Course parameters
- Instructional time
- 10 × 45–60 min
- Delivery
- Email-paced, self-directed
- Prerequisites
- None
- Materials
- Browser · no purchase
- Cost
- $0
- Student data collected
- None
Hybrid AI architecture. Module 5 is fully air-gapped — Gemma runs
on-device in the browser via WebGPU with the network disabled, so that work never leaves the
machine. The remaining nine modules use cloud inference, sending the prompt to Google's
Gemini API for processing. We operate no student database and retain no prompts; see the data
handling section below. Credentialing is optional throughout and requires parental consent.
Assessment model
Each module produces one concrete student artifact. Assessment is performance-based:
the artifact is the evidence. No module relies on a multiple-choice instrument, and no
student work is transmitted to or stored by the provider — teachers collect artifacts
through their existing LMS or classroom workflow.
Module sequence
01
Running an AI Model on Your Own Device
The Tech Shaman
45–60 min
Essential QWhere does an AI model actually run, and who can see what I type into it?
Objectives
- Initialize a language model running locally on the student's own device
- Contrast local inference with cloud inference in terms of data exposure
- Generate a unique identifier as proof of a completed local run
ArtifactA working local Gemma instance and a generated call-sign identifier
Vocabularymodel, inference, local vs. cloud, prompt, surveillance
StandardsCSTA 2-IC-23 CSTA 2-NI-05 ISTE 2d
02
Digital Identity & Proving Ownership
The Cypher-Nomad
45–60 min
Essential QHow can you prove something online belongs to you without a company vouching for you?
Objectives
- Explain the relationship between a public address and a private key
- Locate and read a public account address
- Describe why a public identifier can be shared but a private key cannot
ArtifactWritten explanation of public/private key roles, with the student's public address
Vocabularypublic key, private key, address, custody, verification
StandardsCSTA 2-NI-06 CSTA 3A-NI-06 ISTE 2b
03
Precision Prompting & Decoding
The Prompt Whisperer
45–60 min
Essential QHow much does the exact wording of an instruction change what a machine gives back?
Objectives
- Construct prompts that produce a specific, verifiable output
- Decode an encoded message using iterative prompt refinement
- Perform and check a multi-step operation produced by the model
ArtifactDecoded message plus the prompt sequence used to obtain it
Vocabularyprompt engineering, encoding, iteration, specificity, verification
StandardsCSTA 2-AP-13 CSTA 2-DA-07 ISTE 5c
04
Verifying a Claim Against a Public Record
The Urban Samurai
45–60 min
Essential QWhat makes a record trustworthy — and how would you check it yourself?
Objectives
- Independently look up a record in a public ledger explorer
- Compare a claimed credential against its published record
- Explain why a publicly auditable record resists tampering
ArtifactVerification write-up matching a credential to its public record
Vocabularyledger, explorer, immutability, audit, authenticity
StandardsCSTA 3A-IC-24 ISTE 3b ISTE 3d
05
Air-Gapped AI: Working With No Network
The Neon Ninja
45–60 min
Essential QCan a computer still be intelligent with the internet switched off?
Objectives
- Operate a language model with network access fully disabled
- Generate a hash value while air-gapped and explain what a hash represents
- Evaluate the trade-offs of edge computing vs. cloud computing
ArtifactOffline-generated hash plus a short trade-off analysis
Vocabularyair gap, hash, edge computing, latency, network isolation
StandardsCSTA 2-IC-20 CSTA 2-NI-05 ISTE 5a
06
Deepfakes, Hallucinations & Fact-Checking
The Solar Monk
45–60 min
Essential QIf an AI can state a falsehood confidently, how do you catch it?
Objectives
- Define hallucination and identify one in generated output
- Use a model to interrogate a fabricated news item against evidence
- Build a reusable checklist for evaluating source credibility
ArtifactStudent-authored fact-checking framework applied to a sample article
Vocabularyhallucination, deepfake, misinformation, corroboration, provenance
StandardsCSTA 3A-IC-25 ISTE 3b ISTE 2a
07
Explaining a Technical Concept by Analogy
The Tech-Bazaar Trader
45–60 min
Essential QDo you understand something well enough to explain it to someone who doesn't?
Objectives
- Distinguish a unique non-interchangeable asset from an interchangeable one
- Construct an original analogy that conveys the distinction accurately
- Test the explanation on a non-technical audience and revise it
ArtifactWritten or recorded analogy, plus a revision note after audience testing
Vocabularyfungible, non-fungible, scarcity, analogy, audience
StandardsISTE 6a ISTE 6d CSTA 2-AP-15
08
AI Personas as a Study Tool
The Steampunk Scrapper
45–60 min
Essential QCan you direct an AI to teach you something — and then check whether it taught you correctly?
Objectives
- Assign a model a persona to explain a mathematical concept
- Independently verify the mathematical accuracy of the explanation
- Assess where persona framing helps comprehension and where it distorts
ArtifactPersona-generated explanation with the student's accuracy verification noted
Vocabularypersona, system prompt, verification, pedagogy, bias
StandardsISTE 1c ISTE 3b CSTA 3A-IC-25
09
How Value Moves Between People Digitally
The Street Artist
45–60 min
Essential QWhat actually happens between "send" and "received"?
Objectives
- Trace the lifecycle of a transaction from initiation to confirmation
- Identify the intermediaries present in, or absent from, the path
- Explain what a confirmation does and does not guarantee
ArtifactAnnotated diagram of a transaction lifecycle
Vocabularytransaction, confirmation, intermediary, network fee, settlement
StandardsCSTA 2-NI-04 CSTA 3A-IC-24 ISTE 5a
10
Capstone: Build an AI Agent for Your Community
The Prime
90–120 min
Essential QWhat problem near you is worth solving, and can you build something that helps?
Objectives
- Identify an authentic problem in the student's own community
- Decompose it into requirements an agent could address
- Design, build, and test a working AI agent against those requirements
- Evaluate the result and articulate its limitations
ArtifactWorking AI agent + design rationale + limitations statement
Vocabularyagent, requirements, decomposition, iteration, limitation
StandardsCSTA 3A-AP-13 CSTA 2-AP-13 ISTE 4a ISTE 4c ISTE 5c
Suggested pilot configurations
| Configuration | Modules | Time | Approval needed |
|---|
| 4-week taster — AI literacy focus | 1, 3, 6, 8 | ~4 hrs | Email address only |
| 6-week unit — privacy & security | 1, 2, 3, 5, 6, 8 | ~6 hrs | Email address only |
| Full course — club or elective | 1–10 | ~10 hrs | + parental consent if credentialing |
| Terminal demo — single lesson | Live demo, any agent | ~1 hr | None — no signup at all |
Data handling
| Data | Processed by | Retention |
|---|
| Module 5 prompts | Student's own device (WebGPU, offline) | Never transmitted |
| Modules 1–4, 6–10 prompts | Google Gemini API | Not retained by us |
| Student name / school records | Not requested | Never collected |
| Email address | Course delivery only | Not sold or shared |
Our AI endpoint is a stateless pass-through and writes no prompt data to storage.
Cloud-processed prompts are handled under Google's API terms; districts requiring a
fully offline deployment should contact us to discuss options.
A note on standards alignment
Codes above reference the CSTA K–12 Computer Science Standards (2017 revision) and the
ISTE Standards for Students (2016 revision). These are proposed mappings offered as a
starting point for curriculum review — they are not a certification, and districts
should verify them against their own adopted framework and version. We are glad to work with
curriculum staff on alignment to a specific state framework on request.