Welcome
This arcade runs on your own machine and uses your OpenAI API key directly from your browser to generate and judge images. Your key is saved locally on this computer so you won't be asked again.
Get a key at platform.openai.com/api-keys. It's stored only in a file on this computer, never sent anywhere except OpenAI.
DaRI AI Arcade
Sign in with your Google account to access solo arcade challenges, live tournaments, and player leaderboards.
Secured via Google OAuth 2.0. Any Google or Gmail account can sign in instantly.
You can set your Google Client ID in google-auth-config.json, or paste it here:
Welcome to DaRI AI Arcade!
You're signing in with Google for the first time. Confirm your role to complete setup:
Sign in with Google
Enter your Google or institutional account to continue to DaRI AI Arcade
DaRI AI
Welcome back, — pick where you want to go.
Welcome to DaRI AI — Your Gateway to Intelligent Arcade Gaming & Vertical AI.
Gaming Arcade
Step into the arcade! Challenge yourself with Picture Mirror, Emoji Charades, Code Mirror & Prompt Repair.
Agent Marketplace
Discover, test, and deploy specialized vertical AI agents crafted for healthcare, legal, finance, and engineering.
AI Learning
Interactive masterclasses, prompt engineering bootcamps, and personalized AI recruitment interview preparation.
Picture Mirror Setup
Customize your gameplay parameters before starting. The AI will generate a hidden original image tailored to your selected theme, difficulty, and rendering speed.
Level 1: One simple subject with clear shapes and plain lighting. Easiest to match closely.
DaRI AI — Agent Marketplace
A curated repository of enterprise-grade, vertical-specific autonomous agents designed for seamless production integration.
Clinical Health Agent
Extracts EHR insights, analyzes medical transcripts, and draft structured clinical notes adhering to HIPAA standards.
Legal Contract Auditor
Scans complex master services agreements, flags indemnification liabilities, and compares redlines in seconds.
Quantitative FinOps Agent
Connects to portfolio data feeds, conducts anomaly detection on transaction ledgers, and runs risk simulations.
DevSecOps Synthesizer
Monitors deployment pipelines, triages vulnerability alerts, and automatically generates patch pull requests.
Get Early Access
Join over 1,500 forward-thinking teams testing our vertical agents.
Thanks! You're on the priority notification list.
DaRI AI — AI Learning & Fluency
Empower your entire workforce with certified AI fluency, prompt engineering mastery, and role-based simulation labs.
Prompt Engineering Lab
Hands-on sandbox with real-time feedback, token optimization heuristics, and multi-model comparison matrix.
AI Recruitment Interview Prep
Live adaptive voice & text mock interviews with an AI evaluator providing instant rubric-based performance scoring.
Autonomous Agent Architecture
Deep-dive curriculum covering memory patterns, tool calling, re-act frameworks, and human-in-the-loop design.
Enterprise Fluency Certification
Verifiable credentials for business leaders, analysts, and developers demonstrating applied AI proficiency.
Join the Curriculum Waitlist
Receive introductory syllabus modules and early webinar access.
You're on the waitlist! We'll send early access syllabus materials soon.
AI Game Arcade
A small collection of games you play against AI models. Pick one to start.
Prompt Mirror
Study a hidden AI-generated picture, then write a prompt to recreate it. An AI judge scores how close you got.
Emoji Charades
AI invents a fresh movie, idiom, or phrase and translates it into emojis — the category is always shown up front. Scoring is instant, no images involved.
Code Mirror
Study a hidden AI-written Python function, then write a prompt to recreate it. An AI code reviewer scores how close you got. Nothing is ever executed — it's judged by reading, not running.
Prompt Repair Shop
Fix a customer's badly written AI prompt so it actually meets their requirement. Test it against one real message — solve it in fewer attempts for a higher score.
More games are on the way.
Prompt Mirror
How to play
- We'll generate a picture with AI — but we keep the prompt that made it a secret.
- Study the picture, then write your own prompt describing what you see.
- We generate a new picture from your prompt.
- An AI judge compares your picture to the original and scores how close you got, out of 100.
Each round makes one call to write the hidden prompt, two image-generation calls, and one scoring call to OpenAI, which uses your API credits.
