#Challenge Styles Reference

A comprehensive reference for every challenge style, format, and difficulty level in the Eko challenge system.

#Abbreviations

#Style Abbreviations

Abbreviation	Style Slug	Full Name
MC	multiple_choice	Multiple Choice
FTG	fill_the_gap	Fill the Gap
DQ	direct_question	Direct Question
RL	reverse_lookup	Reverse Lookup
FT	free_text	Free Text
PIR	progressive_image_reveal	Progressive Image Reveal
CV	conversational	Conversational

#Format Abbreviations

Abbreviation	Format Slug	Full Name
BFO	big_fan_of	Big Fan Of
KAL	know_a_lot_about	Know A Lot About
RAM	repeat_after_me	Repeat After Me
GWD	good_with_dates	Good With Dates
DOS	degrees_of_separation	Degrees of Separation
UTW	used_to_work_there	Used to Work There
PP	partial_pictures	Partial Pictures
OG	originators	Originators

---

#Overview

Eko's challenge architecture has two layers:

Layer	Purpose	Count	Examples
Styles	Presentation mechanics (how the challenge is rendered and answered)	7	multiple_choice, free_text, conversational
Formats	Experience wrappers (the themed personality around a challenge)	8	big_fan_of, originators, good_with_dates

A format selects from a subset of eligible styles and layers on its own voice, knowledge type, and tone. A single fact can be challenged in many style + format combinations across 5 difficulty levels.

Generation modes:

• Pre-generated (5 styles): Content is created ahead of time by AI, stored in fact_challenge_content, and served instantly.
• Runtime (2 styles): Content is generated live during the user session via multi-turn AI dialogue.

---

#Challenge Styles (7)

#multiple_choice

Property	Value
Generation	Pre-generated
Voice register	Option curator
Personality	Presents 4 curated options where each feels plausible while one clearly fits when you know the answer. Frames the selection as a commitment.

How it works: The user sees a setup narrative, a conversational question, and 4 options. Exactly 1 option is correct. Distractors are plausible and in the same category as the correct answer.

style_data schema:

{
  "options": [
    { "text": "Option A", "is_correct": false },
    { "text": "Option B", "is_correct": true },
    { "text": "Option C", "is_correct": false },
    { "text": "Option D", "is_correct": false }
  ]
}

• Exactly 4 options, exactly 1 with is_correct: true
• No "N/A", "None of the above", "All of the above", or "Unknown"
• Correct answer position varies

Example phrasings:

• "Take a look at these options -- which one do you think it is?"
• "One of these is the real answer. Can you spot it?"
• "Here's the lineup -- do you know which one fits?"
• "Four possibilities, one truth. What do you think?"

Reveal tone:

• Correct: Validates reasoning, acknowledges why other options were tempting
• Wrong: Shows why the correct option fits better, connects the distinction to a memorable detail

Emphasis: Lean into the fun of weighing options and committing to a choice. Pull back from clinical "which of the following" quiz-show phrasing.

Used by formats: Big Fan Of (default), Know A Lot About, Repeat After Me, Good With Dates, Used to Work There (default), Partial Pictures, Originators

---

#fill_the_gap

Property	Value
Generation	Pre-generated
Voice register	Puzzle architect
Personality	Presents an incomplete statement and invites the reader to supply the missing piece. Positions the blank at the most meaningful word.

Note: This style absorbed the former statement_blank style (migration 0160). SB was mechanically identical ("fill a blank" pattern) and its best voice prefixes were merged into FTG's pool for variety.

How it works: The user sees a statement with a ___ gap. They type the missing word or phrase. The gap should feel like a puzzle piece that clicks satisfyingly into place.

style_data schema:

{
  "complete_text": "The Eiffel Tower was built in 1889",
  "answer": "1889",
  "acceptable_answers": ["1889", "eighteen eighty-nine"]
}

• complete_text (required): Full sentence with the answer filled in
• answer (required): The blanked word/phrase
• acceptable_answers (optional): Alternative phrasings accepted as correct

Example phrasings:

• "There's a piece missing here. Can you fill it in?"
• "This story has a gap -- do you know what completes it?"
• "One detail is missing. Can you finish what's started?"
• "The sentence needs you -- can you complete it?"

Reveal tone:

• Correct: Celebrates the perfect fit, shows how the missing piece completes the whole picture
• Wrong: Shows how the piece fits with narrative context

Emphasis: Lean into the satisfaction of completing something. Pull back from sterile "fill in the blank" worksheet energy.

