Wibbit Curriculum Alignment at a Glance

Version: v2 — May 2026 Status: Published Coverage: Course 1 (fully shipped) + Course 2 (fully shipped) + Course 3 (fully shipped) Detailed analysis: Full alignment documentation is available to schools and districts on request — contact hello@wibbit.ai

---

Summary

Wibbit's curriculum is designed around a single question: what does a middle-schooler actually need to understand about the AI systems they're already using? Through Courses 1–3, that question now covers three distinct domains: how language AI works (C1–C2), how AI sees and creates (C3), and how AI-generated content affects trust and authenticity (C3-M5).

Courses 1 and 2 deliver LLM-depth coverage — tokenization through backpropagation, bias through adversarial examples — exceeding any K-12 framework's requirements on how AI learns. Course 3 extends into computer vision, generative AI, multimodal systems, and synthetic media. The combination covers four of AI4K12's Five Big Ideas with hands-on depth, closes several previously identified gaps, and opens a few new ones honestly named below.

The table below shows how Courses 1–3 map against the five major frameworks. Strong means the framework's requirements for this area are fully met. Partial means the concepts are introduced but coverage is thinner than the framework's full scope. Gap means the gap is confirmed and not yet addressed. Out of scope means the framework's requirement is in a domain Wibbit does not address (typically programming skills, not literacy skills).

---

Coverage by Framework

AI4K12 Five Big Ideas

AI4K12 is the primary US K-12 AI literacy reference — the framework educators and CS curriculum buyers check first.

Big Idea	Coverage	Notes
1 — Perception (sensors, vision, audio)	Strong	C3-M1 (A New Way of Seeing): pixels as number grids, drag-to-modify R/G/B sliders, convolution filter lab, edge detection as the first CNN layer. C3-M2 (Seeing in Layers): CNN elevator through 4 depth levels (edges → textures → parts → objects), pooling, image classifier confidence. Full hands-on coverage; exceeds 6–8 grade-band expectations on how visual AI perceives and processes input.
2 — Representation & Reasoning	Partial	Strong on connectionist representation: token encoding (C1-M1), embeddings and attention (C1-M4), CNN feature maps (C3-M1–M2), latent space geometry (C3-M3), shared image-text embedding space (C3-M4). Thinner on symbolic reasoning, search, and non-LLM approaches; C2-M2 introduces Five Tribes conceptually. Bayesian and evolutionary paradigms are names, not owned experiences.
3 — Learning	Strong	Wibbit's deepest coverage area. Training loop, loss, gradient descent, data quality, RLHF, supervised learning, labels, overfitting, backpropagation (C1-M3, C2-M1, C2-M3, C2-M4). C3-M3 adds diffusion model training as a distinct generative learning paradigm: students watch the reverse-noise process and understand it as learned un-noising. Exceeds 9–12 grade-band depth on several concepts.
4 — Natural Interaction	Partial→Strong	C1–C2: full coverage of language/LLM interaction (prompt engineering, hallucination, output evaluation). C3-M4 adds multimodal natural interaction: image patches as tokens, visual attention heatmaps ("which patches did AI look at?"), shared image-text embedding space, visual QA including absence detection. Missing: speech interfaces, dialogue system design, embodied AI. The multimodal addition substantially closes the gap for ages 10–13.
5 — Societal Impact	Strong	Hallucination introduced mechanistically (C1-M2, C1-M5), bias and stereotyping (C1-M5), adversarial examples and disparate impact (C2-M5). C3-M5 (Real or Fake?): deepfake timeline 2014–2025, detection arms race, artist consent and data rights, C2PA provenance standards. Exceeds typical 9–12 grade-band expectations on synthetic media and content authenticity.

AI4K12 grade band: Wibbit's primary target is the 6–8 band; on Big Ideas 1, 3, and 5, content reaches or exceeds 9–12 depth.

---

CSTA K-12 Computer Science Standards

The most widely adopted CS standards framework in US K-12 — the default alignment baseline in district RFPs.

