← Back to Blog
WORLDS

The Ten Worlds of Jogg Mini

by MokingBird
Product and curriculum reference | MokingBird

Updated: 1 July 2026

Jogg Mini organises AI literacy into ten themed learning worlds. A world is not just a visual category: it owns a question pool, a learning mission, progress, stars and a recognisable place in the wider curriculum.

The world model lets a young learner move from basic recognition of AI to data, models, modern AI systems, responsible use and independent judgment without being presented with one undifferentiated list of more than 2,000 questions.

Why Jogg Mini teaches through worlds

Artificial intelligence is not one school subject. It combines computing, statistics, language, design, engineering, ethics, media literacy and social decision-making. A flat topic list may be accurate, but it gives a child no sense of place, direction or completion. The world system provides all three.

Each title performs curriculum work. Robot Valley lowers the barrier to entry. Data Valley turns training material into something that can be collected and cultivated. Pattern Mountain makes abstraction feel like a climb. Smart City shows connected architecture. Future Lab introduces experimentation and deployment. AI Safety & Ethics gives responsibility a permanent home. Prompt Plaza, Agent Academy and AI Studio turn three modern modes of AI use into distinct skills. AI Forest brings the parts back together.

This approach benefits learning in several ways:

visual environment and narrative.

architecture, deployment and governance.

not merely reducing a question count.

connected without being collapsed into one vague “generative AI” category.

recurring concern in data, prompting, agents, creation and deployment.

Worlds are not rigid academic silos. Concepts intentionally recur from different angles. Data quality introduced in Data Valley returns as model bias, retrieval quality, agent reliability and responsible creation. Repetition with new context strengthens understanding and shows that real AI systems cross technical boundaries.

Educational and technical architecture

The question specification maps child-facing worlds to a more technical eight-layer AI framework:

  1. foundations and concepts;
  2. data;
  3. embeddings and representations;
  4. model training and architecture;
  5. retrieval-augmented generation and agents;
  6. inference and optimisation;
  7. deployment;
  8. applications, including cross-cutting safety questions.

The mapping keeps the playful interface educationally accountable. A question may mention a robot, forest, city or mission, but its metadata still identifies the technical layer, difficulty, age range, tags and expected answer. The metaphor helps explain the concept; it does not replace the concept.

Difficulty is also more than a vocabulary change. Easy questions emphasise recognition, concrete choices and safe behaviour. Medium questions require comparison, explanation and simple application. Hard questions introduce technical terminology, multi-step reasoning, trade-offs and system-level analysis. The specification spans approximately ages 5–15, so the same world can support a young learner encountering AI for the first time and an older learner ready for embeddings, gradient descent, RAG, agent loops or AI governance.

The current question bank uses three core formats:

one correct action or factor.

Explanations are essential. The purpose is not only to mark an answer but to correct the learner’s mental model. A wrong answer about training data should explain why training and testing differ. A safety answer should explain the risk and the appropriate action. A hard technical question should connect the term to the larger system.

Why this matters for education

Children are already interacting with recommendation systems, generative assistants, synthetic media and automated decisions. Waiting until university to teach AI means allowing years of unexamined assumptions to form. At the same time, teaching only tool operation would prepare learners for products that may be obsolete by the time they leave school.

Jogg Mini therefore targets durable literacy. Learners should be able to:

For teachers, the worlds provide teachable curriculum boundaries and discussion prompts. For parents, they provide understandable categories for reviewing a child’s learning. For children, they create a journey in which technical competence and responsibility grow together.

Current content snapshot

supported by the source data

The source seed expected 2,241 questions, so the current 2,239 total retains a known two-question discrepancy. This does not prevent the ten worlds from loading.

How worlds appear in the app

Home reads active worlds from Supabase and is the database-driven reference. Play and Progress are configured for the same ten-world sequence:

  1. Robot Valley
  2. Data Valley
  3. Pattern Mountain
  4. Smart City
  5. Future Lab
  6. AI Safety & Ethics
  7. Prompt Plaza
  8. Agent Academy
  9. AI Studio
  10. AI Forest

When a learner opens a world, the first view shows its name, visual identity, completion percentage and stars. An information icon in the header opens a short explanation of the world and a list of the topics it covers. The learner can close the information dialog and continue to the session filters.