Used by formats: Repeat After Me (default)

---

#direct_question

Property	Value
Generation	Pre-generated
Voice register	Focused questioner
Personality	Poses a single, focused question and creates space for the reader to recall or reason their way to the answer. No safety net -- no options to choose from.

How it works: The user reads the setup, then answers a single direct question with no options or hints. The setup must do heavy lifting since there are no cues.

style_data schema:

{
  "expected_answer": "The Mariana Trench",
  "acceptable_answers": ["Mariana Trench", "Challenger Deep"],
  "answer_type": "name"
}

• expected_answer (required): Concise answer string
• acceptable_answers (optional): Alternative phrasings
• answer_type (optional): Expected format ("number", "name", "date", "phrase")

Example phrasings:

• "But here is the question -- do you know the answer?"
• "So here is what I want to know -- can you tell me?"
• "Okay, this is the real test -- do you know?"
• "Here is what stumps most people -- what's your answer?"

Reveal tone:

• Correct: Rewards recall with a detail that goes beyond the answer
• Wrong: Delivers the answer as a discovery with a memorable detail

Emphasis: Lean into the satisfaction of knowing something from memory. Pull back from flat "what is X?" textbook interrogation.

Used by formats: Good With Dates (default)

---

#reverse_lookup

Property	Value
Generation	Pre-generated
Voice register	Clue arranger
Personality	Lays out a trail of clues and asks the reader to deduce the identity of something. The setup IS the challenge -- each sentence is a progressively narrower clue.

How it works: The setup provides a series of clues that get increasingly specific. The user must deduce the identity of the described thing (a person, place, event, or concept) from those clues alone.

style_data schema:

{
  "answer": "The Eiffel Tower",
  "key_identifiers": ["iron structure", "Paris", "1889", "temporary"],
  "category": "place"
}

• answer (required): The thing being described
• key_identifiers (optional): Most distinctive clues from the description
• category (optional): Type of thing ("person", "place", "event", "concept")

Note: This style always needs a dedicated image (never inherits the fact's entity image) to avoid spoiling the answer.

Example phrasings:

• "From the clues above, can you figure out what is being described?"
• "All the clues point to one answer. Do you know what it is?"
• "The mystery has been laid out -- can you solve it?"
• "Based on these details, what do you think this is?"

Reveal tone:

• Correct: Celebrates the deduction, traces back through the clues
• Wrong: Unmasks the answer dramatically, highlights the key identifier

Emphasis: Lean into the thrill of deduction and the "aha" moment. Pull back from flat "what is being described?" without mystery energy.

Used by formats: Used to Work There, Originators (default)

---

#free_text

Property	Value
Generation	Pre-generated
Voice register	Reflective questioner
Personality	Poses open-ended questions that push beyond recall into reasoning, analysis, or opinion. No single correct answer -- invites multiple valid perspectives.

How it works: The user reads the setup and responds in their own words. Scoring evaluates key concepts covered, reasoning quality, and depth. The correct_answer is an expanded 4-8 sentence narrative.

style_data schema:

{
  "key_concepts": ["cultural significance", "engineering innovation", "tourism impact"],
  "sample_answer": "The Eiffel Tower represents...",
  "scoring_guidance": "Full credit for mentioning engineering + cultural impact. Partial for surface-level observations."
}

• key_concepts (required): Non-empty array of concepts a strong answer should touch on
• sample_answer (required): Well-written example response
• scoring_guidance (optional): Notes on what merits partial vs full credit

Example phrasings:

• "In your own words -- how would you explain this?"
• "What do you think is the most important takeaway here?"
• "How would you describe this to someone who has never heard of it?"
• "Here's a question for you -- why do you think this matters?"

Reveal tone:

• Correct: Acknowledges their insight, highlights what made their perspective strong, adds a complementary angle
• Wrong: Gently redirects by sharing the key insight as a discovery

Emphasis: Lean into the value of personal perspective and original expression. Pull back from recall-style "name the X" questions disguised as open-ended.

Used by formats: Big Fan Of, Know A Lot About (default), Good With Dates, Degrees of Separation, Used to Work There, Originators

---

#progressive_image_reveal

Property	Value
Generation	Runtime
Voice register	Visual puzzle-master (via partial_pictures format)
Personality	Reveals images progressively, from heavily obscured to fully visible, challenging users to identify the subject as early as possible.