Standard	Coverage	Notes
2-DA-07 (represent data, multiple encoding schemes)	Strong	Token-to-integer encoding (C1-M1); continuous vector encoding / embeddings (C1-M4); pixel as three integers 0–255 with interactive slider (C3-M1); image as 2D number grid (C3-M1). Multiple encoding schemes taught through interactive manipulation at every layer.
2-IC-21 (bias and accessibility in technology design)	Strong	Jobs, culture, and names bias scenarios (C1-M5); disparate impact and protected attributes (C2-M5). Among Wibbit's strongest alignments.
2-IC-20 (tradeoffs in computing technologies)	Partial	Strengths/weaknesses framing (C1-M5), hallucination as a structural limitation (C1-M2), guidance slider tradeoff in text-to-image (C3-M3), detection-vs.-generation arms race (C3-M5). Not structured as a single formal comparative analysis exercise, but tradeoff reasoning appears throughout the curriculum.
2-DA-08 (collect and transform data)	Partial	Training data concepts and data quality (C1-M3); labeled examples and supervised signal (C2-M1); image data representation and transformation through convolution (C3-M1). No student-facing data collection; conceptual layer only.
2-DA-09 (refine computational models)	Partial	Overfitting / train-test split / generalization (C2-M3). Students interact with the concept; no hands-on model retraining.
2-IC-23 (privacy tradeoffs)	Partial	C3-M5 covers consent and data rights specifically in the AI training data context: artist consent, opt-in vs. opt-out defaults, the scale of unconsented data use, artist-side tools (Glaze/Nightshade), and governance-side tools (EU AI Act). This is a depth-first treatment of one specific privacy context, not a general digital privacy survey. General data broker, surveillance, and personal privacy tradeoffs remain a gap.
2-AP-12–14 (control structures, procedures)	Out of scope	Programming standards; Wibbit teaches AI literacy, not programming. Intentionally out of scope.

CSTA AI Learning Priorities (2025): Five-category supplement to the 2026 standards revision.

Category	Coverage
1 — Humans and AI (pattern-matching vs. cognition; human oversight)	Strong — anti-anthropomorphism framing from first lesson; evaluating AI output quality throughout; C3-M5 adds evaluating synthetic media as a human cognitive task.
2 — Representation and Reasoning (encoding, knowledge representation)	Partial — strong on embeddings, attention, CNN feature maps; thinner on symbolic and Bayesian approaches.
3 — Machine Learning (training, data, model evaluation)	Strong — Wibbit's primary depth area; training loop through backpropagation; C3-M3 adds diffusion model training as a generative learning paradigm.
4 — Ethical AI System Design (fairness, accountability, testing)	Partial — bias and disparate impact covered deeply; adversarial robustness (C2-M5); detection limits and arms race (C3-M5). Auditing and system design are observer-level, not practitioner-level.
5 — Societal Impacts (environmental, IP, workforce effects)	Partial — societal harms and disparate impact (C1-M5, C2-M5); IP ownership and training data consent (C3-M5). Environmental impact of AI remains a gap.

---

ISTE Standards for Students

ISTE is the technology coordinators' checklist — no dedicated AI strand, so alignment is through existing standards.

ISTE Standard	Coverage	Notes
Computational Thinker (5)	Partial	Algorithmic thinking (training loop, predict-check-adjust), data systems (training data, encoding, image as numbers), model behavior (CNN layers, diffusion process). No programming or formal algorithm design.
Digital Citizen (2)	Strong	AI ethics and fairness (C1-M5, C2-M5); hallucination and verification habits (C1-M2, C1-M5); synthetic media and deepfake detection literacy (C3-M5); consent and data rights (C3-M5); content provenance (C3-M5). C3-M5 strengthens Digital Citizen alignment substantially — students are equipped to evaluate AI-generated content authenticity.
Innovative Designer (4)	Partial	C3-M3 introduces creating with generative AI: latent space navigation (drag dot → see face morph), text-to-image with guidance slider, discovering the creativity-fidelity tradeoff. C3-M4 adds multimodal creation and understanding. Students create with AI tools but do not yet design AI systems end-to-end.
Knowledge Constructor (3)	Partial	Prompt engineering as a research tool (C1-M5); evaluating AI sources for accuracy and bias; C3-M5 adds lateral reading and source triangulation for AI-generated content.

