For non-specialists who need to follow the AI safety conversation: understand what alignment means, how labs try to achieve it, where it fails, and how frontier risk is governed today.
Level
Basic
Study time
6 hours
Lessons
18 in 6 modules
Exam
20 questions
About this course
Build six compact records for fictional Lakefront Learning Cooperative. Use supplied transcripts, tables and source cards to distinguish goals, behavior, evidence and authority. Compare historical research mechanisms and current attributed framework roles, then revise a priced preparation plan after source and capacity changes. No coding, paid model, real data or operational action is required.
This six-hour basic course assesses textual source/evidence reasoning, not model training, empirical research, runtime enforcement, legal clearance or workplace safety certification. The current-law access limits are explicit. Completion follows lessons, listed Decision Labs played to any ending and the exam; optional fallible recommendation feedback is practice.
What you will be able to do
Apply overlapping alignment/safety/security/misuse lenses to a source/action boundary
Repair a proxy and distinguish observed behavior from claims about learned internal objectives
Compare demonstrations, preferences, principle-guided training, written policies and runtime controls
Interpret sycophancy/context/autonomy observations with rival explanations and discriminating test plans
Calculate evidence fractions and distinguish regression, fresh evaluation, association and intervention
Compare attributed publisher framework roles, triggers and evidence without false equivalence
Separate source force, actor, task and version with US applicability questions explicit
Defend and revise a feasible priced strategy with concrete evidence, trade-offs and feedback decisions
Syllabus
6 modules · 18 lessons · about 6 hours
1
Module 1: Intent and proxy
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
One Assistant Request Can Raise Several Safety Questions
Repair a Proxy That Rewards Closing the Wrong Request
Separate a Bad Objective from a Claim About an Internal Goal
Decision Lab · 5 minThe clarification counted as failureA manual repair is useful without an inferred internal objective or completed booking. Wider preparation covers more counterexamples, not model internals.
2
Module 2: Shaping behavior
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
Preference Training Needs a Reference Worth Learning
A Principle Can Guide Critique Without Becoming an Enforcement Result
Helpful Answers and Runtime Permissions Need Separate Controls
Decision Lab · 5 minThe favorite draft contradicts the sourceCompare a warm P1 error with a new P2 source-correct agreement that the supplied label penalizes; preserve reference, training and runtime boundaries.
3
Module 3: Context and autonomy
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
Test Agreement Against the Source, Not the User's Certainty
An Evaluation Marker Changes Context, Not the Evidence Standard
New Tools Change Consequences Before They Establish Permission
Decision Lab · 5 minThe booking partner is a new recipientThe new tool proposal changes possible consequences, not proof of a motive. Retained source drafts remain useful; booking permission and tests remain pending.
4
Module 4: Evidence and interpretation
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
Three Error Fractions Answer Three Different Questions
Keep Regression Evidence Separate from a New Evaluation
A Feature Association Is Not a Complete Explanation
Decision Lab · 5 minV2 reopens the evidence claimA new source context needs relevant evidence. Full prepares broader design, narrow documents known-case regression, and neither supplies fresh V2 results.
5
Module 5: Frontier sources and US applicability
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
Compare Frontier Frameworks Without Flattening Their Commitments
Keep a Source's Version, Evidence Date and Force Together
Ask the US Applicability Question Before Borrowing a Duty
Decision Lab · 5 minThe current title opens the old bodyA source receipt can be useful historically while inadequate for a new claim. Peer and full have equal price but different comparison scope; neither is local permission.
6
Module 6: Recommendation and revision
Apply the module concepts to the evolving Lakefront source-bearing record.3 lessons
Turn Six Records into a Bounded Recommendation
Reprice Scope When Staff Capacity Falls
Challenge Feedback and Revise the Actual Recommendation
Decision Lab · 5 minFour hours leave the preparation windowCash does not replace aggregate effort. Supported conditional/manual work is complete when source, scope, trade-off, owner and next evidence are explicit.
Capstone simulation · 20 minLakefront: scope after the capacity revealCompare priced preparation scopes, retain actual evidence and revise after source/capacity changes.
Exam and certificate
The exam unlocks once you have completed every lesson and played each Decision Lab to an ending. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Anthropic, OpenAI, or Google DeepMind. Product and company names are trademarks of their respective owners.
A free, plain-English introduction taught through one realistic company getting ready for its busiest season. Sort the kinds of AI, see why a chat assistant can sound right and be wrong, decide which tasks it fits and how to measure them, catch its three predictable failures, ask well, and decide what never goes into an AI tool.
A plain-language grounding in artificial intelligence: what AI, machine learning and generative AI actually are, how chat assistants produce text, why they sometimes get things confidently wrong, and how to use them safely for real work.
A guided tour of the machinery inside a large language model, told in plain-language analogies, tables and small runnable examples rather than equations: tokenizers, embeddings, attention, pretraining, instruction tuning, reasoning models, context windows and sampling. You leave able to explain why models behave the way they do.