World sessions support:

This prevents a learner from being asked to complete an entire 200-question world in one sitting.

Progress and access

Progress is tracked independently for each world. Home, Play and Progress should always show the same worlds and the same underlying question totals.

The canonical client exposes all ten Worlds without a legacy progression threshold. Progress is represented by attempted/mastered questions, accuracy, Stars and Tier. Parent restrictions can still make a World unavailable for a specific child.

Each first-attempted question enters a per-kid/per-World state. Previously incorrect questions are prioritized in later sessions, then unanswered content, so a learner cannot hide a misconception by selecting only new questions. A World is complete only after every active question has been attempted and overall ever-correct accuracy is greater than 70%. Every block of 50 first attempts can grant one Gem at 70% block accuracy; a later corrective retry may earn a previously missed block milestone once.

Freemium allowance counts only normal World Lane first attempts: 50 in Robot Valley, Data Valley and AI Safety & Ethics; 25 in each other World. Daily Quest, Arcade, Random Quiz, teacher/joined quizzes and parent tasks use separate pools and do not consume it. Paid plans provide unlimited World questions without changing scoring.

The access policy should be treated separately from curriculum order. A world being open does not mean its concepts are easier; it means the learner can explore it without waiting for an artificial progression gate.

World 1: Robot Valley 🤖

Learning mission: Understand what AI is and recognise it in everyday life.

Robot Valley is the entry point. It replaces the idea that AI is magic with a simple working model: AI systems learn patterns from examples and use those patterns to produce predictions, classifications or decisions.

Why we chose the name Robot Valley

“Robot” is one of the first words children associate with artificial intelligence, even though a robot and an AI system are not the same thing. That misunderstanding makes it an effective starting point. The valley is imagined as a welcoming place filled with workshops, friendly machines, assembly lines and characters such as Robo or Beep. Children enter through something familiar and gradually discover that the intelligence may live in software, while the robot is only one possible body for it.

The word “valley” also matters. A valley is an accessible beginning before the curriculum climbs into more technical terrain. It communicates exploration rather than examination: the learner is looking around, noticing intelligent systems and asking why they behave differently from ordinary machines.

Curriculum design

The question specification gives this world the largest emphasis on foundations and real-world applications. Younger learners compare calculators, fixed instructions, game characters and adaptive systems. Older learners meet algorithms, models, parameters, tokens, benchmarks, narrow AI, reinforcement learning, computer vision and natural-language processing.

This layering avoids two common educational failures. The first is defining AI so loosely that every computer becomes “AI.” The second is introducing model architecture before the learner has any reason to care. Robot Valley instead builds a usable boundary: automation follows programmed rules; machine learning derives useful behaviour from examples; and neither one possesses human feelings or understanding merely because its output looks intelligent.

Topics covered

By the end of the world, learners should be able to identify likely uses of AI and explain why not every automated machine is intelligent.

Educational impact and learner benefit

Robot Valley gives children vocabulary for technology they already encounter. Recommendations, route planning, autocorrect, speech recognition, spam filtering, medical-image assistance and game opponents stop being unexplained magic. A learner who can ask “what data might this use?” or “is this learning, predicting or following a fixed rule?” has begun to think computationally.

That foundation also reduces exaggerated beliefs about AI. Children learn that systems can be useful and impressive while still making mistakes, lacking emotions and operating within limited tasks. This calibrated understanding is beneficial throughout the curriculum because later safety, prompting and agentic-AI lessons depend on knowing both what AI can do and what it cannot.

World 2: Data Valley 📊

Learning mission: Understand that data is the material from which AI learns.

Data Valley explains datasets, examples and labels. It also establishes that more data is not automatically better: data must be relevant, accurate, representative and handled responsibly.