---

UNESCO AI Competency Framework for School Students (2024)

UNESCO's 2024 framework is the primary international credibility signal for cross-border contexts.

Competency Area	Coverage	Notes
AI Fundamentals	Strong	How AI systems work (tokenization through transformers; pixels through CNNs; latent space through diffusion models); training and learning mechanisms; LLM and vision AI behavior and limitations.
Critical Evaluation of AI Outputs	Strong	Hallucination detection and verification habits (C1-M5); adversarial examples (C2-M5); bias identification (C1-M5, C2-M5); deepfake detection and detection limits (C3-M5); C2PA provenance reading (C3-M5). The addition of synthetic media literacy in C3-M5 substantially deepens this competency.
AI Ethics	Strong	Bias, fairness, and disparate impact throughout C1-M5 and C2-M5; RLHF as values-shaping process (C1-M3); artist consent and data rights (C3-M5); detection arms race ethics (C3-M5); EU AI Act governance (C3-M5). "Who gets hurt when machines fail" is first-class curriculum across all three courses.
AI Design and Creation	Partial	C3-M3 covers creating with generative AI tools (latent space navigation, text-to-image with guidance); C3-M4 covers understanding how multimodal AI architectures work. Students create with AI and understand AI architecture at the level of informed users, not engineers. UNESCO's full "AI Design and Creation" competency implies system-level design choices; that remains a gap.
Privacy in AI Contexts	Partial	C3-M5 covers data rights in the AI training data context: artist consent, unconsented scraping at scale, opt-out mechanisms, governance-side protections. General digital privacy (surveillance, personal data brokers, terms of service) is addressed in spirit but not as a dedicated module. General digital privacy is a confirmed gap for future content.

Notable gap (unchanged from v1): UNESCO strongly emphasizes equity, linguistic diversity, and cultural accessibility. Wibbit's current content does not explicitly address these dimensions. This is a curriculum consideration for international expansion and a framing consideration for international market positioning.

---

OECD/EC AILit Framework (Review Draft, May 2025)

The fastest-rising international framework; final version with grade-level exemplars expected 2026.

Interaction Domain	Coverage	Notes
Engaging with AI (recognize, critically evaluate, understand)	Strong	Wibbit's core curriculum lives here — conceptual understanding of LLM and vision AI, critical evaluation of outputs, bias and failure modes, synthetic media literacy (C3-M5). Anti-anthropomorphism framing from first contact.
Creating with AI (collaborate, ideate, problem-solve)	Strong	C3-M3 (AI the Artist) directly addresses this domain: students navigate latent space to explore the AI's "idea map," use text-to-image generation with a guidance slider, and discover the creativity-vs.-fidelity tradeoff through interactive experimentation. C1-M5 (prompt engineering) provides the language-AI foundation. Full coverage for the 10–13 age range interpretation of this domain.
Managing AI (strategically delegate, enhance human work)	Not yet	Beyond C1–C3 scope.
Designing AI (understand principles, fairness, societal impact)	Not yet	System-level AI design decisions (training data choices, fairness constraints, deployment tradeoffs) are not yet a student-facing design task in Wibbit's curriculum. The societal impact dimension is covered (C1-M5, C2-M5, C3-M5); the design-decision dimension is not.

---

What Wibbit Teaches Beyond Any Framework

Through Courses 1–3, Wibbit goes beyond what any framework requires on a specific cluster of concepts. This is deliberate. The depth-first trade-off is Wibbit's core positioning.

