Learn AI.
My WorkYour learning, collected
ONE STEP AT A TIME

Pick up where you left off.

Your progress, experiments, and discoveries. Saved in this browser, ready when you are.

Saved in this browser · Back up your work. No automatic sync between devices or site addresses.

YOUR NEXT STEP

Your first neuron

Understand Neural Networks · try

Start learning
YOUR GOAL

Choose where to begin.

Hands-on starter paths for students, software engineers and managers. Skip familiar topics, keep your work, and explore freely.

Choose a learning path Explore the expanded catalog

Ideas to review

No completed lessons need an understanding check. Checks appear here as you work through the course.

LESSONS COMPLETED0 / 134 available

Projects and independent transfer checks are tracked separately below.

UNDERSTANDING CHECK0 / 134 passed

Tracked separately from lesson completion.

Progress by course

Understand Neural Networks

How does a network learn?

Your first neuronTo explore · Check to tryFollow the slope of a mistakeTo explore · Check to tryMake a problem easier to seeTo explore · Check to trySend the error backwardTo explore · Check to tryLearn from small batchesTo explore · Check to tryUse momentum and adaptive stepsTo explore · Check to trySchedule the learning rateTo explore · Check to tryInitialize and normalizeTo explore · Check to tryTrace disappearing and exploding gradientsTo explore · Check to tryMake numerical computation trustworthyTo explore · Check to tryChoose complexity using development evidenceTo explore · Check to tryCompare manual and automatic gradientsTo explore · Check to try

What has the network learned?

Learn the pattern, not the examplesTo explore · Check to tryWhat does this neuron notice?To explore · Check to tryCatch a model taking a shortcutTo explore · Check to trySlide a detector across an imageTo explore · Check to tryTeach useful invariancesTo explore · Check to tryCompare convolution with image attentionTo explore · Check to tryTurn sound into featuresTo explore · Check to tryOrder changes meaningTo explore · Check to tryLearn from relationships between examplesTo explore · Check to tryReuse a representationTo explore · Check to tryLearn with objects and relationshipsTo explore · Check to tryLocate and segment objectsTo explore · Check to tryEstimate depth and poseTo explore · Check to try

How does AI process language?

Choose the pieces a model predictsTo explore · Check to tryGive words a place in spaceTo explore · Check to tryCompare, normalize, combineTo explore · Check to tryFollow a token through a TransformerTo explore · Check to tryPredict the next tokenTo explore · Check to tryPrepare a text corpusTo explore · Check to tryTrain a tiny language modelTo explore · Check to tryChoose how to sampleTo explore · Check to tryAdapt a pretrained modelTo explore · Check to tryLearn from preferencesTo explore · Check to tryConnect instruction tuning, preferences and verified rewardsTo explore · Check to tryUse a pretrained model responsiblyTo explore · Check to tryEvaluate language tasks with evidenceTo explore · Check to try

Why does AI need computation?

Spend a limited compute budgetTo explore · Check to tryPackage a model as a componentTo explore · Check to tryMeasure and optimize inferenceTo explore · Check to tryVersion data and experimentsTo explore · Check to tryDeploy with a reversible planTo explore · Check to tryTest the complete systemTo explore · Check to tryMonitor what changedTo explore · Check to tryRespond to an incidentTo explore · Check to try

Data audits and training diagnosis

What should the machine predict?To explore · Check to tryExamples are measurementsTo explore · Check to tryDecide where AI belongsTo explore · Check to tryRepair data without leaking answersTo explore · Check to tryRepair a broken training runTo explore · Check to tryDifferent mistakes have different costsTo explore · Check to tryHow much should a small result convince you?To explore · Check to tryWho receives the errors?To explore · Check to tryUse data with a purposeTo explore · Check to tryExplain what your model is forTo explore · Check to try

Simple models and useful baselines

Start with a simple baselineTo explore · Check to tryFit a line and explain its errorsTo explore · Check to tryPredict a class probabilityTo explore · Check to tryCombine many imperfect treesTo explore · Check to tryCorrect residual errors with boostingTo explore · Check to tryFind structure without labelsTo explore · Check to tryFind unusual examplesTo explore · Check to tryRecommend from incomplete feedbackTo explore · Check to tryValidate without reusing the answersTo explore · Check to tryFinish a tabular experimentTo explore · Check to try

Generative models beyond text

Compress and reconstructTo explore · Check to tryGenerate with a variational autoencoderTo explore · Check to tryInfer what cannot be directly observedTo explore · Check to tryGenerate by removing noiseTo explore · Check to tryGenerate by learning a flowTo explore · Check to tryControl the generated resultTo explore · Check to tryLearn with a generator and discriminatorTo explore · Check to tryJudge a generated distributionTo explore · Check to try
Build AI Systems

Learning through actions

Explore before you decideTo explore · Check to tryLearn from delayed rewardsTo explore · Check to trySolve the Bellman equationsTo explore · Check to tryWhat can the agent observe?To explore · Check to tryCopy a demonstrated actionTo explore · Check to tryLearn from verified self-playTo explore · Check to tryKnow which policy produced the dataTo explore · Check to tryStabilize neural Q-learningTo explore · Check to tryIncrease the probability of useful actionsTo explore · Check to tryUse a critic to guide learningTo explore · Check to tryLimit a policy updateTo explore · Check to tryTrade off reward and exploration with entropyTo explore · Check to tryLearn without new interactionsTo explore · Check to tryEvaluate learning fairlyTo explore · Check to try