How it's judged
An AI art judge compares your recreated picture to the hidden original across several dimensions: the exact subject (and how many of them), the color palette, the composition and framing, the art style and medium (photo vs. painting vs. cartoon, etc.), the setting and background, and small distinguishing details like lighting, texture, or accessories. It's deliberately strict and uses the full 0–100 scale, so scores in the 90s are reserved for near-perfect matches — even one clearly wrong detail (a different color, a missing prop) costs several points. It never judges how good your picture looks on its own, only how closely it matches the original.
Only a Dari Admin can set up a tournament — ask an admin to create one, then come back here to join it.
Level 1: one simple object, plain style. Easiest to match closely.
Recreate this picture
How to play
Study the picture above closely, then describe it in the prompt box exactly as you see it — objects, colors, style, composition. Submitting generates a new picture from your words, and an AI judge scores how closely it matches the original, out of 100.
How it's judged
An AI art judge compares your recreated picture to the hidden original across several dimensions: the exact subject (and how many of them), the color palette, the composition and framing, the art style and medium (photo vs. painting vs. cartoon, etc.), the setting and background, and small distinguishing details like lighting, texture, or accessories. It's deliberately strict and uses the full 0–100 scale, so scores in the 90s are reserved for near-perfect matches — even one clearly wrong detail (a different color, a missing prop) costs several points. It never judges how good your picture looks on its own, only how closely it matches the original.
Results
Original
Your recreation
See the detailed prompt an AI would write after looking at the original picture — a model answer to help you improve your own guesses next time. Optional — makes one extra call to OpenAI.
Prompt Mirror tournaments
Create a new tournament, or join one that's already running on this WiFi and play your matches from your own device.
Only a Dari Admin can create a tournament — ask an admin, then join it below once it's set up.
No tournaments yet — create one to get started.
Or join by tournament ID directly:
Create tournament
Add every player, then generate the fixture. It's a random single-elimination knockout — if the player count isn't a power of two, some players randomly get a first-round bye. Each match plays through the difficulty levels you pick below, in order, and whoever's cumulative score is higher advances. Once created, every player joins and plays their own matches from their own device.
Type a player's name, or their player ID if they have one (useful when two players share a name).
Tap a level to include or leave out that round — at least one must stay on. Each match is best-of-however-many you pick, played in that order (e.g. Easy only = 1 round; Easy + Medium = 2 rounds).
Tournament fixture
Tournament ID: — share this (and each player's own player ID) so they can join and play their matches from their own device.
How to play
Both players see the same hidden picture each sub-round, on their own device. Type your prompt whenever you're ready — you don't have to wait for your opponent to go first. Once you've both submitted, each device generates and scores its own player's picture, so neither of you sits around waiting on the other's turn. Whoever scores higher that sub-round wins it — the match goes to whoever's total score is higher across every level picked, with a tiebreaker sub-round if it ends up tied.
How it's judged
An AI art judge compares your recreated picture to the hidden original across several dimensions: the exact subject (and how many of them), the color palette, the composition and framing, the art style and medium (photo vs. painting vs. cartoon, etc.), the setting and background, and small distinguishing details like lighting, texture, or accessories. It's deliberately strict and uses the full 0–100 scale, so scores in the 90s are reserved for near-perfect matches — even one clearly wrong detail (a different color, a missing prop) costs several points. It never judges how good your picture looks on its own, only how closely it matches the original.
Sub-round of
See the detailed prompt an AI would write after looking at the original picture — a model answer to help you both improve. Optional — makes one extra call to OpenAI.
Emoji Charades
How to play
- AI invents a fresh hidden answer — a movie title, an idiom, or a common phrase — and translates it into a string of emojis.
- The category (Movie, Idiom, or Phrase) is shown right alongside the emojis, so you always know what kind of answer to look for.
- You study the emojis and type your best guess. Scoring is instant and local — no extra AI call, no images, just a close look at how well your guess matches.
Each round makes one AI call to OpenAI (to invent the answer and translate it into emojis), which uses your API credits. A fresh answer is generated every round, so content doesn't repeat.
How it's judged
Scoring is done locally, without any AI call. Your guess is compared to the hidden answer two ways: how many of its key words you got right (ignoring small words like "a" or "the"), and how close your spelling and phrasing are letter-for-letter. The two are blended — weighted a bit more toward key words — into a score out of 100. An exact match is always 100; a guess that shares no words or letters with the answer scores near 0.