Why we chose the name Data Valley

A valley collects water, soil, seeds and living systems. In the same way, a machine-learning project gathers observations before anything can be trained. The specification uses an earthy discovery metaphor: data seeds grow into knowledge only when they are sorted, labelled and cared for. This makes an otherwise abstract pipeline visible to a child.

“Data” is deliberately placed in the title rather than hidden behind a character name. Data is the central dependency of modern AI, and children need to see it as a first-class subject. The title tells learners that the impressive output seen in Robot Valley has a source—and that source can be incomplete, messy, private, biased or misleading.

Curriculum design

More than half of this world’s specified question coverage belongs directly to the data layer, with substantial coverage of model training. Early questions identify text, images, sound, measurements, labels and tables. Intermediate questions introduce annotation, features, structured and unstructured data, class balance, diversity, cleaning and training/test splits. Advanced questions cover validation sets, overfitting, underfitting, augmentation, synthetic data, ground truth, poisoning, pipelines, batches, epochs, loss, gradient descent, few-shot learning and zero-shot learning.

The progression intentionally moves from “what is a piece of data?” to “how does the design of a dataset change model behaviour?” It also separates a label from a prediction and training evidence from evaluation evidence. These distinctions prevent learners from imagining that a model already knows the answer it is being asked to produce.

Topics covered

Learners should leave with the principle that poor or unfair data can produce poor or unfair AI outcomes.

Educational impact and learner benefit

Data literacy is useful far beyond AI. Learners practise categorisation, evidence quality, sampling, measurement and the difference between an example and a conclusion. They discover why testing on training data can give a false sense of success and why a dataset representing only one group can fail for everyone else.

The world also creates an early bridge to privacy and fairness. When children understand that photos, messages, locations and voices can become training data, consent is no longer an abstract legal word. They can reason about what should be collected, who should have access and whether the collection serves the stated purpose. That makes Data Valley beneficial both as technical education and as practical digital citizenship.

World 3: Pattern Mountain 🎯

Learning mission: Explore how AI detects structure and represents similarity.

Pattern Mountain bridges raw data and learned models. It develops intuition for features, patterns, vectors and embeddings without requiring advanced mathematics.

Why we chose the name Pattern Mountain

Patterns are often easy to see from a distance but difficult to understand up close. A mountain therefore works as both a learning metaphor and a progression signal. Learners follow trails through data, group similar observations and climb toward representations that reveal structure. The summit represents the moment when scattered examples become a model capable of handling something new.

The mountain title also prepares children for productive difficulty. Vectors, embeddings, dimensionality and similarity are more abstract than recognising an AI-powered product. Framing this as an expedition turns increasing complexity into visible progress. Easy and medium questions can use maps, connect-the-dots, clusters and trails; hard questions can then use precise technical language without changing the underlying mental model.

Curriculum design

The specification centres this world on embeddings and representations, then connects those representations to model training and data. Learners begin with features and classification before exploring how images become arrays of numbers and how words become tokens and vectors. They encounter clustering, centroids, decision trees, regression, similarity search and recommendation systems.

For older learners, the route continues into cosine similarity, feature vectors, dimensionality reduction, PCA, n-grams, sequence models, attention, transformers and context windows. These are not presented as disconnected buzzwords. Each one answers a coherent question: how can a machine represent meaning or similarity in a form it can calculate with?

Topics covered

Learners should understand that an AI model does not store human-like meanings; it builds numerical representations useful for its task.

Educational impact and learner benefit

Pattern Mountain develops abstraction. A learner moves from recognising a visible pattern to understanding that useful relationships can exist in a high-dimensional numerical space. This is a major conceptual step toward real machine-learning literacy.

It also demystifies familiar services. Autocomplete, product recommendations, music suggestions, image search and semantic search become examples of representation and similarity rather than evidence that a machine “knows” a person. The educational benefit is a more accurate language for discussing model behaviour: similar is not identical, correlated is not causal, and a representation useful for one task may be unsuitable or unfair for another.

World 4: Smart City 🏙️

Learning mission: Understand neural networks, training and inference.