Wibbit concept	Status in external frameworks
Transformer architecture internals (embeddings, attention, layer-wise understanding)	Not required by any K-12 framework; Wibbit opens the black box
RLHF as a named, teachable concept	Not addressed in any current K-12 standard
Five Tribes / ML paradigm diversity	No framework asks students to understand competing ML philosophies
Hallucination explained mechanistically (next-token prediction → confident-but-wrong)	Frameworks name the phenomenon; Wibbit explains the mechanism
Overfitting, train/test split, generalization at 9–12 depth	C2-M3 substantially exceeds any 6–8 grade-band standard
Backpropagation as an earned concept	Not required by any K-12 framework; Wibbit teaches it experientially
Anti-anthropomorphism from first contact	CSTA AI Learning Priorities Cat. 1 names this outcome; Wibbit places it in the first lesson, not a late-stage module
Convolution and CNN feature maps taught interactively	AI4K12 Big Idea 1 requires students to understand computer perception; Wibbit teaches it at the filter-mathematics level through a hands-on filter lab
Latent space geometry made navigable	No framework requires students to interact with a 2D latent space; Wibbit teaches it as a drag-and-explore experience
Detection arms race framed as systemic, not solvable	Frameworks acknowledge AI-generated content risks; Wibbit explicitly teaches why detection cannot "win" — the framing of an arms race, not a solved problem
C2PA / content provenance standards at middle school level	No K-12 framework addresses content credential standards; Wibbit teaches it through a ProvenanceInspector interaction and explains the adoption-dependency problem
GAN historical framing (Faker/Detective animation, explicitly "the first big idea — today's tools work differently")	No framework covers the history of generative model development; Wibbit teaches it to build correct mental models of why current tools differ from 2014 approaches

---

Remaining Gaps

The following gaps were identified in v1 and remain open after Course 3:

Gap	Frameworks citing it	Status
General digital privacy (data brokers, surveillance, personal consent)	CSTA 2-IC-23; UNESCO Privacy; AILit digital wellbeing	Open — C3-M5 covers training data consent; general digital privacy is not yet a Wibbit module
Environmental impact of AI	CSTA AI Learning Priorities Cat. 5; AILit Domain 1	Open — not addressed in Courses 1–3
Full AI system design (not creation with tools, but designing systems)	AILit "Designing AI"; ISTE Innovative Designer (full scope)	Partial — C3-M3–M4 cover creating with and understanding generative AI; designing systems with chosen tradeoffs is not yet taught
UNESCO equity, linguistic diversity, cultural accessibility	UNESCO core posture	Open — a curriculum and positioning consideration for international expansion

The following gaps identified in v1 are now closed by Course 3:

Gap	Resolved by
Perception / computer vision / sensor-based AI	C3-M1 (A New Way of Seeing) + C3-M2 (Seeing in Layers)
IP ownership and training data rights	C3-M5 (consent and data rights, Glaze/Nightshade, EU AI Act)
Creating with AI (OECD/AILit domain)	C3-M3 (AI the Artist)
Synthetic media and deepfakes as a literacy domain	C3-M5 (Real or Fake?)
Multimodal natural interaction	C3-M4 (Words and Pictures Together)

---

For School and District Procurement

If you are evaluating Wibbit for classroom use or district adoption, the following documentation is available on request:

• Full alignment report — concept-level citations mapping Wibbit modules to specific AI4K12 grade-band outcomes, CSTA standards, ISTE outcomes, UNESCO competencies, and AILit domains
• Alignment statement letters — signed documentation of framework alignment for use in grant reporting or RFP responses
• Classroom implementation guide — how Wibbit's Lesson → Experiment → Challenge structure maps to classroom periods

Contact hello@wibbit.ai with "Standards Documentation" in the subject line.

---

Version Notes

Version	Date	Coverage	Notes
v1	May 2026	Course 1 + Course 2	Initial publication. Course 3 gaps documented as planned.
v2	May 2026	Course 1 + Course 2 + Course 3	Full Course 3 audit. AI4K12 Big Idea 1 upgraded to Strong; Big Idea 4 upgraded; OECD/AILit "Creating with AI" upgraded to Strong; ISTE Digital Citizen strengthened by C3-M5. Remaining gaps table updated.

Alignment is assessed against framework versions known as of May 2026. Specific framework version references:

• AI4K12: Five Big Ideas grade-band progressions (2022) + AI Learning Priorities report (May 2025)
• CSTA: K-12 CS Standards 2017 (v2.1) + AI Learning Priorities supplement (2025)
• ISTE: Standards 2024 update + "Bringing AI to School" (2023)
• UNESCO: AI Competency Framework for School Students (September 2024)
• OECD/AILit: Review draft (May 2025); final version pending

This document will be updated when the CSTA 2026 revision publishes (expected summer 2026) and when Course 4 modules ship.

---

Part of the Wibbit AI Literacy Standards series.