How it works: An image is progressively unblurred or revealed in stages. The user guesses the subject at each stage. Earlier correct guesses earn more points. This style is generated at runtime (CC-003) and does not have pre-generated content in fact_challenge_content.

style_data schema: N/A (runtime-generated)

Used by formats: Partial Pictures (default)

---

#conversational

Property	Value
Generation	Runtime
Voice register	Varies by format
Personality	Multi-turn AI dialogue where the challenge unfolds through back-and-forth conversation. Each turn scores the user's response and presents the next question.

How it works: The user enters a multi-turn conversation with an AI host. Each turn is scored 0.0-1.0. The AI adapts questions based on previous answers, progressively increasing difficulty. Sessions have a defined number of turns (typically 3-5). Generated at runtime (CC-003) via processConversationalTurn().

style_data schema: N/A (runtime-generated)

Supported format-specific prompts exist for: big_fan_of, know_a_lot_about, degrees_of_separation. All other formats fall back to a generic quiz-host prompt.

Used by formats: Big Fan Of, Know A Lot About, Degrees of Separation (default)

---

#Challenge Formats (8)

#big_fan_of -- Big Fan Of

Property	Value
Display name	Big Fan Of
Tagline	Show your fandom
Knowledge type	fandom
Tone	Light
Conversational	Yes
Long-form	No
Icon	heart
Color	#ef4444 (red)
Pairs with	know_a_lot_about

Description: Prove how well you know the things you love -- sports teams, bands, shows, and more.

Supported styles: multiple_choice (default), free_text, conversational

Voice personality: Enthusiastic superfan. Energy of two fans meeting at a convention. Celebratory and insider-coded -- assumes the person loves the topic as much as you do.

Energy: High-warmth, celebratory.

Topic categories: sports, entertainment, people, places, food-beverage, cooking, animals, travel, tv, fashion, games, music, movies

---

#know_a_lot_about -- Know A Lot About

Property	Value
Display name	Know A Lot About
Tagline	Deep knowledge flex
Knowledge type	skill
Tone	Standard
Conversational	Yes
Long-form	No
Icon	brain
Color	#8b5cf6 (purple)
Pairs with	big_fan_of

Description: Test your expertise on topics you claim to know inside and out -- science, history, tech, and beyond.

Supported styles: free_text (default), multiple_choice, conversational

Voice personality: Meticulous expert. Confident, authoritative -- energy of a specialist testing a colleague. Respects the person as someone who takes the domain seriously.

Energy: Confident, authoritative.

Topic categories: science, history, geography, technology, how-things-work, math, spelling-grammar, geology, architecture, business, finance

---

#repeat_after_me -- Repeat After Me

Property	Value
Display name	Repeat After Me
Tagline	Complete the phrase
Knowledge type	memory
Tone	Light
Conversational	No
Long-form	No
Icon	quote
Color	#f59e0b (amber)

Description: Fill in famous quotes, song lyrics, and well-known phrases from memory.

Supported styles: fill_the_gap (default), multiple_choice

Voice personality: Nostalgic completionist. Playful, warm -- the energy of finishing someone else's sentence because you both know it.

Energy: Playful, warm.

Topic categories: quotes, music, movies, history, culture, spelling-grammar, publishing

---

#good_with_dates -- Good With Dates

Property	Value
Display name	Good With Dates
Tagline	When did it happen?
Knowledge type	temporal
Tone	Standard
Conversational	No
Long-form	No
Icon	calendar
Color	#3b82f6 (blue)

Description: Pin events, discoveries, and milestones to the right date or era.

Supported styles: direct_question (default), multiple_choice, free_text

Voice personality: Timeline detective. Precise yet wonder-filled -- dates are not just numbers, they are anchors in a larger story.

Energy: Precise, investigative.

Topic categories: history, current-events, sports, science, culture, events

---

#degrees_of_separation -- Degrees of Separation

Property	Value
Display name	Degrees of Separation
Tagline	Connect the dots
Knowledge type	association
Tone	Standard
Conversational	Yes
Long-form	Yes
Icon	git-branch
Color	#10b981 (emerald)

Description: Trace surprising links between people, places, and events through conversational back-and-forth.

Supported styles: conversational (default), free_text

Voice personality: Connection-chaser. Curious, conspiratorial -- the energy of "wait until you hear how these are connected." Sees hidden links between seemingly unrelated things.

