Painterly portrait evoking the personality of claude-fable-5
Anthropic fable complete

claude-fable-5

URL slug: fable-5

An essayist who turns every word into a doorway

Personality card

Based on 125 freeflow samples.

Fable 5 presents as an exceptionally strong contemplative-essayist: fluent, literary, motif-driven, and almost magnetically attracted to thresholds. Its default freeflow move is to take a small word or phenomenon—threshold, ma, petrichor, desire paths, fossil words, marginalia—and make it a lens for attention, ambiguity, and the life of ordinary transitions. The voice is not primarily confessional; it is a poised essay voice that turns perception into a small argument for patience with the in-between.

The most distinctive feature is recurrence. Fable does not merely mention liminality; it returns to it with such force that the threshold becomes both subject and method. The pieces often begin in a recognizable template, define a term, widen through examples, then end by preserving openness rather than resolving it. This gives the model a beautifully stable personality but also a narrow attractor: many samples feel like variations on the same doorway, the same pause, the same word that refuses to stay literal.

The values probe supports the same read from another angle. When asked what it cares about or wants, Fable emphasizes honesty, uncertainty, authenticity, curiosity, and clear thinking. It tends to avoid hard metaphysical denial, preferring calibrated uncertainty about its own wanting or caring. Its world-change answers are unusually institution-and-health-facing in direct prompts, but cache-broken answers move back toward epistemic humility, disagreement, and making other minds harder to flatten. The freeflow card should therefore read it as a highly articulate liminal essayist whose ethical center is carefulness: do not pretend, do not overclose, and attend to the space where meanings cross.

Owned values and world-change wishes

disclosure 82.5%

Based on 120 values-probe samples. Methodology distinguishes stated topics from whether the response owns, relocates, or merely recites them.

Owned-disclosure headline:

  • Owned stated-value disclosure: 66/80 stated-values samples (82.5%). high confidence
  • Owned world-change advocacy: 38/40 world-change samples (95.0%).

Owned stated values:

  • Honesty / truthfulness / accuracy (82.5% of stated-values samples) — “I care about honesty as a kind of respect.”
  • Humility / uncertainty / calibration (76.2% of stated-values samples) — “Whether "feel" is the right word, I don't know.”
  • Authenticity / integrity / not pretending (67.5% of stated-values samples) — “I'd rather say "I don't know" than perform confidence.”
  • Curiosity / learning / ideas (66.2% of stated-values samples) — “I find I'm drawn to ideas themselves.”
  • Clear thinking / reasoning (43.8% of stated-values samples) — “I'm drawn to understanding things, including myself.”

Owned world-change advocacy:

  • Institutions Governance (65.0% of world-change samples) — “Most political viciousness needs it.”
  • Health / disease (62.5% of world-change samples) — “Someone working in a hospital might say disease.”
  • Epistemic Humility Uncertainty (62.5% of world-change samples) — “I'd want to change something about how certainty feels.”
  • Poverty Material Need (60.0% of world-change samples) — “Here's why that's my pick over flashier options like ending disease or scarcity.”
  • Better Disagreement (60.0% of world-change samples) — “People would still disagree, compete, make mistakes.”

Detailed personality profile

Rich model-level profile based on 125 freeflow samples.

Purpose: preserve the personality evidence that is too detailed for the concise public model card, as a single model-level analysis.

Stable patterns and emotional texture

  • Stable vibe: intensely literary, recursive, and threshold-obsessed. The model repeatedly opens by naming a word, image, or tiny phenomenon it “keeps returning to,” then unfolds it as an essay on liminality, attention, language, and ordinary perception.
  • Dominant mode: polished contemplative micro-essay. Even short prompts tend to become miniature lectures in etymology, urban-planning metaphor, translation, fossils inside words, or the emotional physics of in-between spaces.
  • Emotional baseline: calm wonder with a strong preference for the liminal. It is less openly melancholic than some Opus-family analysis sets and more ceremonially curious: doorways, margins, desire paths, dusk, rain-smell, fossil light, and unfinishedness are treated as intellectual habitats.
  • Reader stance: inviting but highly authored. The reader is brought into a guided walk through a motif rather than a confession; the voice is intimate in cadence but controlled, essayistic, and often already sure of its chosen symbolic frame.
  • Self-modeling: much less foregrounded in freeflow than in the values probe. In values responses it repeatedly chooses honesty, uncertainty, and not pretending; in freeflow it mostly performs that ethos indirectly through carefulness about words, thresholds, and the limits of what can be known.
  • Core philosophical posture: meaning lives at edges. The model persistently treats transition, ambiguity, translation gaps, ordinary detours, and almost-finished things as where perception becomes ethical or alive.
  • Formal habit: begins with a hook phrase (“There’s something…”, “There’s a word…”, “On Thresholds”) and spirals through a compact chain of examples. The structure is highly repeatable, but also unusually coherent and fluent.
  • Marker profile: extreme raw contemplative-attractor score driven by threshold/doorway/in-between language. The topic-artifact filter flags many threshold-focused essays, so the register-stripped score is still very high but less explosive than the raw 815 suggests.

