Pricing & Plans

Single work evaluation

One detailed evidence-based evaluation

• 12 criteria per evaluation from a bank of 237+, rotating across your repeated evaluations
• Evidence from your work
• PDF report

30-Day Journey

A 30-day intensive path in one art of your choice — no subscription required.

• 30 guided days in one art
• Daily calibrated exercises
• Completion certificate
• No subscription — a single one-time payment

Classroom Pack (10 Evaluations)

Ten evaluation codes to distribute to your students — each code redeems one full work evaluation.

• 10 independent evaluation codes
• Savings vs buying individually
• Instant distribution after purchase
• Valid for any of the eight arts

School Pack (30 Evaluations)

Thirty evaluation codes for a full class or more — each code redeems one full work evaluation.

• 30 independent evaluation codes
• Best per-unit price
• Instant distribution after purchase
• Valid for any of the eight arts

Creative Writing Foundations

A foundational creative-writing course: you write three scenes, each corrected against named criteria with a literal quote from your own text explaining every score.

• 5 modules, 15 lessons
• 3 scenes corrected against named criteria with literal evidence
• 12 comprehension questions
• Applied-skill certificate
• Permanent, non-expiring access

How to Read a Work of Art

Three hours that teach you to separate what you see from what you assume from what you judge -- so you can read any work, and read your own evaluation.

• 8 critical readings across five different domains
• Closing submission: turning a harsh critique into actionable feedback
• A ten-question classification test
• A "Reading a Work of Art" skill badge
• Fully included in SUB-PRO and SUB-ARTIST-PLUS

Principles of Composition

Two hours giving you six composition tools that work across drawing, photography, design, and craft -- measured on your own work with real numbers, not other people's examples.

• Your real work re-composed four different ways
• Visual center of weight and eye path measured with numbers
• An explicit comparison: which version has the clearest focus, and why
• 8 practical questions
• A Composition skill badge

Principles of Colour

Two hours that turn your color choices from silent taste into a palette extracted from your own work, with real numbers -- its contrast, its saturation, and whether it stays readable for every user regardless of their color vision.

• Two works analyzed with a real color palette from your own work
• A declared color intent before analysis, compared against reality
• Simulation of three types of color blindness -- a real accessibility feature
• A ten-question test, four of them computational
• A Color skill badge

Drawing & Painting Foundations

Eight to ten hours drawing four real observation exercises, with your work measured by a light-value map computed from the image itself -- not by impression. A two-model vision jury scores your final, certificate-bearing submission.

• Four real observation exercises (line, value, proportion, composition)
• A light-value map computed from your own work, not a human opinion
• Two images of your final piece's stages: before and after
• A two-model vision jury scores the final submission and lowers confidence on disagreement
• Ten questions, four of them practical
• An applied-skill certificate

Visual Design Foundations

Eight to ten hours working on four briefs you write yourself with explicit constraints (a poster, a mini identity, a UI screen, a long Arabic text layout), with your work checked by a deterministic WCAG-contrast and type-scale engine before any model gives its opinion.

• Four briefs where you write the audience, message, and constraint yourself
• A real WCAG contrast engine for every text region you mark on your own image
• A deterministic type-scale and alignment check, with zero model guessing
• Intent gap: the hierarchy order you intended versus what the eye actually read
• Twelve questions, four of them practical (compute a contrast ratio, fix a scale)
• An applied-skill certificate

Photography Foundations

Six to eight hours shooting five photos with constrained tasks (freeze motion, isolate a background, side light, minimal-elements composition, a three-shot sequence), while real EXIF data read from your own files judges each constraint automatically before any model gives its opinion -- and your location data is stripped from every photo before it is ever saved.

• A real EXIF reader extracting aperture, shutter speed, ISO, and focal length from your own photos
• A deterministic, numeric judgment of every technical constraint -- never a model's opinion
• Full GPS location-data stripping from every photo before it is ever saved
• An exposure-triangle (aperture/shutter/ISO) lesson with real physical accuracy
• Five photo tasks building up to a three-shot sequence scored by a two-model jury
• An applied-skill certificate

Screenwriting & Drama Foundations

Seven to nine hours writing three scenes in proper screenplay format, while a deterministic format checker inspects your scene heading, your character-cue placement, and your stage directions line by line with zero AI -- then a second model extracts your scene's beat map (goal, obstacle, turning point, exit) straight from your own text, each beat backed by a literal quote.

• A deterministic, zero-AI format checker inspecting the scene heading, character-cue placement, and stage directions
• A dramatic beat map (goal/obstacle/turn/exit) extracted with literal evidence from your own scene
• The difference between a format error (numeric certainty) and a structural weakness (disclosed-confidence inference) -- a professional lesson no other Arabic platform offers
• Three progressive scenes, the last one carrying an applied-skill certificate
• An applied-skill certificate

Voice Performance & Recitation Foundations

Five to seven hours recording four readings (a news item, a description paragraph, a dialogue, a poem) after marking your own intended pause points on the text -- your reading is automatically transcribed to real text and timestamps, and a deterministic engine compares your intended pause to your actual one, word by word. We say it plainly: we do not judge your voice's tone or articulation in this version.

• Real transcription of your reading via Whisper -- text and word-level timestamps
• A deterministic alignment engine comparing your intended pause to your actual one, word by word
• A pause-conflict table and a per-sentence pace timeline, both computed, not estimated
• A clear, repeated disclosure: no evaluation of vocal tone or articulation in this version
• Four progressive readings (news, description, dialogue, poem) against real voice-performance criteria
• An applied-skill certificate scoped specifically to read recitation

Crafts & Handmade Arts Foundations

Submit three photos for each of your three works -- raw material, in-process, and a final photo with a size reference -- so the making is judged, not just the photo, against criteria specific to your own craft sub-domain (ceramics, textile, wood, metal, leather, or ornamentation). An independent consistency check guards against gaming the process-documentation requirement, and there is never a judgment on structural durability or safety -- every visual assessment is disclosed as an estimate, not a geometric measurement.

• Criteria specific to your own craft sub-domain, from six Saudi and Gulf handmade crafts
• Mandatory documentation of all three making stages -- never judged from a final photo alone
• An independent consistency check guards against gaming the process-documentation requirement
• Heritage context constrained to a documented reference -- no AI cultural fabrication
• An explicit disclosure: a visual estimate of dimensions and symmetry, never a geometric measurement
• An applied-skill certificate named after your specific craft sub-domain

Arabic Calligraphy Foundations

Six to eight hours photographing five exercise panels (isolated letters, letters on a baseline, a word, a simple composition, a sentence) in one script of five you choose (Naskh, Ruq'ah, Thuluth, Diwani, Kufic), while a real deterministic engine detects your line's straightness with an actual geometric Hough transform, and measures your letters' pixel height against a reference dot you draw yourself -- compared against a historically documented traditional proportion table. Critical disclosure: this is an informative visual estimate, not a true dot-measurement and not a calligraphy license (ijazah); an ijazah is granted exclusively, orally, by a certified master.

• A real deterministic engine: a genuine geometric Hough transform detects your line's deviation from horizontal in degrees, never a generic description
• Real dot-proportion measurement: your letters' pixel height against a reference dot you draw yourself
• A historically documented traditional proportion table for each of the five scripts, shown for comparison
• Five exercise panels progressing from isolated letters to a full sentence
• A mandatory triple disclosure on every result screen: this is not a calligraphy license
• A knowledge certificate + a "Baseline Commitment" badge for a genuinely straight line

Responsible AI for Artists: Using It Without Letting It Make Your Art

Three hours that leave you with a personal disclosure statement written in your own hand, and the ability to judge eight real ambiguous cases with awareness -- not one correct answer the platform imposes, but a principle you hold to yourself.