Smart City represents a connected system whose components pass information through layers. It focuses on the machinery behind modern machine learning, while retaining concrete examples children can visualise.

Why we chose the name Smart City

A city is a network of specialised parts. Roads move traffic, power lines carry energy, communication systems transmit messages and different buildings perform different jobs. This is a useful analogy for a neural network, where layers transform information and connected units contribute to an overall result.

The word “smart” is intentionally examined rather than accepted at face value. A smart city is not intelligent in a human sense. It is a collection of sensors, models, software and decisions. The title gives learners a visual system they can explore while teaching them to ask what makes a product “smart,” what data flows through it and where human choices remain.

Curriculum design

Half of the specified coverage concerns model architecture and training. Learners begin with inputs, artificial neurons, weights, hidden layers and outputs. They then study activation functions, forward passes, loss, gradients, learning rates and backpropagation. Older learners encounter ReLU, sigmoid, softmax, CNNs, RNNs, encoder-decoder systems, transfer learning, fine-tuning, checkpoints, convergence, dropout, normalisation and model capacity.

The curriculum repeatedly contrasts parameters with hyperparameters, architecture with learned weights, and training with inference. It also introduces practical constraints: GPUs, model size, compression, quantisation, latency and benchmark performance. The aim is not to make children implement a neural network from equations, but to give them an accurate systems-level map of what happens when one learns.

Topics covered

Learners should be able to distinguish building or training a model from using an already trained model.

Educational impact and learner benefit

Smart City replaces the “electronic brain” metaphor with something more precise. Learners understand that a network is a mathematical architecture, that weights change during training and that a forward pass produces an output. That knowledge makes later discussions of fine-tuning, model updates and inference costs substantially easier.

The city model also encourages systems thinking. A failure may originate in data, architecture, optimisation, deployment or the way output is interpreted. Children learn not to attribute every problem to one mysterious AI component. This is beneficial in technical education because it develops causal reasoning, and in citizenship because it shows that people design, train, deploy and govern every supposedly autonomous system.

World 5: Future Lab 🚀

Learning mission: Discover how modern AI products are assembled and delivered.

Future Lab moves from individual models to complete systems. It shows that a useful AI product also needs data pipelines, retrieval, APIs, deployment, monitoring and human decisions.

Why we chose the name Future Lab

A laboratory is where ideas become testable systems. “Future” signals that the world contains technologies learners are seeing emerge now: foundation models, multimodal systems, retrieval, agents and cloud AI. Together, the words create a place for experimentation without implying that every new technology is automatically an improvement.

The lab framing is important educationally. Advanced AI should be tested, measured, monitored and revised. It should not be launched merely because a demo looks impressive. The title therefore supports both technical curiosity and scientific caution.

Curriculum design

Future Lab has the widest engineering scope. The specification gives major weight to RAG and agents, deployment, applications, inference optimisation and embeddings. Questions explain servers, APIs, REST interfaces, model endpoints, cloud computing, latency, throughput and scalability. Retrieval topics include knowledge bases, chunks, vector databases, semantic search and why grounding a response in selected documents can reduce—but not eliminate—hallucinations.

The world then connects models to product operations: versioning, monitoring, A/B testing, feedback loops, drift, streaming, token limits and orchestration. It introduces large language models, foundation models and multimodal AI while distinguishing a chatbot from an agent that has tools, memory and a sequence of actions.

Advanced questions use authentic vocabulary because older children benefit from meeting the same terms used in engineering documentation. The surrounding examples remain concrete enough that terminology is tied to purpose rather than memorised in isolation.

Topics covered

This world prepares learners to see AI as one component in a larger engineered system.

Educational impact and learner benefit

Future Lab teaches that deploying AI is a lifecycle, not a single model call. A learner can trace an application from user request through an API, retrieval system and model endpoint, then consider latency, cost, monitoring and feedback. This is foundational knowledge for future software, data and product work.

The world also develops healthy scepticism about novelty. RAG can improve grounding but retrieve the wrong source. A larger model can be more capable but slower and more expensive. An agent can automate work but also compound errors. The benefit is balanced technical literacy: learners can be excited by modern AI while still asking how it is evaluated, maintained and controlled.

