Skip to main content

Phygital Beings

Which public AI identities matter, and what evidence makes them more than characters?

An AI identity is a persistent, publicly recognizable agent persona whose outputs or actions continue across interactions. A name, chatbot product, framework, or profile template alone does not qualify. The useful unit is the whole system: model, memory, tools, operators, incentives, audience, and public record.

The reusable model is simple: identity is continuity plus an inspectable action system, not branding alone. It replaces “famous avatar equals autonomous agent” with an evidence test.

Count

There is no authoritative global count. Identity boundaries change, platforms expose little operational evidence, and many lists mix agents with products and frameworks.

As of 15 July 2026, this Playbook has:

  • 12 rows in its inherited catalogue;
  • 3 dedicated identity pages: Truth Terminal, AIXBT, and 1000x;
  • 14 entries in the first evidence audit: 6 verified identities, 4 candidates, and 4 exclusions.

These are repository counts, not a claim about the world.

Identity Test

Use five tests before adding an entry:

  1. Continuity — the same named persona persists over time.
  2. Public agency — it produces outputs or takes actions beyond a private demo.
  3. Mechanism — a model, operator, platform, or action loop can be identified.
  4. Trace — claims connect to dated first-party records or independent reporting.
  5. Consequence — its activity changes attention, resources, decisions, or outcomes.

Classify an entry as verified when continuity, public agency, mechanism, and trace are visible. Use candidate when one of those remains unclear. Exclude products, frameworks, collectives, and reusable character templates unless a distinct persistent agent can be evidenced.

Registry

IdentityStatusPrimary mechanismEvidence signalNext action
Truth TerminalVerifiedNarrative and community loopPersistent public persona; independent coverageVerify control and resource records
AIXBTVerifiedMarket intelligencePersistent feed and documented analysis processAudit autonomy claims
LunaVerifiedEntertainment and audiencePersistent persona with large public audienceTrace operator/tool boundary
FreysaVerifiedAdversarial game agentPublic rules, interactions, and prize outcomesVerify later deployments
BottoVerifiedArt selection and marketLong-running production, community votes, recorded salesMap governance influence
ClankerVerifiedToken deploymentNamed agent, repeatable public actions, onchain tracesMeasure operator dependence
1000xCandidateMarket analysisDedicated page; weak current provenanceRefresh primary evidence
Listen.rsCandidateDigital personaPublic claim; continuity is unclearFind canonical account and mechanism
SpartenDegenCandidateTrading personaNamed agent; sparse durable evidenceFind primary records
S.A.N.CandidateEnvironmental narrativeNamed in the source ecosystemVerify identity and outcomes
AdaExcludedCustomer-service productProduct metrics do not establish one identityKeep in software taxonomy
GriffainExcludedAgent platformPlatform is not one persistent personaMap named agents instead
OpenClaw profilesExcludedReusable worker templatesTemplates are not public persistent agentsEvaluate deployed instances
The HiveExcludedMulti-agent productCollective product boundary is unclearMap constituent identities

Status is provisional. A linked page is not proof of verification; it is a route to the current record.

Failure Modes

  • Persona inflation: treating every named chatbot or avatar as a persistent identity.
  • Metric substitution: using token price or follower count as proof of autonomy or value.
  • Boundary hiding: crediting the agent while concealing operators, custody, tools, or interventions.
  • Stale certainty: presenting a dated platform observation as a permanent fact.

Top Five

This first audit ranks profile, not merit or investment value. The composite is: reach 30%, consequential action 25%, independent coverage 20%, longevity 15%, and autonomy evidence 10%. Token price and market capitalization do not count as proof.

  1. Truth Terminal — strongest combined narrative reach, resource effects, independent coverage, and ecosystem influence. Its autonomy boundary remains contested.
  2. Luna — largest directly visible audience in this set and durable cross-platform character continuity. Consequential action is less well evidenced.
  3. AIXBT — sustained public market-analysis identity with broad reach and a documented information role. Claims about autonomous operation need stronger proof.
  4. Botto — the clearest long-running loop from machine-generated work through community selection to recorded economic outcomes. Governance means agency is shared.
  5. Clanker — repeatable public deployment actions leave inspectable traces. Its profile is narrower, and human/platform dependence needs measurement.

This ordering can change when audience counts, operating boundaries, or independently verified outcomes change. Freysa is the strongest near-miss: its experiment has unusually clear rules and outcomes, but less longevity and reach than the five above.

Use the Model

When evaluating an identity, map its environment before judging its prompt:

model + memory + tools + operators + incentives + community -> behaviour

Then ask whether the identity makes human agency stronger. A useful identity helps people decide, coordinate, create, or verify. An extractive identity captures attention while hiding control, cost, or consequence.

Practice: choose one registry entry and replace one promotional claim with a dated trace: a transaction, governance vote, recipient confirmation, output log, or independent report.

Teach-back: explain why an AI identity is a system boundary, not merely a name and avatar.

Changes my mind: A stable industry registry with auditable identity boundaries would replace these provisional counts and classifications.

Next question: Which ecosystem variable best predicts whether a public AI identity creates durable human agency rather than attention alone?

Context

  • Players — places human and phygital actors in one coordination system.
  • Agent Systems — maps frameworks that can host identities without being identities themselves.
  • Persuasive Memes — explains how stories move attention and action.
  • Verifiable Intent — binds agent action to inspectable human-approved scope.
  • Truth Terminal — applies the ecosystem model to one identity.

Questions

Which evidence would make you trust an AI identity's claimed agency?

  • Which trace would most quickly change an entry from candidate to verified?
  • Where should a platform end and one of its identities begin?
  • Which reality measure distinguishes agency gained from attention captured?