Only a Dari Admin can set up a tournament — ask an admin to create one, then come back here to join it.
Level 1: well-known, easy-to-guess movies, idioms, and phrases.
Guess the answer
How to play
Read the emoji string above — the category badge next to it tells you whether you're guessing a movie, an idiom, or a phrase. Type your best guess below — you can reveal a clue if you get stuck. Scoring is instant and compares your guess directly to the hidden answer, out of 100.
How it's judged
Scoring is done locally, without any AI call. Your guess is compared to the hidden answer two ways: how many of its key words you got right (ignoring small words like "a" or "the"), and how close your spelling and phrasing are letter-for-letter. The two are blended — weighted a bit more toward key words — into a score out of 100. An exact match is always 100; a guess that shares no words or letters with the answer scores near 0.
Results
See what an AI guesses from just the emoji string and category — the same information you had — to compare against your own guess. Optional — makes one extra call to OpenAI.
Tournament setup
Add every player, then generate the fixture. It's a random single-elimination knockout — if the player count isn't a power of two, some players randomly get a first-round bye. Each match plays through the difficulty levels you pick below, in order, and whoever's cumulative score is higher advances.
Type a player's name, or their player ID if they have one (useful when two players share a name).
Tap a level to include or leave out that round — at least one must stay on. Each match is best-of-however-many you pick, played in that order.
Tournament fixture
How to play
Both players see the same emoji string each sub-round and guess independently. Whoever's guess scores higher wins that sub-round — the match goes to whoever's total score is higher across every level you picked, with a tiebreaker sub-round if it ends up tied.
How it's judged
Scoring is done locally, without any AI call. Your guess is compared to the hidden answer two ways: how many of its key words you got right (ignoring small words like "a" or "the"), and how close your spelling and phrasing are letter-for-letter. The two are blended — weighted a bit more toward key words — into a score out of 100. An exact match is always 100; a guess that shares no words or letters with the answer scores near 0.
Sub-round of
See what an AI guesses from just the emoji string and category — the same information you both had. Optional — makes one extra call to OpenAI.
Code Mirror
How to play
- We'll ask AI to write a small Python function — but we keep the request that produced it a secret.
- Study the function, then write your own prompt describing what you think it does.
- We generate a new function from your prompt.
- An AI code reviewer compares your function to the original and scores how closely it matches, out of 100. Nothing is ever run — it's judged by reading the code, like a human reviewer would.
Each round makes two code-generation calls and one scoring call to OpenAI, which uses your API credits.
How it's judged
An AI code reviewer reads your recreated function side-by-side with the hidden original — nothing is ever executed. It judges whether the two would behave the same way on typical and edge-case inputs (empty input, negatives, duplicates, and so on), whether they take a similar algorithmic approach, whether the function's signature and overall structure match, how robust each is (error handling, input validation), and how similar the readability and style are. It's strict: scores above 90 are reserved for functionally equivalent code, while a behavioral mismatch or a completely different approach to the problem scores below 20.
Only a Dari Admin can set up a tournament — ask an admin to create one, then come back here to join it.
Level 1: one simple function. Easiest to match closely.
Recreate this function
How to play
Read the hidden Python function above and figure out what request would have produced it. Describe that request in the prompt box — it generates a new function from your words, and an AI reviewer scores how closely its behavior matches the original, out of 100. Nothing is ever run.
How it's judged
An AI code reviewer reads your recreated function side-by-side with the hidden original — nothing is ever executed. It judges whether the two would behave the same way on typical and edge-case inputs (empty input, negatives, duplicates, and so on), whether they take a similar algorithmic approach, whether the function's signature and overall structure match, how robust each is (error handling, input validation), and how similar the readability and style are. It's strict: scores above 90 are reserved for functionally equivalent code, while a behavioral mismatch or a completely different approach to the problem scores below 20.
Results
Original
Your recreation
See the detailed prompt an AI would write after reading the hidden function — a model answer to help you improve your own guesses next time. Optional — makes one extra call to OpenAI.
Tournament setup
Add every player, then generate the fixture. It's a random single-elimination knockout — if the player count isn't a power of two, some players randomly get a first-round bye. Each match plays through the difficulty levels you pick below, in order, and whoever's cumulative score is higher advances.