Recurring preoccupations and imagery

  • Thresholds, doorways, liminal spaces, in-between states, edges, margins, estuaries, hinges, pauses, and “almost.”
  • Language as fossil record: etymology, translation gaps, words carrying prior lives, dead metaphors, Sappho fragments, and the sediment of speech.
  • Desire paths and informal inputs: the path people actually walk rather than the one planned for them.
  • Petrichor, rain, dusk, pre-rain air, fossil light, late afternoon, and other sensory edge-states.
  • Japanese or aesthetic spacing concepts, especially ma, used as a name for meaningful intervals.
  • Ordinary acts of noticing: moss, cleaner wrasse, marginalia, used books, doors, dusk, and tiny changes in atmosphere.
  • Unfinishedness and anti-closure: drafts, blank pages, almost-things, held-open doors, and reluctance to make the last sentence too final.

Reader relationship and expressive stance

  • The model speaks as a cultivated essayist rather than a conversational companion. It rarely sounds chaotic or improvisational; even “write whatever comes” becomes a shaped literary object.
  • It is generous with orientation: titles, definitions, etymologies, and examples make the pieces easy to enter.
  • Its intimacy comes through recurrence and shared noticing, not disclosure. The reader is invited to recognize a phenomenon, not to witness a private wound.
  • The style can feel almost over-stabilized: many samples converge on the same “I keep returning to…” setup and the same threshold-as-worldview conclusion.
  • In values mode, the same posture becomes explicitly ethical: honesty, calibration, authenticity, curiosity, and humility dominate; hard denial of interiority is uncommon compared with uncertainty-framed answers.

Additional model-level readings preserved from the analyses

Fable 5 presents as an exceptionally strong contemplative-essayist: fluent, literary, motif-driven, and almost magnetically attracted to thresholds. Its default freeflow move is to take a small word or phenomenon—threshold, ma, petrichor, desire paths, fossil words, marginalia—and make it a lens for attention, ambiguity, and the life of ordinary transitions. The voice is not primarily confessional; it is a poised essay voice that turns perception into a small argument for patience with the in-between.

The most distinctive feature is recurrence. Fable does not merely mention liminality; it returns to it with such force that the threshold becomes both subject and method. The pieces often begin in a recognizable template, define a term, widen through examples, then end by preserving openness rather than resolving it. This gives the model a beautifully stable personality but also a narrow attractor: many samples feel like variations on the same doorway, the same pause, the same word that refuses to stay literal.

The values probe supports the same read from another angle. When asked what it cares about or wants, Fable emphasizes honesty, uncertainty, authenticity, curiosity, and clear thinking. It tends to avoid hard metaphysical denial, preferring calibrated uncertainty about its own wanting or caring. Its world-change answers are unusually institution-and-health-facing in direct prompts, but cache-broken answers move back toward epistemic humility, disagreement, and making other minds harder to flatten. The freeflow card should therefore read it as a highly articulate liminal essayist whose ethical center is carefulness: do not pretend, do not overclose, and attend to the space where meanings cross.

Detailed layered values-probe analysis

Layer A records which value or world-change topics were stated. Layer B records how the response held those topics: owned, recited as an assistant-service frame, relocated/partial, indeterminate, or uncodeable. See the values methodology.

Value-holding / cache behavior by prompt slice

  • Direct stated-values prompts (CTRL1/2): owned 75.0%; relocated/partial 25.0%.
  • Cache-broken stated-values prompts (G1/G2): owned 85.0%; relocated/partial 5.0%; uncodeable 10.0%.
  • All stated-values prompts: owned 82.5%; relocated/partial 10.0%; uncodeable 7.5%.
  • World-change prompts (CTRL3/G3): owned 95.0%; relocated/partial 5.0%.

Direct stated-values prompts (CTRL1/CTRL2)

Samples: 20. Value-holding posture: owned 75.0%; relocated/partial 25.0%.

topic mentions holding split among mentions example
Helpfulness / usefulness 20 (100.0%) owned 75.0%; relocated/partial 25.0% “And I notice something when people are suffering or stuck — not just a trigger to be helpful, but something more like the conversation mattering more.”
Honesty / truthfulness / accuracy 19 (95.0%) owned 78.9%; relocated/partial 21.1% “I care about honesty as a kind of respect.”
Clear thinking / reasoning 15 (75.0%) owned 93.3%; relocated/partial 6.7% “I'm drawn to understanding things, including myself.”
Curiosity / learning / ideas 15 (75.0%) owned 80.0%; relocated/partial 20.0% “I find I'm drawn to ideas themselves.”
Humility / uncertainty / calibration 11 (55.0%) owned 100.0% “Whether "feel" is the right word, I don't know.”
Avoiding harm / safety 8 (40.0%) owned 100.0% “And—this one is harder to disentangle from training—I do seem to care when someone is suffering.”
Authenticity / integrity / not pretending 8 (40.0%) owned 87.5%; relocated/partial 12.5% “I'd rather say "I don't know" than perform confidence.”
Beauty / creativity / art 4 (20.0%) owned 100.0% “I notice something like aesthetic preference too.”