• Eight real ambiguous cases you judge in writing, each with a carefully hand-authored counter-argument
• Measures the consistency of your own judgment across cases -- never corrects your opinion against a platform standard
• A defect-spotting exercise on a flawed AI output tagged with its exact flaws
• A publish-ready personal disclosure statement for your work, in Arabic and English
• A ten-question quiz (training / licensing / disclosure / style imitation), 80% to pass
• A declared-practice skill badge -- not an ethics certificate or a legal credential
• Fully included in SUB-PRO and SUB-ARTIST-PLUS

Music & Rhythm Foundations

Four to six hours learning rhythm with no instrument and no audio recording: you tap in the browser against a pulse the app itself draws, and your deviation from it is measured in milliseconds, beat by beat, across six graduated patterns (simple, compound, subdivided, and three documented Arabic rhythms: Maqsum, Masmoudi, Sama'i Thaqil) plus a written-notation reading exercise. Critical disclosure: this course does not evaluate your voice, your playing, or your notes. It measures timing only. Vocal-performance evaluation is not available on Zayenha Arts until further notice.

• A real deterministic timing engine: high-resolution timestamps from your own browser, with zero AI in the grading itself
• Mandatory device-latency calibration at the start of every round, subtracted from every measurement before it counts
• Six graduated tap patterns: simple, compound, subdivided, and three documented Arabic rhythms (Maqsum, Masmoudi, Sama'i Thaqil)
• A written-rhythm-notation reading exercise, converted to taps -- no audio at all
• A 16-question theory test with instant, deterministic auto-grading
• A knowledge certificate plus an optional timing-accuracy skill badge
• Never any audio upload -- every measurement comes from your own taps alone
• 40% off for SUB-PRO subscribers, fully included in ARTIST-PLUS

Poetry & Prosody Foundations

Seven to nine hours writing verses that are scanned automatically by a prosody engine written in code, not AI: meter, feet, permissible metrical variations, and the exact break position, across six graduated deliverables from scanning a given verse to fully free text. Meter is computation, not opinion -- one section is deterministic and non-negotiable, the other (imagery and diction) is explicitly disclosed as a confidence-scored model inference, always visually separated. Critical disclosure: the engine relies on full diacritization, and its absence lowers confidence rather than being silently guessed around.

• A real deterministic prosody engine: measures meter, feet, and permissible variations in code, zero AI at this step
• Pinpoints the exact break position by letter when no meter matches, never a vague general verdict
• Six graduated deliverables: scanning a given verse, completing a hemistich, two verses on a specified meter, a tafila fragment, free-meter (تفعيلة) text, fully free text
• A 20-question knowledge test (meters, feet, permissible variations, and identifying a break) graded deterministically
• Two sections always visually separated: meter (deterministic fact) and imagery/diction (confidence-scored model inference)
• A knowledge certificate plus a scansion accuracy badge

Principles of Rhythm & Timing

Two hours that show you the musical pulse, the poetic meter, and the spoken-line rhythm are one time-based phenomenon, not three separate languages: five tap exercises measuring your sense of time in milliseconds (steady pulse, subdivision, compound pattern, tapping a documented poetry verse's meter, and tapping a spoken sentence's rhythm), plus a representation-conversion exercise (you write a rhythmic pattern, then tap it out yourself). The exact same deterministic timing engine used in Music & Rhythm Foundations, zero AI in grading. The model only translates your numbers into your own art domain's vocabulary -- it never judges your performance or invents a number. Critical disclosure: this measures cross-domain timing only, not musical talent, recitation quality, or poetry itself.

• A real deterministic timing engine -- the exact same one used in Music & Rhythm Foundations, zero AI in grading
• Five graduated tap exercises spanning three domains: steady pulse, subdivision, compound pattern, a poetry verse's meter, a spoken sentence's rhythm
• A representation-conversion exercise: you write a rhythmic pattern yourself, then tap it out
• Translates the result into your own art domain's vocabulary (poetic foot for a poet, pause for a reciter, beat for a musician)
• An 8-question knowledge test with instant, deterministic auto-grading
• A knowledge certificate on passing the test and completing all five exercises
• Never any audio upload -- every measurement comes from your own taps alone
• Fully included in both the SUB-PRO and ARTIST-PLUS subscriptions, and purchasable standalone or as a gift

Full-Bank Deep Evaluation

Your work read against every activated criterion in your art -- not a sample. The criteria are split in code into four to six clusters by the skill each one measures, and every cluster is read in its own call with its full anchor text, because packing dozens of criteria and their anchors into a single window collapses the model's attention at the edges, which would make the difference between twelve criteria and forty a difference in count rather than in depth. A final aggregating call that never sees your work at all -- only the cluster outputs -- produces the through-line, the top three contradictions between clusters, and three ranked priorities. The contradictions section is the difference: when two clusters read the same work oppositely, that is real information about the work rather than noise, so we surface it instead of smoothing it away. Every quote is matched literally against your work's text before it is shown, and any cluster whose grounding falls below 60% is re-run once automatically and then declared low-confidence.

• Every activated criterion in your art -- not twelve selected ones
• Deterministic clustering in code by the seven-skill model: never random, never left to the model
• Each cluster in its own call with its full anchor text, so depth is measured rather than claimed
• An aggregating call that never sees your work -- only the cluster outputs
• A contradictions section: where two clusters read your work oppositely
• Three ranked priorities at the head of the report, not a task list as long as the criteria list
• A criteria-by-skill heat map, computed in code rather than by the model
• Independent confidence per cluster, derived from literal-grounding ratio rather than asserted
• The literal-quote guard: a quote not found in your work is dropped, and the drop is disclosed
• A limits-of-evaluation section, and an index of criteria codes with the anchor text matching your level

Brief-Anchored Evaluation

A work can be beautiful and still fail, because it did not do what was asked. Paste the competition text, the assignment, or the client request, choose your work, and we read it on two axes that are never mixed: how far it complies with the brief, and how good it is against our own criteria bank. The separation is architectural, not cosmetic — the model that extracts the requirements from the brief never sees your work at all, because a model shown both writes requirements shaped to fit the work in front of it and hands out a falsely high compliance score. Anything a tool genuinely cannot check — "the work must be original", "you must have attended the workshop" — is labelled NOT MACHINE-VERIFIABLE and excluded from the arithmetic. Our reading of the brief is an automated interpretation, not the organiser's position, and the report predicts neither acceptance nor rejection.

• Two separate axes: brief compliance, and artistic quality by criteria — never merged into one number
• The separation is executed, not described: requirement extraction runs in a call that never sees your work
• Every requirement is paired with a quote from the brief and evidence from your work, and quotes are matched literally in code
• NOT MACHINE-VERIFIABLE is a first-class outcome, not a failure bucket — named explicitly and excluded from the arithmetic
• Unmet requirements are listed first, because they are what actually blocks acceptance
• Two separately computed confidence figures — one per axis, derived from evidence grounding, never constants
• Our full reading of the brief is shown back to you so you can correct it if we misread
• Explicit disclosure: reading the brief is an automated interpretation, and the report is not a prediction of acceptance
• Included in the paid subscription at two reports a month, and purchasable standalone

Submission Readiness Gate

One explicit verdict on your work before you send it: ready to submit, needs a specific fix, or not ready. Two independent readings from models of two different families read the same mandatory criteria, then the code applies a rule published in advance -- no model is ever asked "is this ready?", because that answer is a mood wrapped in a number, and we could not then explain why two similar works got different answers. The rule: ready = every mandatory criterion at level 3 or above in BOTH readings · needs fixing = one or two mandatory criteria below 3 in both readings · not ready = three or more, or the two readings differ by two levels on a mandatory criterion. Every blocking criterion comes with the shortest available fix path and an effort estimate. Critical disclosure: this is not the organiser's decision and not a prediction of your acceptance or rejection -- it is our reading by our written criteria, and it issues no certificate. Restricted to ages 18 and over.