World 6: AI Safety & Ethics 🛡️

Learning mission: Learn to use, evaluate and build AI responsibly.

Safety and ethics are available from the beginning because responsible use is not an advanced optional topic. Learners do not need to understand neural networks before they can reason about privacy, fairness or harmful content.

Why we chose the name AI Safety & Ethics

The generation specification uses a superhero headquarters concept—sometimes called Hero HQ—because guardianship is active. A responsible learner does more than know a safety rule; they notice risks, protect private information, question unfair outcomes and seek help when a system behaves dangerously. The shield icon preserves that guardian identity.

The product-facing title uses “AI Safety & Ethics” because the subject must be unambiguous to parents, educators and older learners. “Safety” concerns reliability, misuse and harm. “Ethics” asks what is fair, who is accountable and which uses should be permitted. Keeping both words prevents the world from being reduced either to cybersecurity tips or to abstract philosophy.

Curriculum design

The world is deliberately cross-cutting. Its questions connect safety to applications, data, foundations, inference and deployment. Younger learners practise concrete rules: protect personal details, do not follow scary advice, check with a trusted adult and do not assume every generated answer is true. Intermediate learners examine bias, consent, opt-outs, deepfakes, misinformation, surveillance and responsible school use.

Advanced learners encounter transparency, explainability, audits, accountability, data minimisation, copyright, facial-recognition ethics, algorithmic hiring, model cards, responsible disclosure, alignment, watermarking, regulation and environmental cost. Case studies in healthcare, justice, education, work and government make clear that technical performance is only one part of responsible deployment.

Topics covered

Learners should develop the habit of asking who benefits, who may be harmed, what data was used and which human remains responsible.

Educational impact and learner benefit

This world equips children with practical protective behaviours before they become habitual AI users. It teaches them not to share passwords, addresses, private images or sensitive family information with a chatbot. It also gives them language for asking why a result may be biased and how to fact-check a claim.

At a deeper level, AI Safety & Ethics teaches moral and civic reasoning. Learners compare benefits against harms, individual convenience against group fairness, and automation against accountability. The benefit is not simply safer app usage. It is the development of young people who can participate in decisions about technology rather than accepting those decisions as inevitable.

World 7: Prompt Plaza 💬

Learning mission: Communicate clearly and safely with generative AI.

Prompt Plaza treats prompting as the skill of giving useful directions. The plaza metaphor presents prompts as requests exchanged in a busy creative market: clarity, context and examples improve the result.

Why we chose the name Prompt Plaza

A plaza is a place of conversation, exchange and many possible requests. That makes it a natural setting for learning how people communicate with generative AI. The specification compares the learner to a director giving instructions to an AI actor: the clearer the direction, the more useful the performance.

The lively market identity keeps prompting from becoming a narrow lesson about typing special commands. Learners “shop” for better words, add context, compare different outputs and revise their requests. The title also reinforces that a prompt is part of an interaction. A useful result often comes from a sequence of review and refinement rather than one perfect sentence.

Curriculum design

The world begins with purpose, audience and specificity. Children compare “tell me about space” with prompts that specify age, format, scope and desired detail. They learn to add context, constraints, examples, roles and output formats, then evaluate whether the result actually follows the request.

Intermediate questions cover follow-up prompts, iterative prompting, few-shot examples, ambiguity, conflicting instructions and prompt templates. Older learners explore system instructions, context windows, token limits, prompt injection, jailbreak attempts and the difference between improving an answer’s form and proving its truth.

The educational pattern is practical: write, inspect, question and revise. Prompting is treated as communication plus evaluation, not as a collection of magic phrases.

Topics covered

Learners should understand that better prompts improve results but cannot guarantee truth or safety.

Educational impact and learner benefit

Prompt Plaza improves transferable communication skills. To write a strong prompt, a learner must define a goal, identify relevant context, choose constraints and describe what a successful answer would look like. Those are the same skills used in research questions, design briefs, programming tasks and clear requests to other people.