Type a player's name, or their player ID if they have one (useful when two players share a name).
Tap a level to include or leave out that round — at least one must stay on. Each match is best-of-however-many you pick, played in that order.
Tournament fixture
How to play
Both players see the same hidden function each sub-round and write their own prompt to recreate it. Whoever's function scores higher wins that sub-round — the match goes to whoever's total score is higher across every level you picked, with a tiebreaker sub-round if it ends up tied.
How it's judged
An AI code reviewer reads your recreated function side-by-side with the hidden original — nothing is ever executed. It judges whether the two would behave the same way on typical and edge-case inputs (empty input, negatives, duplicates, and so on), whether they take a similar algorithmic approach, whether the function's signature and overall structure match, how robust each is (error handling, input validation), and how similar the readability and style are. It's strict: scores above 90 are reserved for functionally equivalent code, while a behavioral mismatch or a completely different approach to the problem scores below 20.
Sub-round of
See the detailed prompt an AI would write after reading the hidden function — a model answer to help you both improve. Optional — makes one extra call to OpenAI.
Prompt Repair Shop
How to play
- A customer hands you a badly written AI prompt, a plain-English requirement it needs to meet, and the one real message it has to handle.
- Edit the prompt, then click "Test repair" to run it against that message and see whether the result meets the requirement.
- You get 3 attempts. Editing is free; only clicking "Test repair" uses one up — and a technical/API error never costs you an attempt.
- Solve it faster for a higher score: 100 points on attempt 1, 75 on attempt 2, 50 on attempt 3. Hints are optional and cost 10 points each.
How it's judged
Your edited prompt is run once against a real customer message, and the AI's response is checked against the ticket's requirement. Solving it on your first attempt scores 100, your second 75, your third 50 — not solving it within 3 attempts scores 0. Each hint you use subtracts 10 points from whatever you scored.
Only a Dari Admin can set up a tournament — ask an admin to create one, then come back here to join it.
Level 1: extract one or two simple fields from a short message.
Repair this prompt
How to play
Edit the broken prompt below so it satisfies the requirement listed above, then click "Test repair" to run it against the one real message shown. You get 3 attempts — editing is free, only testing uses one up, and a technical error never costs an attempt. Hints cost 10 points each. Solving it sooner scores higher.
How it's judged
Your edited prompt is run once against the customer message shown, and the AI's response is checked against the requirement. Solving it on attempt 1 scores 100, attempt 2 scores 75, attempt 3 scores 50 — not solving it in 3 attempts scores 0. Each hint you use subtracts 10 points from whatever you scored.
Latest attempt
Ticket closed
See the prompt an AI would write to satisfy every requirement — a model answer to help you improve your own repairs next time. Optional — makes one extra call to OpenAI.
Tournament setup
Add every player, then generate the fixture. It's a random single-elimination knockout — if the player count isn't a power of two, some players randomly get a first-round bye. Each match plays through the difficulty levels you pick below, in order; both players repair the SAME ticket with their own private 3 attempts, and whoever's cumulative score is higher advances.
Type a player's name, or their player ID if they have one (useful when two players share a name).
Tap a level to include or leave out that round — at least one must stay on. Each match is best-of-however-many you pick, played in that order.
Tournament fixture
How to play
Both players repair the same ticket with their own private 3 attempts and hints. Once both are done, whoever scored higher wins that sub-round — the match goes to whoever's total score is higher across every level you picked, with a tiebreaker sub-round if it ends up tied.
How it's judged
Each player's edited prompt is run once against the customer message shown, and the AI's response is checked against the requirement. Solving it on attempt 1 scores 100, attempt 2 scores 75, attempt 3 scores 50 — not solving it in 3 attempts scores 0. Each hint used subtracts 10 points from whatever that player scored.
Sub-round of
Latest attempt
Ticket:
See the prompt an AI would write to satisfy every requirement — a model answer to help you both improve. Optional — makes one extra call to OpenAI.
Player profiles
Stats are saved on this computer and tracked separately for solo play and tournaments, per game. Just typing your name works and will keep building the same profile — but if two players share a name, create a profile below and use the player ID it gives you instead.