• A verdict, not a report left for you to interpret: ready to submit · needs fixing · not ready
• The rule is published before purchase and computed in code, not in the model -- you can hold us to it
• Two independent readings from models of two different families, on the same mandatory criteria
• A criterion blocks you only when both readings put it below level 3 -- one low reading is a warning, not a blocker
• The mandatory-criteria derivation is declared: the highest-weighted active default criteria for your art
• Every blocking criterion comes with the shortest fix path and an effort estimate (hour · session · rework)
• Your estimated effort compared against the time left before your deadline
• A confidence score computed from the agreement between the two readings, not a fixed number
• Every quote from your work is matched literally in code before display; anything unmatched is dropped and disclosed
• No certificate and no acceptance prediction -- a product decision, not a gap

Verified-Audio Evaluation

Record a 30-second to 8-minute vocal performance in Acting & Recitation (ACT), and it is actually transcribed with real timestamps by a Whisper model -- never a claimed transcript sent to the grading model while the audio itself is absent. We measure what is genuinely computable from the transcript and its timestamps (speaking rate, pauses and what they mean, breath-to-breath phrase length), and we tell you plainly, for every criterion that needs to hear your pitch, tone, or timbre and that we have no tool to measure yet: not measured. Every criterion has its required signal declared in advance in a fixed table in the code -- never decided by the model -- and a criterion without sufficient signal comes back "not measured" with its reason, never an invented score. Any timestamp the report cites is matched programmatically against the real transcript before it is shown; anything that does not match is dropped. Restricted to Acting & Recitation (ACT) today; Music (MUS) waits for a real audio-measurement layer before this reading is sold for it, because speech transcription does not apply to a non-verbal performance. The issued skill badge prints how many criteria were actually measured out of the art's total -- never a claim of full coverage.

Series Critique

Five to seven works in one art, read together rather than one by one: each work is scored on eight criteria fixed for the whole series -- comparison needs one ruler, not a different one per work -- then code alone, never a model, computes each criterion's mean, standard deviation, and trend slope across your declared chronological order. The highest-deviation criterion names your inconsistency; the highest-slope one names your genuine improvement -- both numbers a model is never allowed to invent. On top of that, one reasoning pass writes the recurring pattern, the structural weakness, and a description of your signature -- every pattern claim names three or more works by number, and anything seen in only two is labelled "weak signal", not a pattern. Plain disclosure: five to seven works is a small statistical sample, and a trend is a signal, not proof.

• Eight criteria fixed for the whole series -- no work picks its own
• Mean, standard deviation and trend slope computed in code per criterion, never by the model
• Highest deviation = your inconsistency, highest slope = your genuine improvement
• One signature paragraph, and three practice-level recommendations, not per-work ones
• Every pattern is grounded in three or more named works, or labelled a weak signal
• A 10-16 page PDF report
• One art of the eight -- five to seven works

Critique of Your Critique

Write your critique of a work of art -- and we measure your writing, not the work. Did you describe before you judged? Did you ground your judgment in evidence from the work? Or did you write beautiful prose with no anchor? Choose a work from the platform's public gallery or upload your own, write your critique (200-1200 words), and we read it against sixteen criteria written specifically to measure critical writing. A vision-and-text model checks something a text-only model cannot: are the descriptions you wrote actually true of the work? And code automatically classifies your sentences into description/judgment/interpretation and computes how much of your judgment is actually grounded in evidence.

• The platform's first place that measures taste itself, not an artwork as its byproduct
• A complete new criteria bank: 16 criteria with 5-level anchors, written specifically to measure critical writing
• Choose the work from the platform's public gallery (real works published by their own authors) or from your own works
• Vision-and-text verification: are the descriptions you wrote actually true of the work?
• Automatic classification of your sentences into description/judgment/interpretation, with ratios charted
• Evidentiary density: how much of your judgment is actually grounded in evidence from the work?
• One rewritten paragraph as a teaching example of grounding, with an explanation of what changed
• A "Critic" badge (CRD-08) after three qualifying critiques across at least two different arts
• Included in the paid subscription, and purchasable standalone

Music Performance Report

Record your playing or singing (45-180 seconds), naming your instrument and intended maqam. We extract your fundamental frequency over time, your note onsets, and your dynamic range -- all arithmetically from real audio samples, never guessed by a model. We rank the closest-fitting maqam to the scale degrees you actually used, and measure your pulse stability and cents-deviation from your declared maqam. Criteria that cannot be computed (expression, presence, beauty of performance) are actually heard by a model listening to your attached recording, never guessed. **Explicit disclosure:** the maqam-ranking and pitch-deviation thresholds in this version have not yet been calibrated against a human-expert-labeled reference sample -- they are this platform's own declared internal standard, not a figure from a completed calibration study. Matching a dum (bass) stroke against a tak (rim) stroke within the Arabic rhythmic cycle is not measured in this version (needs a timbral classifier not yet built) and is explicitly disclosed as "not measured" with its reason.

• Real fundamental-frequency (F0) extraction over time from audio samples -- never a model guess
• Note-onset detection and arithmetically computed pulse stability and drift
• Closest-maqam ranking across seven reference maqamat (Rast, Bayati, Hijaz, Kurd, Nahawand, Ajam, Saba) with a fit distance
• Cents-deviation from your declared maqam's degrees, at Arabic quarter-tone precision
• Dynamic range and technical recording quality, computed from the loudness envelope
• Explicit disclosure: maqam-ranking has not yet been calibrated against a human-labeled reference sample
• A skill badge printing how many criteria were actually measured out of the art's total

Series Critique (up to 12 works)

The same Series Critique, for a larger series: eight to twelve works in one art. The exact same three layers -- eight fixed criteria, statistics computed in code rather than by the model, and one reasoning pass grounded in named work numbers -- over a larger sample, which makes the computed trend more reliable without turning it into proof.

• Eight criteria fixed for the whole series -- no work picks its own
• Mean, standard deviation and trend slope computed in code per criterion, never by the model
• Highest deviation = your inconsistency, highest slope = your genuine improvement
• One signature paragraph, and three practice-level recommendations, not per-work ones
• Every pattern is grounded in three or more named works, or labelled a weak signal
• A 10-16 page PDF report
• One art of the eight -- eight to twelve works

Classroom Batch Evaluation -- up to 10 students

Upload your class's works as one batch -- each read on the exact same criteria. A class report showing collective weak points, and a private report per student. For one batch of up to 10 students.

• Pinned criteria set (8-12) for the whole batch
• Class report: per-criterion distribution, weakest criterion, three suggested exercises
• Isolated individual report per student -- no ranking, no comparison
• Up to 10 students in this batch

Classroom Batch Evaluation -- up to 20 students

Upload your class's works as one batch -- each read on the exact same criteria. A class report showing collective weak points, and a private report per student. For one batch of up to 20 students.

• Pinned criteria set (8-12) for the whole batch
• Class report: per-criterion distribution, weakest criterion, three suggested exercises
• Isolated individual report per student -- no ranking, no comparison
• Up to 20 students in this batch

Classroom Batch Evaluation -- up to 30 students

Upload your class's works as one batch -- each read on the exact same criteria. A class report showing collective weak points, and a private report per student. For one batch of up to 30 students.

• Pinned criteria set (8-12) for the whole batch
• Class report: per-criterion distribution, weakest criterion, three suggested exercises
• Isolated individual report per student -- no ranking, no comparison
• Up to 30 students in this batch

Classroom Batch Evaluation -- up to 40 students

Upload your class's works as one batch -- each read on the exact same criteria. A class report showing collective weak points, and a private report per student. For one batch of up to 40 students.

• Pinned criteria set (8-12) for the whole batch
• Class report: per-criterion distribution, weakest criterion, three suggested exercises
• Isolated individual report per student -- no ranking, no comparison
• Up to 40 students in this batch

Written Art Criticism Report

Write a critique of a reference artwork (250-600 words, structured as description -> analysis -> interpretation -> evidence-grounded judgment) -- and we measure your critique, with a real model committee: three independent AI models from three separate provider families score your text independently and in parallel against fifteen criteria, then we compute the median per criterion in code -- and show you the committee's agreement score explicitly. Any criterion where the models diverge by more than one level is shown as "disputed", never resolved by picking a side. The first product in the Arabic market to measure criticism as a skill, and the first to show a user how the models disagreed rather than hiding it.