Planning and reasoning

Reason with procedures and searchTo explore · Check to tryUse an informed searchTo explore · Check to trySolve a constraint problemTo explore · Check to tryUpdate beliefs from evidenceTo explore · Check to tryFactor a probabilistic modelTo explore · Check to tryInfer state across timeTo explore · Check to tryApproximate a belief with particlesTo explore · Check to tryReason about an opponentTo explore · Check to tryAsk what would change under interventionTo explore · Check to tryPredict what an action will changeTo explore · Check to tryRefine an idea over several passesTo explore · Check to tryTest the parts of a planning systemTo explore · Check to try

Learning without losing old skills

Build a trustworthy practice loopTo explore · Check to tryLearn something new without losing the oldTo explore · Check to tryLearn a better way to adaptTo explore · Check to try

Connecting tools and specialists

Define an assistant’s jobTo explore · Check to tryCall a tool, then check the evidenceTo explore · Check to tryHelp specialists understand each otherTo explore · Check to tryLearn a shared representation contractTo explore · Check to tryChoose where a connected system learnsTo explore · Check to tryKeep documents from becoming instructionsTo explore · Check to try

Memory and shared working state

Find relevant informationTo explore · Check to tryAnswer with evidenceTo explore · Check to tryShare the right information at the right timeTo explore · Check to tryRemember selectively and retrieve evidenceTo explore · Check to try

Choosing what to do next

Decide whether another computation is worth itTo explore · Check to tryCoordinate dependencies and reusable skillsTo explore · Check to tryLearn from routing evidence without fooling yourselfTo explore · Check to tryTrain and evaluate a small neural routerTo explore · Check to try

Testing whether improvement is real

Measure the whole workflowTo explore · Check to tryTest a proposed system improvementTo explore · Check to tryFind out whether improvement generalizesTo explore · Check to try

Assemble an experimental agent

Connect an agent’s decisions, actions, and learningTo explore · Check to tryBuild an architecture that earns its complexityTo explore · Check to try

Geometry and control foundations

Describe where the robot isTo explore · Check to tryRepresent motion in three dimensionsTo explore · Check to tryUse a JacobianTo explore · Check to tryTurn a target into motionTo explore · Check to tryCorrect errors using feedbackTo explore · Check to tryChoose continuous actionsTo explore · Check to tryStop when the controller leaves its limitsTo explore · Check to tryTurn observations into estimatesTo explore · Check to tryFuse noisy measurementsTo explore · Check to tryReason about forces and contactTo explore · Check to tryPlan a collision-free trajectoryTo explore · Check to tryEnforce operating limitsTo explore · Check to try

A verified simulated controller

Report draft · 0 saved configurations · human review pending.

Open controller project

A reward-learning agent under evaluation

Report draft · 0 saved captures · 0 known evaluation openings · first-attempt transfer 0/5 · human review pending.

Open reward-agent project

A planner with uncertain observations

Report draft · 0 saved captures · human review pending.

Open planner project

A controlled generative experiment

Report draft · 0 saved snapshots · first-attempt transfer 0/5 · human review pending.

Open generative project

A perception system under stress

0 saved captures · no local reserved reveal · first-attempt transfer 0/3 · human review pending.

Open perception workbench

A generator and an adaptation comparison

Report draft · 0 saved snapshots · first-attempt transfer 0/4 · human review pending.

Open language project

A model with an operating plan

Report draft · 0 saved captures · first-attempt transfer 0/2 · human review pending.

Open operating-plan project

An assistant that uses evidence

Report draft · 0 saved comparisons · human artifact review pending.

Open evidence-assistant project

A real-data model comparison

Report draft · 0 retained comparisons · 0 known evaluations. Public-data exposure and human review remain explicit.

Open tabular comparison project

An evidence-based AI decision

Report draft · guided variant. First-attempt transfer: not submitted. Human review pending.

Open cumulative decision project

Foundation projects and assessments

Audit a dataset and defend the evidence

0/4 captured runs · guided variant · report draft.

First-attempt objective cases: 0/2. Latest attempts: 0/2. Written reasoning and implementation review: pending.

Open data audit project

Diagnose, repair and report a training failure

0/4 captured runs · guided variant · report draft.

First-attempt objective cases: 0/2. Latest attempts: 0/2. Written reasoning and implementation review: pending.

Open training diagnosis project

Additional training-debugging transfer

First-attempt cases: 0/4. Latest attempts: 0/4. These additional checks do not change existing path requirements. Human artifact review remains pending.

Open debugging transfer cases

Your classifier project

Understand Neural Networks milestone

Build and evaluate a classifier

0 training epochs · 0 saved candidates · Model editable.

Understanding check: to try. Finish the original Understand Neural Networks source lessons and classifier project to record this milestone. New curriculum activities have separate path requirements.

Open classifier project

Your agent project

Build AI Systems milestone

Build and evaluate a puzzle agent

0 training updates · 0 candidates · Developing. Understanding check: to try.

Complete the original Build AI Systems source lessons and agent project to record this milestone. New curriculum activities have separate path requirements.

Open agent project

Changed experiments

Only experiments with settings or results different from their starting state appear here. Browse all experiments

Recently changed lesson experiments

No changed lesson experiments yet.

Optional research

Your research notebook

Protocol and stage notes are included in progress backups. Export comparison results before leaving the research page.

Continue research
Optional coding

Your coding notebook

Six program drafts are included in progress backups. Imported code is not run automatically. Practice results stay separate from course completion.

Continue coding

Take your progress with you.

Your work stays on this device. Export a copy before clearing browser data, or import it on another device. Import replaces the progress currently shown here. Recorded exposure to the fixed test sets is retained, so a backup cannot make previously viewed examples unseen.

Learn AI · Understand by doing.Loading local progress…Copyright (c) 2026 Mobi Weave, Inc. All rights reserved