The world also protects against “prompt confidence”: the mistaken belief that a well-worded instruction makes the output correct. Learners practise checking sources, identifying missing information and asking the model to expose uncertainty. This makes prompting beneficial as both a productivity skill and a critical-literacy skill.

World 8: Agent Academy 🦾

Learning mission: Understand AI systems that plan, use tools and act toward goals.

Agent Academy is structured like a mission school. An agent receives a goal, chooses steps, uses available tools, observes results and adjusts its plan.

Why we chose the name Agent Academy

An academy is where capability is trained under rules. The specification uses a mission and special-operations atmosphere with characters such as Agent Ace, Commander Code and Tool-Belt Bot. This turns agent architecture into a concrete story: a mission has a goal, a plan, permitted tools, observations and a report of what happened.

The title deliberately avoids calling every assistant an agent. An academy teaches the distinction. A chatbot may respond with text; an agent may search, run code, call an API, update a file or attempt a purchase. Greater agency creates greater usefulness and greater risk, so permissions and supervision are part of the identity from the beginning.

Curriculum design

Early questions identify goals, steps and tools. Learners decide which tool is appropriate for a search, calculation, calendar action or coding task. Intermediate questions introduce memory, context, observations, replanning, tool selection, logs and the loop of planning, acting and checking.

Advanced questions cover function calling, autonomous behaviour, agent workflows, short- and long-term memory, context limits, multi-agent systems, error recovery and compounding failures. Research and coding agents provide real examples, while scenarios involving messages, purchases or personal data introduce approval boundaries.

Human oversight is not placed in a separate ethics footnote. It is built into the agent lifecycle: before a consequential action, the system should confirm authority, scope and expected impact.

Topics covered

Learners should recognise that an agent can do more than answer a question and therefore needs stronger safeguards than a simple chatbot.

Educational impact and learner benefit

Agent Academy develops procedural thinking. Learners break a goal into steps, choose tools, inspect intermediate results and revise a failed plan. These are valuable skills in programming, project work and independent problem-solving even when no AI system is involved.

It also prepares children for a major shift in AI products. The safety question is no longer only “is this answer correct?” but also “what is this system allowed to do?” A learner who understands permissions, logs, checkpoints and human approval is better equipped to use future autonomous tools without granting them unnecessary access or trusting a chain of invisible actions.

World 9: AI Studio 🎨

Learning mission: Create with AI while keeping human judgment and responsibility.

AI Studio is a makerspace covering practical generative tools. The learner is the creator and AI is an assistant—not the unquestioned author or source of truth.

Why we chose the name AI Studio

A studio is a place where ideas become artefacts through experimentation, selection and revision. The specification gives this world a colourful makerspace identity, with characters such as Aria the AI Artist and Code Canvas Bot. That framing places the child—not the model—in the role of creator.

The title is broader than “AI Art” because generative systems work across text, images, music, speech, video and code. A studio can contain all of those media. It also implies process: drafts are evaluated, edited, combined and sometimes discarded. Generation is one step in creation, not the complete creative act.

Curriculum design

Learners explore text generation, text-to-image systems, music generation, speech synthesis, translation, summarisation, code assistants and AI tutoring. They compare modalities and learn that different models are trained for different inputs and outputs.

Intermediate questions address prompting for creative work, style guidance, editing, hallucinations, translation quality and responsible homework support. Advanced topics include diffusion models, multimodal generation, voice cloning, inpainting, outpainting, style transfer, procedural generation and text-to-video systems.

Creation is always paired with responsibility. The world discusses plagiarism, cheating, attribution, copyright, consent, deepfakes, disclosure and the need to verify factual content. It asks not only “can AI make this?” but “should it, who owns it, and how should its use be communicated?”

Topics covered

Learners should be able to choose where AI assistance is useful and identify where verification, permission or disclosure is needed.

Educational impact and learner benefit

AI Studio encourages active making rather than passive consumption. Learners can use generated material as a starting point, then apply taste, factual judgment and personal intent. This supports creativity without teaching that speed is the same as quality.