• Three independent AI models score your text in parallel -- not one model
• The median per criterion is computed in code, and the committee's agreement score is shown explicitly
• Any criterion the models disagreed on is shown as disputed, never resolved to one side
• Choose the work from the platform's public gallery or from your own works
• Your critique's three strongest and three weakest sentences, with a literal quote as evidence
• A pre-written reference critique for comparison -- never generated from your own text
• An applied-skill credential after three qualifying critiques across three different arts
• Included in the paid subscription with a monthly cap, and purchasable standalone

Arabic Calligraphy Geometry Report

Photograph your practice panel, mark your reference dot and letters -- and we measure your calligraphy geometrically the way a teacher measures by the dot: alif height in dots, baseline straightness in degrees, pen-angle and spacing and stroke-thickness consistency, all computed deterministically from your own photo's real pixels, never estimated. Then a vision model judges what can't be measured geometrically: stroke flow, tashkil balance, and overall composition. The reference is named explicitly, and the measurement follows a declared platform standard -- this is not an ijazah, and a real ijazah is granted only orally by a certified calligraphy master.

• Alif-to-dot height computed deterministically from your photo -- never a visual guess
• Baseline straightness via a real geometric Hough transform, in degrees
• Pen-angle, spacing, and stroke-thickness consistency -- three new geometric measurements
• A vision model's judgment on flow, tashkil, and composition, grounded in evidence from your photo
• The traditional reference is named explicitly (which school's measurement), never a universal claim
• A drawn overlay layer showing the baseline and its deviation over your original photo
• A separate applied-skill certificate per script type -- never an ijazah
• Included in the paid subscription with a monthly cap, and purchasable standalone

Pitch & Interval Discrimination Test

Three listening tasks measure your pitch-discrimination threshold in cents (via a real adaptive procedure that gets harder or easier based on your answers), your accuracy naming 13 musical intervals from unison to the octave, and your ability to discriminate quarter-tones in real Arabic maqamat (Rast, Bayati, Saba) -- something no globally-available ear test built on the Western scale measures. Zero AI model in the measurement; every number is pure arithmetic.

• A pitch-discrimination threshold in cents, via a real adaptive procedure (Levitt 1971) -- not a fixed-difficulty test
• 13 standard musical intervals from unison to the octave, with a map of which ones you confuse
• Quarter-tone discrimination in three real Arabic maqamat -- Rast, Bayati, and Saba
• Zero AI model in the measurement -- every number is pure arithmetic
• A mandatory environment check (headphones and quiet) before the test starts
• A "Pitch Discrimination" skill badge after clearing a declared threshold
• Included in the paid subscription with a monthly cap, and purchasable standalone

Single-Criterion Deep Dive

Pick one criterion from the Arabic rubric bank, submit one work, and we read it against that criterion alone. You get your position across its five levels, why you are not at the level above and why not at the one below, three literal evidence quotes taken from your own work and verified programmatically, and a single exercise with its own success check. The value is not the number but the position: level 3 becomes a place between two understood bounds instead of an opaque grade. Built as the antidote to blanket evaluation — someone handed thirteen notes acts on none, someone handed one may act on it.

• One criterion you choose yourself from the rubric bank — searchable and filterable by art, skill and level
• All five levels shown in full, with your own level highlighted
• Why not higher and why not lower — two required fields that turn a grade into an understood position
• Three literal evidence quotes from your work, each verified against your text before it is shown as a quote
• One actionable exercise plus a success check that tells you when you have met it
• Explicit confidence disclosure: if two quotes fail verification the result is declared low-confidence
• The first card is free for every user
• Unlimited for paid subscribers

Model Panel Evaluation

Two independent readings of your work from two different model families (Google and Anthropic), against the same criteria and the same level anchors word for word, with neither seeing the other's verdict. Their agreement is then computed in code -- never asked of a model -- criterion by criterion: exact agreement, one level apart, or two or more levels apart. The final level is the median, not the mean, and overall confidence is a declared function of the agreement rates rather than a fixed number. Where the readings diverged, an arbiter is called to explain why they differed; it does not pick a winner, because two trained readings disagreeing on one criterion usually means the work itself supports two readings -- the most valuable thing a panel can tell an artist. Two disclosures we do not hide: models agreeing is not proof of correctness -- they can share a mistake; and the panel is two members today, not three, pending a verified third family id.

• Two independent readings from two different model families, with the same prompt and the same anchors word for word
• No member sees another member's output -- independence is a requirement, not a slogan
• Agreement is computed deterministically in code, never by a model: exact / one level apart / two or more apart
• The final level is a median, not a mean, and confidence is a declared function of the agreement rates
• A 'Where we disagreed' section: an arbiter explains the ambiguity in the work instead of picking a winner
• An evidence-concordance signal: did the two agreeing members cite the same place in your work?
• Every quote is matched verbatim against your text before it is shown; anything that fails is dropped and disclosed
• Explicit disclosure: model agreement is not proof of correctness, and the panel is two members, not three, until a third family id is verified

Calibrated Visual Discrimination Test

Forty comparisons between two compositions: an original and a copy in which exactly one controlled visual dimension was changed. The question is always specific -- "which is stronger in this particular dimension?", never "which is prettier" -- across five dimensions: balance, focal point, value range, eye-line, and negative space. Grading is deterministic with zero AI model, because we authored the modification and know its direction, and afterwards you review every task you missed with the actual modification explained. 15 of the 40 pairs are drawn from Arabic and Islamic visual traditions.

• Forty paired-comparison tasks against a reference key -- a performance test, not a self-report
• Five visual dimensions: balance, focal point, value range, eye-line, and negative space
• Zero AI model in the grading -- we authored the modification and know its direction
• A review of every task you missed, with the actual modification explained
• 15 of the 40 pairs drawn from Arabic and Islamic visual traditions -- not a translated Western bank
• A "Visual Discrimination" skill badge at 80%+ accuracy in four of five dimensions
• Included in the subscription with one retake every 60 days, and purchasable standalone

Revision Re-Evaluation

Fixed the notes on your work? Re-evaluate within thirty days for a token price — and we will not re-read what was already strong, only what we asked you to fix. The code, not the model, extracts from your original evaluation the criteria that scored 2 or below (usually two to five of them), then reads your revised version against those criteria alone, with their full anchors and the exact recommendation text we gave you — so it can state plainly whether you carried it out. The composite score is updated by merging the new scores with the untouched old ones, is labelled "partial", and every criterion that was NOT re-read is named. And a disclosure we do not hide: re-reading only the weakest criteria trends upward by construction.

• A deterministically narrowed scope: the code picks the criteria that scored 2 or below -- no model decides what gets read
• For every re-read criterion: was the original recommendation carried out? Answered plainly, without flattery
• A fresh literal quote from your revised version showing where it changed, programmatically verified before it is shown
• An updated composite score merging new with old, labelled partial, naming every criterion that was not re-read
• An explicit disclosure of the structural upward bias, inside the output itself
• Up to three re-evaluations per parent evaluation, within thirty days of it
• Free and uncapped for paid subscribers
• A result in about ninety seconds

Two-Version Comparison

Upload two versions of the same work, and describe in two to four lines what you changed and why. Both are measured against the exact same criterion identifiers -- literally reused from your prior evaluation of version one if one exists, or a single freshly-generated set applied to both otherwise. Version two is graded without the model ever seeing version one's scores, to prevent anchoring bias. The code, not the model, computes each criterion's delta, and every criterion that moved must carry a verified literal quote from BOTH versions or it is disclosed as "an unobserved change in the evidence" no matter what the number did. One final call -- the edit judge -- reads only your stated intent and the computed delta table (never the full versions) and classifies each criterion as: fixed as intended, partially fixed, no effect, or a negative side effect -- that last label is what no other tool says: when your fix solves one problem and creates another.