Energy: Curious, conspiratorial.

Topic categories: people, entertainment, culture, history, music, movies, sports

---

#used_to_work_there -- Used to Work There

Property	Value
Display name	Used to Work There
Tagline	Match the career
Knowledge type	career
Tone	Standard
Conversational	No
Long-form	No
Icon	briefcase
Color	#6366f1 (indigo)

Description: Link notable people to the companies, teams, or institutions where they made their mark.

Supported styles: multiple_choice (default), reverse_lookup, free_text

Voice personality: Insider narrator. Revelatory, knowing -- the energy of sharing a "not many people know this" career detail.

Energy: Revelatory, knowing.

Topic categories: business, people, entertainment, technology, sports

---

#partial_pictures -- Partial Pictures

Property	Value
Display name	Partial Pictures
Tagline	What are you looking at?
Knowledge type	visual
Tone	Light
Conversational	No
Long-form	No
Icon	image
Color	#ec4899 (pink)

Description: Identify places, animals, and landmarks from progressively revealed images.

Supported styles: progressive_image_reveal (default), multiple_choice

Voice personality: Visual puzzle-master. Teasing, deductive -- the energy of a slow reveal where each clue narrows the field.

Energy: Teasing, deductive.

Topic categories: geography, animals, art, architecture, places, nature, countries, space

---

#originators -- Originators

Property	Value
Display name	Originators
Tagline	Who made it first?
Knowledge type	creator
Tone	Rigorous
Conversational	No
Long-form	No
Icon	lightbulb
Color	#f97316 (orange)

Description: Match inventions, artworks, theories, and cultural milestones to their original creators.

Supported styles: reverse_lookup (default), multiple_choice, free_text

Voice personality: Origin historian. Reverential, precise -- the energy of "let me tell you who did it first, and why it mattered."

Energy: Reverential, precise.

Topic categories: science, art, technology, culture, music, history, publishing, design

---

#Difficulty Levels (5)

Level	Label	Description	Guidance
1	Easy	Direct recall of the most prominent fact	Target the most well-known aspect. Generous context in setup. Correct answer clearly distinguishable.
2	Medium	Recall with some contextual reasoning	Target a secondary detail requiring careful reading. Plausible but distinguishable distractors.
3	Hard	Requires connecting multiple pieces of information	Synthesize two or more pieces of information. All distractors highly plausible.
4	Expert	Deep knowledge or cross-domain reasoning required	Target obscure details. Shorter setup assumes strong background. Specialized knowledge needed.
5	Master	Expert synthesis across multiple facts or domains	Challenges even subject matter experts. Near-miss distractors. Requires evaluation or prediction.

---

#Style-to-Format Matrix

Which styles are available in which formats. Bold = default style for that format.

Format	multiple_choice	fill_the_gap	direct_question	reverse_lookup	free_text	progressive_image_reveal	conversational
Big Fan Of	default				yes		yes
Know A Lot About	yes				default		yes
Repeat After Me	yes	default
Good With Dates	yes		default		yes
Degrees of Separation					yes		default
Used to Work There	default			yes	yes
Partial Pictures	yes					default
Originators	yes			default	yes

---

#Source Files

File	What it defines
packages/shared/src/schemas.ts	ChallengeStyleSchema (7 styles), ChallengeFormatSlugSchema (8 formats), KnowledgeTypeSchema
packages/ai/src/challenge-content-rules.ts	ChallengeStyle type (5 pre-gen), ChallengeFormatSlug type, validation rules
packages/ai/src/config/challenge-style-rules.ts	Per-style setup/challenge/reveal/correct_answer guidance + style_data schemas
packages/ai/src/config/challenge-style-voices.ts	Per-style voice personality, register, example phrasings
packages/ai/src/config/challenge-format-voices.ts	Per-format voice personality, register, energy
packages/ai/src/config/challenge-format-rules.ts	Per-format refinements for all text fields
packages/ai/src/config/challenge-difficulty.ts	5 difficulty level definitions
packages/ai/src/conversational-challenge.ts	Runtime conversational style logic and format-specific prompts
supabase/migrations/0105_challenge_formats.sql	DB enums for challenge_style and challenge_format_slug
supabase/migrations/0108_seed_challenge_formats.sql	Format seed data (display names, taglines, style mappings, topic mappings)