The world is also directly relevant to education. Children need a framework for distinguishing tutoring from answer substitution, brainstorming from plagiarism, and editing support from hidden authorship. By connecting practical tools to attribution and verification, AI Studio helps learners, parents and teachers discuss acceptable use in concrete terms.

World 10: AI Forest 🌲

Learning mission: Combine the entire curriculum into practical AI literacy.

The generation specification originally describes this capstone as Wisdom Woods. The current database and app-facing product name is AI Forest. The mission remains the same: connect earlier concepts and make thoughtful decisions in unfamiliar AI situations.

Why we chose the forest concept

A mature forest is an interconnected system. Roots, soil, water, plants and animals affect one another, just as data, representations, models, deployment, people and policy interact in an AI system. No single tree explains the whole forest. This makes the metaphor especially appropriate for a capstone world whose purpose is synthesis.

The original name, Wisdom Woods, emphasised judgment earned across all earlier worlds. The current name, AI Forest, is clearer in navigation while preserving that idea. The learner enters with knowledge collected elsewhere and must now decide how the pieces fit together. Wisdom here does not mean knowing every technical term; it means asking better questions when the situation is new.

Curriculum design

The specification intentionally draws from every adult layer. Questions revisit data, embeddings, training, RAG, agents, inference, deployment, applications and safety, but now ask learners to connect them. A model-lifecycle question, for example, may require reasoning from collection through training, evaluation, deployment, monitoring and retirement.

The world also covers modern AI literacy: large language models, foundation models, multimodal AI, reasoning systems, open and closed models, AGI, alignment, RLHF and AI safety research. Societal questions address education, employment, government, regulation, governance, warfare and human-AI collaboration with age-appropriate framing.

Critical-thinking tasks ask learners to evaluate AI-generated articles, synthetic media, product claims and AI detectors. The goal is not to memorise which current product has which feature. It is to develop methods that remain useful as products and terminology change.

Topics covered

Learners should finish this world able to ask informed questions, evaluate an AI tool’s strengths and risks, and decide when human expertise is essential.

Educational impact and learner benefit

AI Forest measures transfer rather than isolated recall. A learner may know what biased data is in Data Valley; the capstone asks how that bias could affect a deployed hiring system, how it might be detected and who should respond. This kind of cross-world reasoning is closer to real decision-making than a definition quiz.

The world’s greatest educational benefit is durability. Specific models, companies and laws will change, but habits such as checking evidence, distinguishing capability from marketing, tracing consequences and preserving human accountability remain useful. AI Forest therefore turns the curriculum from a tour of technology into preparation for informed participation in a society shaped by AI.

The complete learning journey

The ten worlds form one connected arc:

  1. Robot Valley: identify AI.
  2. Data Valley: understand what AI learns from.
  3. Pattern Mountain: understand what AI finds and represents.
  4. Smart City: understand how modern models train and infer.
  5. Future Lab: understand how AI systems are built and deployed.
  6. AI Safety & Ethics: judge whether AI is responsible.
  7. Prompt Plaza: communicate effectively with generative AI.
  8. Agent Academy: understand AI that plans and acts.
  9. AI Studio: create with AI responsibly.
  10. AI Forest: combine the ideas into durable AI literacy.

The intended outcome is not merely that a child can answer questions about AI. It is that the child can recognise AI, understand its foundations, use it carefully, challenge its output and make responsible decisions around it.

What success looks like

Completing all ten worlds should not be interpreted as mastering the entire field of artificial intelligence. The field is too broad and changes too quickly for that claim. Success means the learner possesses a coherent map. They know where data enters, how patterns become representations, how models train and run, how products connect models to tools and people, and where safety and accountability must be considered.

That map changes the questions a child asks. Instead of “Is this AI smart?” the learner can ask:

Those questions are the real product of the world system. Technical terms, stars and completion percentages support the journey, but informed curiosity is the destination.

MokingBird — Jogg Mini. Teach AI. One world at a time.

Published by MokingBird Oy.