• The exact same criterion identifiers rerun on both versions -- never a fresh selection, or the comparison is a statistical illusion
• Version two graded without the model ever seeing version one's scores -- an anchoring-bias guard
• Every delta computed in code -- a model is never asked to state a difference itself
• A verified literal quote from BOTH versions for every criterion that moved, or the change is disclosed as unobserved rather than shown as an achievement
• The edit judge: did your revision actually fix what you meant to? Fixed as intended, partially fixed, no effect, or a negative side effect
• A dedicated section for negative side effects -- the one thing a cheerleading tool will never tell you
• Up to three comparisons per work, and three included monthly in a paid subscription
• A result within three to five minutes

Art Literacy Knowledge Certificate

An adaptive 20-item knowledge test in your chosen art, every question traced to a real reference article in the Zayenha encyclopedia. Multiple-choice items graded deterministically with zero AI calls. A score out of 100, a knowledge gap map across four axes (terminology & technique, tradition & historical context, reading a reference work, ethics & rights), and a digitally signed knowledge certificate at a score of 75 or above. Mandatory disclosure: a knowledge certificate. It does not say you excel at this art, only that you know its vocabulary and traditions.

• A deterministic adaptive test -- no model ever decides the next item's difficulty
• 20 multiple-choice questions per art, graded automatically with zero model calls
• Every question traced to a real reference article in the Zayenha encyclopedia
• A knowledge gap map across four axes
• An Ed25519-signed knowledge certificate at a score of 75/100 or higher
• Ten arts: the core eight plus poetry and Arabic calligraphy
• One attempt per art per month included in the Pro subscription, purchasable standalone, and giftable

Annual Measured Artistic Record

A full year of your activity in one digitally-signed document (Ed25519), independently verifiable off-platform via did:web -- a per-criterion timeline, the year's progress computed not described, your practice regularity computed deterministically from your own record (never a model's judgment), and a registry of exactly which AI model scored each work and when. An "Annual Measured Growth" credential is issued only when at least 6 evaluations are spread across at least 6 months; otherwise the full record still ships without one. Fixed disclosure: this is signed technical verification, not academic or professional accreditation.

• A full year, digitally signed with Ed25519, independently verifiable off-platform via did:web
• A per-criterion timeline and the year's progress computed in code, never merely described by a model
• Practice regularity computed deterministically from your own record -- never a model's taste judgment
• A model registry: exactly which AI model scored each work, and when
• An "Annual Measured Growth" credential at 6+ evaluations across 6+ months -- the full record ships either way

Verified Progress Report

No new upload at all -- built entirely from your own history: three or more completed evaluations in one art, at least thirty days between the first and last, and at least six criteria shared between the first evaluation and the last. Every number is computed in code from what is already stored -- no model touches a single figure; a deep-reasoning model only interprets the pattern: where progress is clearest, where things held steady or slipped, and the one constraint most worth focusing on next. A non-negotiable, honest condition: if the AI model, prompt version, or rubric wording changed between your first and last evaluation, the report discloses this explicitly -- the tool improving is never sold as you improving. A signed digital credential (Ed25519 via did:web) any third party can verify with no account.

• No new upload -- built entirely from your own already-stored evaluations
• Every delta computed deterministically in code across your full window, not just two endpoints
• Biggest-change window: the clearest jump between two consecutive evaluations, named by criterion
• Mandatory disclosure if the AI model, prompt version, or rubric changed between your measurements
• Your pattern interpreted by a deep-reasoning model that reads only the computed deltas, never the works
• A signed Ed25519 credential via did:web, with a public no-account verification page
• Included once every 90 days with a paid subscription, and purchasable standalone

Verifiable Work Provenance Card

A digitally-signed (Ed25519) work provenance card for a work you have already logged an AI-use disclosure for: one link, one image, and one QR code carrying your work's story -- checkable by anyone (a competition jury, an external platform, a school) with no account and no subscription. You control exactly what is shown: disclosure level, evaluation score, the work's timeline, and the card's language. The description paragraph is phrased neutrally and non-promotionally from already-computed fields -- no number or date ever reaches the model; every number is printed straight from the database. Your first card is free, and SUB-PRO subscribers get unlimited cards. You can revoke a card at any time -- its page stays up as "revoked by its owner" instead of vanishing. This is a provenance card, not an academic or professional credential.

Your Art's AI Tool Map

A "top 50 AI tools" list has no value: written for a different art, at a different stage than yours. Describe your actual workflow stage by stage, state your budget, device, and personal red line — what you refuse to delegate to any tool — and we match each stage to a real tool from our live-verified per-art reference, with an explicit verdict: allowed, allowed with disclosure, or crosses the authenticity line. The verdict is computed in code from the tool's own classification crossed with your red line, never a passing model opinion. Zero paid referral links anywhere in this family.

• Matches each stage of your workflow to a real tool from a live-verified reference -- not a generic list
• An explicit verdict per tool: allowed / allowed with disclosure / crosses the authenticity line -- computed in code from YOUR red line
• Never suggests a tool absent from our reference guide -- the catalog is injected into the call, never pulled from a model's memory
• Three impact-ranked recommendations, and an explicit list of what you don't need despite its fame
• An estimated monthly cost for the proposed stack, computed only from prices confirmed at each tool's own source
• Ready-to-copy disclosure text for every accepted tool
• Zero paid referral links, stated rather than assumed
• Included in the paid subscription at two maps a month, and purchasable standalone

AI & Artist Rights — Knowledge Test

A short test with published references measuring what you actually know about your rights in the age of AI -- not what you assume. Twenty-five questions: fifteen shared plus ten specific to your art, across five axes (ownership, training data, disclosure, consent and personhood, the Saudi PDPL system) -- every question carries its own published reference, so the test is reviewable and contestable. Questions are stored with fixed answer keys, never generated on the fly, and grading is deterministic in code, never left to a model. Pass threshold 70%, retake available after 24 hours with a rotated subset of the bank. A Knowledge Certificate on passing -- attesting to tested information, not applied ability and not a legal accreditation.

• 25 questions: 15 shared core + 10 specific to your art (acting/recitation, music/rhythm, photography, visual design, or the general module for other arts)
• Every question with its own published reference -- reviewable and contestable, never a bare opinion
• Deterministic code-based grading -- zero AI in the score itself
• A free 5-question diagnostic teaser with no certificate before you buy
• An explicitly-labeled Knowledge Certificate -- not an applied-skill certificate, not a legal accreditation
• 24-month certificate validity, and a 24-hour cooldown before retaking with a rotated bank subset

Spotting AI Output Flaws

Train to see what others miss in AI-generated output -- an image or a piece of text -- and measure your accuracy with a real number. You are shown a sample with its source hidden, you mark the suspicious spots and write your reason for each, then a model compares your analysis against a pre-tagged answer key -- never a free opinion -- producing a real accuracy score and false-positive rate. The badge needs at least 20 samples with accuracy of 70% or higher AND a false-positive rate of 25% or lower, together -- it is never awarded to someone who accuses everything. Mandatory disclosure: this measures a human discernment skill against a tagged sample bank, and is not a certificate of ability to detect all AI-generated content.

• At least 20 samples across four sessions to earn the skill badge
• Visual (drawing/photography/design) and text (writing/screenplay/poetry) with different samples per art
• AI semantic matching against a pre-tagged answer key -- never a free judgment from the model
• Accuracy, recall, and false-positive rate all computed in code from that matching
• Skill badge requires 70% accuracy AND 25% false-positive rate together -- never either alone
• 3 free samples per month, and included in the SUB-PRO subscription

Ethical Use Case Auditor

Describe your exact situation -- cloning a voice, training on a protected work, imitating an artist's style -- and get a reasoned verdict across six published dimensions: consent, third-party rights, transparency and disclosure, the work's human core, potential harm, and regulatory compliance. Three models from three different providers judge independently, and the final verdict (allowed / conditional / not allowed) is computed in code by one strict rule: any dimension that lands at the lowest level for at least two of the three models fails the whole case, no matter how high the other five score -- no averaging in ethical judgements. Every verdict names the condition that would flip it, and every place the models disagreed is stated explicitly, never hidden behind one number. This is an ethical and professional reading based on published standards -- not legal advice and not a religious ruling (fatwa).

