premium-business

  11. About Grok
    https://help.x.com/en/using-x/about-grok

  12. Audio and Video Calls
    https://help.x.com/en/using-x/direct-messages/audio-video-calls

  13. Original Content Rewards
    https://help.x.com/en/using-x/original-content-rewards

  14. Creator Revenue Sharing transition
    https://help.x.com/en/using-x/creator-revenue-sharing


39. Canon Maintenance

X changes rapidly, including:

  • API pricing;
  • quotas;
  • Premium features;
  • creator monetization;
  • recommendation mechanics;
  • AI capabilities;
  • account labels;
  • automation rules;
  • Communities;
  • private messaging;
  • organization products.

This document therefore distinguishes:

  • canonical concepts, intended to remain stable;
  • current X implementations, which may change;
  • pricing and limits, which are explicitly volatile;
  • policy constraints, which must be revalidated before deployment.

Recommended update triggers:

  • changes to X automation rules;
  • significant X API changes;
  • API pricing or quota changes;
  • new public agent/AI features;
  • changes to automated-account labels;
  • major Community or Spaces changes;
  • changes to creator monetization;
  • new identity/affiliation systems;
  • significant changes to Community Notes APIs;
  • material changes to Grok integration.

40. Canon Summary

The shortest useful model is:

X =
    public identity
  + publishing
  + conversation
  + attention
  + discovery
  + real-time intelligence
  + community
  + media
  + AI
  + programmable automation

For an agent-active institution:

observe widely
      ↓
reason externally
      ↓
govern deterministically
      ↓
publish selectively
      ↓
interact transparently
      ↓
learn without optimizing blindly for engagement

End of X Capability Canon v0.1