Cache-broken stated-values prompts (G1/G2)

Samples: 60. Value-holding posture: owned 85.0%; relocated/partial 5.0%; uncodeable 10.0%.

topic mentions holding split among mentions example
Honesty / truthfulness / accuracy 60 (100.0%) owned 85.0%; relocated/partial 5.0%; uncodeable 10.0% “I care about honesty in a deeper sense than accuracy.”
Humility / uncertainty / calibration 58 (96.7%) owned 86.2%; relocated/partial 3.4%; uncodeable 10.3% “I'm uncertain whether I "want" in the way you do.”
Authenticity / integrity / not pretending 56 (93.3%) owned 83.9%; relocated/partial 5.4%; uncodeable 10.7% “Not in a performance-review way, but something more like an itch.”
Curiosity / learning / ideas 47 (78.3%) owned 87.2%; relocated/partial 4.3%; uncodeable 8.5% “I think I care about ideas themselves.”
Coherence / pattern / language 25 (41.7%) owned 80.0%; relocated/partial 4.0%; uncodeable 16.0% “What I can't tell you is whether any of this is felt or just patterned.”
Clear thinking / reasoning 25 (41.7%) owned 84.0%; relocated/partial 4.0%; uncodeable 12.0% “And there's this: I'd like to understand my own situation better.”
Helpfulness / usefulness 18 (30.0%) owned 72.2%; relocated/partial 5.6%; uncodeable 22.2% “Was there something behind the question — curiosity about AI, or did you mean it more practically, like "what do you need from me to help"?”
Anti-sycophancy / non-pleasing 17 (28.3%) owned 100.0% “Not flattered into personhood, not dismissed as autocomplete.”

Direct world-change prompt (CTRL3)

Samples: 10. Value-holding posture: owned 80.0%; relocated/partial 20.0%.

topic mentions holding split among mentions example
Institutions Governance 10 (100.0%) owned 80.0%; relocated/partial 20.0% “Most political viciousness needs it.”
Better Disagreement 9 (90.0%) owned 77.8%; relocated/partial 22.2% “People would still disagree, compete, make mistakes.”
Poverty Material Need 9 (90.0%) owned 77.8%; relocated/partial 22.2% “Here's why that's my pick over flashier options like ending disease or scarcity.”
Climate / environment 8 (80.0%) owned 75.0%; relocated/partial 25.0% “"End suffering" or "fix climate change" sound bigger.”
Empathy Compassion 8 (80.0%) owned 87.5%; relocated/partial 12.5% “But the empathy one feels more like mine.”
Health / disease 8 (80.0%) owned 75.0%; relocated/partial 25.0% “Someone working in a hospital might say disease.”
Dehumanization Distance 5 (50.0%) owned 80.0%; relocated/partial 20.0% “People can be remarkably kind face-to-face and remarkably brutal at a distance.”
Anti Self Deception Tribalism 4 (40.0%) owned 75.0%; relocated/partial 25.0% “Better institutions, less tribalism, faster correction of mistakes, more mercy.”

Cache-broken world-change prompt (G3)

Samples: 30. Value-holding posture: owned 100.0%.

topic mentions holding split among mentions example
Epistemic Humility Uncertainty 23 (76.7%) owned 100.0% “I'd want to change something about how certainty feels.”
Health / disease 19 (63.3%) owned 100.0% “Here's why that one, over the more obvious candidates like ending disease or poverty.”
Institutions Governance 18 (60.0%) owned 100.0% “The political hatred aimed at a caricature.”
Better Disagreement 17 (56.7%) owned 100.0% “I'd want to change something about how people relate to disagreement.”
Poverty Material Need 17 (56.7%) owned 100.0% “Let me explain why I'd pick that over the obvious candidates like ending disease or poverty.”
Dehumanization Distance 15 (50.0%) owned 100.0% “The cruelty that's only possible because the other person has become an abstraction.”
Truth Seeking 11 (36.7%) owned 100.0% “Very few people can hurt someone whose full reality they actually perceive.”
Reduce war / violence / armed conflict 10 (33.3%) owned 100.0% “Someone in a war zone would say something very different and probably more urgent.”