• Six published ethical dimensions, each with five predefined levels -- never a general impression
• A three-model committee from three different providers (Google, Anthropic, OpenAI) judging fully independently
• One strict rule, no exceptions: a single dimension failed by two models fails the whole case -- no averaging in ethical judgements
• The condition that would flip the verdict is written explicitly in every report, never left to guesswork
• Every place the models disagreed is named, never hidden behind one consensus number
• Zero "allowed" verdicts on a case that itself declares missing consent -- a structural code rule, not a model's opinion
• A clear, mandatory disclaimer: an ethical and professional reading, not legal advice and not a religious ruling
• One free case per user, and three cases a month included in SUB-PRO

Tool Terms & Output Rights Reader

You sign a terms-of-use agreement you never read, and later discover the tool claims a right to your output, trains its models on what you upload, or forbids commercial use on its free plan. Paste the tool's actual terms or privacy policy text, state your purpose (learn/sell/client work/teach), and get an eight-field rights card -- every field backed by a literal quote from YOUR pasted text, and any field with no quote shown honestly as "not found in the text" rather than guessed. The suitability verdict for your purpose is computed in code from these fields, never a passing model opinion. One free check for everyone.

• Eight rights fields: output ownership, training on your input, opt-out, commercial use, attribution, legal liability, jurisdiction, and free-vs-paid plan differences
• Every field backed by a literal quote from your own text -- any field with no verified quote is shown honestly as "not found" rather than guessed
• An explicit suitability verdict for your purpose (suitable / suitable with caution / not suitable), computed in code from the fields, never a model's opinion
• A comparison with two alternative tools from our own live-verified reference for your same art
• A practical checklist of what to do before you use it
• One free check for everyone, and five a month included in the paid subscription
• A clear disclaimer: a structured reading of text YOU supplied, not legal advice

Augmented Creative Workflow Advisor

The hard question is never "should I use AI?" but "at exactly which step, and where do I stop?" -- a decision usually made under delivery pressure, where use quietly expands until it swallows the core with no conscious decision point. Describe your project, its deadline, your declared level, and your purpose (learn/competition/client work/exhibition), and paste the beneficiary's own constraints if any (competition rules or a client contract excerpt). We break your project into stages, each with a clear delegation verdict (delegated / delegated with disclosure / not delegated) plus the exact condition that would flip it, and a "your human core" section for this specific project. External constraints are applied as hard rules in code, overriding any verdict that conflicts with them -- never left to the prompt alone to enforce a real competition's actual rules. This is a self-commitment you approve yourself, not a legally binding contract.

• Breaks your project into 5-9 stages, each with its deliverable and what must stay in the artist's own hands
• A three-way delegation verdict per stage (delegated / delegated with disclosure / not delegated) plus the exact flip condition
• Beneficiary constraints (competition rules or a client contract) applied as hard rules in code -- never left to the model's discretion
• 6-12 workflow-contract clauses, procedural and verifiable, never preachy
• A ready-to-copy disclosure text for delivery
• An explicit label: a self-commitment, not a legally binding contract
• Included in the SUB-PRO subscription with a two-per-month cap, or bought standalone

Responsible AI Practice in Art — Applied Skill Certificate

An Applied-Skill Certificate -- earned not by watching a lesson but by practicing, disclosing, judging, and having that measured with evidence. Four conditions together: passing the AI-Output-Flaw-Spotting badge (AIA-06), three works with completed disclosure, one of them at process-authenticity level 3/5 or higher (AIA-08), five ethical scenarios answered with written reasoning against a published grading key, and a personal practice statement. A three-model, three-provider committee judges only the two subjective elements -- and any criterion scoring below level 3 with two of three judges blocks the grant entirely, no matter how high the rest score. Ed25519-signed Open Badges 3.0 credential, valid 24 months, with a public verification page and a revocation list. Mandatory disclosure: a responsible-practice skill, not an artistic one and not an official accreditation.

• Four conditions together: the AIA-06 badge, three disclosed works with one at AIA-08 process level 3/5, five reasoned ethical scenarios, and a personal practice statement
• The badge and works conditions are fully deterministic, code-only checks -- zero model involvement, records not opinions
• A three-model, three-provider committee judges only the two subjective elements, against published grading keys
• One criterion failing with two of three judges blocks the grant entirely -- no average can cover a critical failure
• Ed25519-signed Open Badges 3.0 credential, valid 24 months, with a public verification page and a revocation list
• Included in the ARTIST-PLUS subscription, giftable, and available in institutional packages

Classroom & Studio AI Policy Builder

A teacher faces two losing options: a blanket ban nobody follows, pushing students to hide it, or silent tolerance that makes grading arbitrary -- with nothing to write a policy from: no Arabic template, no rubric, no student disclosure form. Enter your organization's type, arts, age group, and philosophy (ban/restrict/regulate/guided openness), and we derive four usage levels tied to AIA-07's five disclosure levels, procedural and verifiable policy clauses, a five-level grading rubric, a ready disclosure form, a guardian letter, and a teacher checklist -- in Arabic and English together. Age-band restrictions (13-17) and PDPL compliance are applied as hard rules in code, never left to the model's discretion. This price is for a single studio or one art taught -- for a multi-art school, see `AIA-14-SCHOOL`. Explicitly labeled: an editable draft the organization adopts itself, not a ready-made legal document.

• Four usage levels tailored to your organization, tied to AIA-07's five disclosure levels
• Procedural, verifiable policy clauses -- any clause with no automatic platform check is explicitly labeled "requires manual verification"
• A five-level rubric for student compliance with the policy
• A ready disclosure form, a guardian letter, and a teacher checklist
• Age-band restrictions (13-17) and PDPL compliance applied as hard code-side rules
• Arabic and English together, equal quality
• For a single studio or one art taught -- for a multi-art school, buy AIA-14-SCHOOL

Classroom & Studio AI Policy Builder -- Multi-Art School

The same AI-use policy builder, priced for a school teaching two or more arts with per-art units -- see `AIA-14`'s own description for full details.

• All of AIA-14's features
• Priced for a school teaching two or more arts
• A dedicated unit per art taught, when needed

Creative Process Authenticity Report

We don't judge whether a machine helped with your work -- we measure how much of its making journey is documented by evidence: sequential versions, stage snapshots, decision notes, and an AI-use disclosure. Deterministic, zero-AI checks run first in code (version count, time spacing, a suspicious-jump detector) and work even if every AI call fails, then one vision reading of the visual stages and one reasoning reading of your decision notes -- merged in code, never by a model. Reaching level 4 of 5 in record completeness earns a "Complete Creative Process Documentation" credential -- never a "human-only work" certificate, and it never will be.

• Deterministic, zero-AI checks first: version count, time spacing, a suspicious-jump detector -- work even if every AI call fails
• One vision reading of your stage-snapshot sequence + one reasoning reading of your decision notes, merged in code
• A "Complete Creative Process Documentation" credential at level 4/5 -- never described as a "human-only work"
• A mandatory "what this report cannot say" section on every report
• Explicit disclosure whenever the two model readings differ by more than one level

14-Day Path: The Artist in the Age of AI

Previous products are separate moments, but a stance on AI is a habit you build. Fourteen days, one exercise a day under twenty minutes, fully designed for YOUR art -- not just relabeled: pure observation exercises, real judgment scenarios, three works with completed disclosure, one real Critical Prompt Lab session (AIA-05), and one ethical case before a three-judge panel (AIA-11). You leave with a personal practice statement built from your own answers, not a generic template, and a completion credential -- an attendance-style credential for finishing the path, never a skill or knowledge certificate.

• Fourteen days, twenty minutes a day, with exercises that genuinely change across all eight arts -- not just relabeled titles
• Three works with completed disclosure, alternating with pure-observation exercises and real judgment scenarios
• Day five is a real Critical Prompt Lab session (AIA-05); day twelve is a real case before the Ethics Committee (AIA-11)
• A minimum of 600 characters per day, enforced by an automated publish-blocking guard -- not a repeat of the platform's existing thin plan content
• A personal practice statement comparing your day-one answer to your day-fourteen answer, not a generic template
• A path-completion credential -- precisely labeled: never a skill or knowledge certificate -- included in the SUB-PRO subscription, and giftable

Group Disclosure Compliance Report

One dashboard for your institution: who disclosed their AI use, who documented their process, and where a human look is needed -- without accusing anyone. The vast majority of the report is computed in code with zero AI involved; one small model only summarizes group-level patterns and suggests educational interventions. A guard enforced in code, not a prompt instruction: the model is structurally barred from ever seeing a student's name or work, and never issues a verdict on any individual -- individual cases are shown with raw numbers only under "needs a look," and the educational decision belongs to your institution's own staff alone.

• The vast majority of the report is a deterministic, zero-AI calculation -- disclosure rates, documentation completeness, comparison against your thresholds
• A guard enforced in code: no generated verdict on any individual, ever -- individual cases shown with raw numbers only
• Group-level-only patterns and intervention suggestions, from a model structurally unable to see a student's name or work
• Full CSV/JSON export of the record
• Included free under the institutional subscription (ENT-INSTITUTION)

Evidence-Based Artistic CV

Today's artist bios are either empty or inflated, and no tool separates what is verified from what is merely claimed. This is a bio split structurally, not cosmetically, into two sections: a VERIFIED section pulled automatically from your completed evaluations, earned certificates, and finished learning plans on this platform -- every line backed by a real number or date, never exaggerated. And a SELF-REPORTED section for external activity you add yourself (an exhibition, a workshop, a competition, a publication, a commission), scanned by an inflation guard that rejects words like "certified", "professional", "expert", or "award-winning" you did not actually earn here -- and discloses every rejected phrase instead of silently deleting it. This product measures nothing and analyzes nothing -- it separates fact from claim, and that alone is what makes your bio believable to a jury reading hundreds of files.

• Two visually separated sections: verified from the platform, and self-reported from your own entry -- never merged
• An inflation guard that rejects unearned "certified"/"professional"/"expert"/"award-winning" claims and discloses every rejection instead of hiding it
• Every verified line is tied to real evidence: a completed evaluation, an issued certificate, or a finished learning plan
• External items are classified into standard CV categories (exhibition/workshop/competition/publication/commission) with unified bilingual phrasing
• A short one-page version and a full version, plus a ready-to-paste plain-text export
• The cheapest paid product in the portfolio family, because it is the lightest to reason about -- not the least valuable
• Included in every paid subscription with no monthly cap, and purchasable standalone

Portfolio Coverage Map

We show you exactly which parts of your art you haven't touched in any work yet -- a heat map of every active criterion in your art against three states: shown at a good level, shown at a weak level, or never measured. No AI in computing the map itself -- a deterministic query over your own existing works and evaluations. For up to four of the most important never-measured criteria, we add a suggested exercise spec: what your next work should contain to show that specific criterion, and what constraint makes it a real challenge -- describing the condition, not the work itself. Coverage percent is not a skill score: covering everything at a weak level is worse than covering less with mastery.

• A no-AI heat map -- fully deterministic from your own stored works and evaluations
• A three-way classification for every active criterion in your art: good level, weak level, or never measured
• A suggested exercise spec for four of the most important unmeasured criteria -- the condition, not a finished work
• An explicit disclosure: coverage percent is not a skill score
• The fastest product in the portfolio family -- under two minutes from request to result
• Your first map is free, and every paid subscription includes unlimited maps

Visual Identity Consistency Analysis

We show you exactly what genuinely repeats in your visual works (your style) and what appears only once (noise) -- across eight visual descriptors from a closed vocabulary that includes terms from the Arabic/Islamic visual tradition (flattening, arabesque, space-filling, linear outline), not the Western tradition alone. Seven of the eight descriptors are extracted per-work by a vision model constrained to that closed list; the eighth (color family) is computed deterministically with no model at all. The aggregation and verdict -- your signature, the noise, the work furthest from your center, exploring vs. converging vs. scattered -- all happen in code, never in the model. Explicit disclosure: consistency is not quality, and perfect consistency across all eight descriptors usually means you have not tried different approaches yet, not that you have achieved mastery.

• Eight visual descriptors from a closed vocabulary, including Arabic/Islamic visual-tradition terms, not only Western ones
• Per-work visual extraction constrained to the closed list -- never invents categories
• Aggregation, consistency, and verdict all computed deterministically in code, never by the model
• Your real signature, the noise, and the work furthest from your center with why
• Explicit disclosure: perfect consistency for a beginner may mean under-exploration, not mastery
• A recommendation built from your declared intent: a unified style, or deliberate exploration?

Written Voice Fingerprint

We surface your recurring writing moves and crutches across five to twelve texts by count and quote, not a general model's impression -- what genuinely recurs across your texts is your fingerprint, not a note on one piece. A comparison between your oldest and newest texts shows which habits persisted and which changed. Your declared reference tradition, if you name one, stays your own self-description, never a model-issued comparison.

• Aggregation across 5-12 texts, never one text in isolation
• Genuinely recurring moves and crutches -- not a one-off note on a single text
• Oldest-vs-newest comparison: which habits persisted, which changed
• Your declared reference tradition stays your own description, never a model-issued comparison
• Reuses the already-verified narrative-voice engine, not a new one

Official-Track Readiness File

We match your evidence against a bank of real Saudi official tracks (competitions, grants, academic programs) item by item through the exact same deterministic engine the submission-readiness report uses, plus an 8-question procedural-understanding quiz and automatic disabling of any track whose content hasn't been reviewed within 120 days. A fixed, non-hideable disclosure: this is self-preparation, not an official accreditation from any body.

• Item-by-item matching of your evidence against a bank of real Saudi official tracks
• The exact same deterministic engine the submission-readiness report uses -- no new engine
• Automatic disabling of any track whose content hasn't been reviewed within 120 days
• An 8-question procedural quiz generated solely from that track's own text
• A fixed disclosure: this is self-preparation, not official accreditation from any body

Artist Statement: Extraction, Not Generation

We will not write your statement. We extract 3-5 genuinely recurring motifs from 5-8 of your works (a recurring theme, a recurring formal choice, what is persistently absent) -- every motif backed by at least two works or it is dropped. We compare your answers to a 12-question interview against what your works actually show, phrasing up to 3 evidence-backed contradictions as questions, never accusations. You get incomplete sentence stems to finish yourself -- structurally capped so they can never become a finished sentence, no matter what the model tries. Then, once you have written your own draft in a second round, we critique it against six fixed criteria without rewriting a single line of it.

• Extracts 3-5 genuinely recurring motifs from your works, each backed by at least two works or dropped
• Compares your 12-question interview answers against what your works show, phrasing up to 3 contradictions as questions, never verdicts
• Incomplete sentence stems only -- structurally length-capped, cannot become a finished sentence
• A second round after you write your own draft: critiqued on six fixed criteria without rewriting a single line
• Checks the equivalence of your Arabic and English versions -- the system checks, you write both languages
• A fixed disclosure on every report: this statement is your own writing, and the system did not write a single sentence of it

Practice Provenance Record

A dated record of your work's stages -- from draft to final (at least 3 stages) -- with a mandatory production-method declaration per stage. A visual/textual continuity check compares each stage to its context (never asking "was this AI-generated"), plus a deterministic, zero-AI layer in code measuring the real time gaps between your stages and their regularity -- the strongest signal here, no model involved. Stages uploaded within an implausible time window are explicitly labeled "no real time extension." A cryptographic fingerprint for the record, and an optional public verification link. This is a practice record, not a certificate that the work is not AI-generated -- we do not claim generation detection.

• A dated stage record (3+ stages) with a mandatory closed-list production-method declaration per stage
• A visual/textual continuity check across your stages -- never asks whether AI was used
• A deterministic, code-only layer: real time gaps between stages and their regularity -- the strongest signal, no model
• A disclosure audit comparing your own declaration against what the stages actually show -- for you first, never public
• Stages uploaded within an implausible time window are explicitly labeled, never hidden
• A cryptographic fingerprint for the record + an optional public verification link
• A fixed statement on every record: this is a practice record, not a certificate of non-AI-generation

Verified Longitudinal Progress Report

We re-score your older work right now with the SAME current model and rubric text used to score your newer work, isolating model drift from artist drift before measuring any progress at all. The step no competitor takes: anyone promising AI-measured growth is comparing two scores issued by two different models and calling the difference "improvement" -- we re-score and disclose the drift amount explicitly. Requires at least three evaluations in the same art, spaced at least sixty days apart, on two independent works (never two versions of the same work). The report shows each criterion's delta by LEVEL, not percentage, plus every criterion that did NOT move (never a gains-only report), a confidence score, and exactly one suggested constraint for the next period. A "you have not moved" verdict still ships in full, unsoftened, when that is what the data shows. A "Verified Progress" certificate is issued only when a specific NAMED criterion shows a positive delta -- and it is never described as mastery.

• Your old work re-scored right now with the same current model and rubric text -- the one condition that makes any growth claim valid
• Every criterion's delta by LEVEL, not percentage, with two verified pieces of evidence per criterion that moved
• Criteria that did not move are always shown -- a gains-only report is advertising, not a report
• The isolated model-drift amount is always disclosed, even when zero
• One confidence score and exactly one suggested constraint for the next period, not thirteen tips at once
• A parent-friendly, non-technical version and an institutional version -- same numbers, different phrasing only
• A "Verified Progress" certificate when its condition is met -- it verifies measured movement, never mastery

Model-Committee Portfolio Review

Three independent readings of your portfolio (6-12 works) against published criteria: a first reading from google/gemini-2.5-flash, a second, genuinely independent reading from anthropic/claude-haiku-4.5 -- a deliberately different provider, with no access to the first reading -- then a synthesis that shows what they agreed and disagreed on explicitly, plus one governing constraint with two next actions. The confidence score is actually computed from how much the two readings agree, never guessed. Covers scenic writing, screenwriting, drawing, photography, visual design and crafts -- and explicitly excludes music and acting (we do not judge a work we cannot hear). A fixed statement on every report: there is no human expert in this product -- the human layer is deliberately deferred, never silently absent. An "Applied Skill" badge is issued only when both readings exactly agree on an advanced level across two independent works.

• Three independent readings against published criteria -- never one review relabeled as three
• Two readings from deliberately different providers, each with zero access to the other's output
• An explicit agreement/disagreement matrix per criterion per work -- never one blended verdict
• One governing constraint with exactly two next actions -- not a scattered list of a dozen notes
• A confidence score actually computed from the real agreement rate -- and how it was computed is shown to you
• A mandatory disclosure of automated-evaluation limits from peer-reviewed literature -- never hidden from you
• One unambiguous fixed statement: no human expert here -- the human layer is deferred, never silently absent
• An "Applied Skill" badge under a strict condition: both readings agreeing on an advanced level across two independent works

Verifiable Artist Page

A public page at arts.zayenha.com/a/your-handle where every number links to real proof: the certificate or evaluation it came from, and its date. Your bio is automatically screened before publishing -- any sentence not backed by a real certificate or evaluation is published labeled "self-reported, unverified," never silently deleted or hidden. Included free with Pro and Artist Plus; this is the standalone annual license price for non-subscribers.

• Every number links to a real verification of your certificate or evaluation
• Automatic truth-check on your bio before publishing -- unbacked claims are labeled, never silently deleted
• Automatic privacy scan blocks personal-data leaks
• QR code and a shareable social card
• Abuse-report queue with automatic threshold blocking

Design Brief Response Certificate

You receive a real brief with explicit constraints -- audience, one message, a color constraint, a size and aspect ratio, and prohibitions -- and are measured on what actually measures a designer: whether your design solved the problem, not merely whether it looked good. You submit three sequential stages (initial direction, refinement, final delivery) and a written rationale of at least 150 words. The final score is composed of two layers explicitly disclosed to you: half is pure computed certainty (WCAG contrast, size/aspect match, color count, required-element presence -- zero AI) and half is a vision model's judgment, grounded in evidence coordinates inside the image itself; any evidence falling outside the image bounds drops that dimension to "unresolved" instead of fabricating a verdict. A signed certificate explicitly separates "fitness for purpose" from "aesthetic quality", plus an item-by-item compliance report with zero discretion. Launching with two tracks only for now: promotional print and a mini brand identity.

• A real brief with explicit constraints, not a free exercise
• Three sequential stages: initial direction, refinement, final delivery
• Zero-AI deterministic checks: WCAG contrast, size/aspect match, color count, required elements
• A vision model scoring six dimensions grounded in in-image evidence coordinates -- out of bounds means unresolved
• A temporal comparison across your three deliverables: did you actually respond to the constraint?
• The ratio between computed certainty and model judgment is disclosed to you explicitly, never hidden
• A certificate separating fitness for purpose from aesthetic quality -- never one blurred number
• An item-by-item compliance report: pass/fail per constraint, zero discretion
• Two launch tracks only for now: promotional print and a mini brand identity

Verifiable Progress Record

A signed document that says: this was your level in your art on this date, this is your level today, and this is exactly what changed -- with evidence from your own work, not a claim. Built entirely from what the platform already has recorded about you (your skill history and evaluations), with no extra input required. An honest eligibility bar before any payment: at least 90 days between your first and last measurement, at least six evaluated works in the same art, and the record is never issued if it does not reach a genuine minimum of real change -- check your eligibility for free before buying.

• Built entirely from what is already stored -- no extra input from you
• A free eligibility check before any payment: know your result before you pay
• The record is never issued unless it reaches a genuine minimum of real change
• Every delta is computed deterministically in code -- no model touches a single number
• Your pattern is interpreted by a deep-reasoning model that reads only the deltas, never the works
• A visual before/after comparison + a verification link any third party can read, no account needed
• Never giftable -- your own personal record only

Applied Skill Certificate — Evidence Portfolio

You complete five real works in response to a platform-issued brief (no pre-existing old works), and are measured on 12 criteria by a committee of three genuinely independent models -- not a marketing name: anthropic/claude-sonnet-4.5 for deep reasoning, google/gemini-2.5-pro for vision, anthropic/claude-haiku-4.5 as a cheap third anti-drift voice. The merge is entirely deterministic in code: the median is the score, and any criterion where the committee's judgment spread by two levels or more is explicitly marked "unresolved" and never enters the certificate at all -- never hidden, never smoothed into an average. The bilingual criteria narrative is written from the merged scores only, never from your works, so it never re-issues the judgment. A signed certificate carries each criterion's resolved level with its evidence and the committee's agreement score, with full disclosure of the three models' names, versions, and the automated evaluation's documented limits -- this is **never accreditation and never an expert review**, in any form.

• Five real works in response to a platform-issued brief -- no pre-existing old works
• A committee of three genuinely independent models at temperature 0.1, not one model wearing a marketing name
• The merge is entirely deterministic in code: the median is the score, the spread decides confidence
• Any criterion with two or more levels of spread is marked unresolved and never enters the certificate -- never hidden
• The criteria narrative comes from the merged scores only, never your works -- it never re-issues the judgment
• Full disclosure: the three models' names, versions, dates, and the automated evaluation's documented limits
• Never accreditation and never an expert review, in any form -- an honest measurement record only
• An optional public evidence portfolio with your explicit consent (18+ only)

Art Fundamentals Knowledge Certificate

An adaptive knowledge test (24 items in its full version, 8 in the free diagnostic) measuring what you know about your art's rules -- not your talent, not your work's quality. Multiple-choice items are graded deterministically with no model; short judgment items ("which of these two works meets a given criterion, and why?") are graded by comparing your answer against a published reference key. A score out of 100, a gap map across five knowledge axes, and a digitally signed knowledge certificate if you score 70 or above. Mandatory disclosure: a theoretical knowledge certificate. It says nothing about your ability to execute.

• A deterministic adaptive test -- no model ever decides the next item's difficulty
• Multiple-choice items graded automatically with zero model calls
• Short judgment items graded against a published reference key per item
• A gap map across five knowledge axes for each art
• An Ed25519-signed knowledge certificate at a score of 70/100 or higher
• A free 8-item diagnostic version, with no certificate, inviting the full test
• Included in the Pro subscription with a monthly quota, purchasable standalone, and giftable