Classes 8-9 · Year two • Ama Dablam SummitProduct & systems

Choose how the system should be built.

By Class 9, the question is no longer only 'Can you implement this architecture?' The learner increasingly decides how the problem should be represented, which data structures and algorithms fit, how the software should be divided, how the model is evaluated and where human judgment belongs.

“I can build and defend increasingly mature systems.”

Who owns whatThe problem is supplied; the learner increasingly owns computational formulation, architecture, data/model decisions, trade-offs, reliability and technical defense.

“I can implement a supplied intelligent-system architecture.”

“I can formulate and architect increasingly mature computational/data/AI solutions and defend the choices.”

What changes this year

Data-structure and algorithm choice becomes serious, including graph/state-space reasoning.
Modular software, external libraries, APIs and persistent data become normal.
Students navigate and modify repositories/code they did not create.
Data work includes schemas/data dictionaries, validation, transformations and multivariate reasoning.
ML moves into a proper train/validation/test cycle with stronger metrics and error analysis.
AI-generated code must be explained, traced, tested, modified and rejected when inappropriate.
Reliability and technical trade-offs become part of the design—not cleanup at the end.
Technical defense becomes a stronger evidence requirement.

Above ordinary 'teen coding class'

Try the thinking

A minute of Ama Dablam Summit thinking.

An illustrative interaction, not a lesson from the curriculum — but the reasoning is the reasoning.

The model is accurate. The evidence is broken.

Interactive example

By Class 9, the learner is not only implementing a model. They must reason about architecture, data and whether the evaluation can be trusted.

The systemAn emergency resource planner predicts demand from historical incidents.

How the pipeline is wired

Inspect a component

Select any component in the pipeline to see what feeds it.

What is the most serious evidence problem?

What they produce

Multi-module software
Architecture/data-flow diagrams
Graph/search or other nontrivial algorithmic components
Validated/cleaned datasets
Model evaluation and error analysis
Tests and reliability evidence
AI-generated-code verification
Technical defense

What you should be able to see

Why a representation/data structure was chosen
An algorithm comparison
Architecture the learner increasingly shaped
Data cleaning/transformation and evaluation evidence
A confusion matrix or equivalent model-error analysis where appropriate
AI-generated technical output the learner challenged or repaired
A reliability trade-off the learner can defend

Support matters. A heavily guided success is not reported as independent mastery.

Twelve Projects. One climb.

A Band is not a pile of unrelated activities. Each standard Band contains 12 Projects arranged so that capability is established, combined, transferred and increasingly owned.

P1–P4Establish / Acquire
P5–P8Combine / Extend
P9–P12Transfer / Own

Illustrative capstone family

Intelligent Emergency Resource Planner

A learner might represent routes/resources as a graph, compare route strategies, process historical data, train/evaluate a demand or risk model, combine probabilistic predictions with deterministic rules, build an interface and explain where human judgment must override the system.

Illustrative advanced project family from current Band design; not a guaranteed final Project title.

Where Ama Dablam Summit lands on each capability

These are Band end-states, not lesson titles. Follow any of them across all eight years in the Capability Telescope.

Algorithms & problem solving: Choose serious data structures and search/reasoning approaches, including graph/state-space ideas, and evaluate correctness, clarity and scale.
Programming & representation: Navigate modular software, APIs and larger unfamiliar programs; treat AI-generated code as a proposal to verify.
Debugging: Debug modular systems and challenge plausible AI-generated code that contains logic, data or evaluation flaws.
Testing & reliability: Test modular software, edge cases and model behavior; reason about train/validation/test evidence and reliability.
Data & evidence: Use multivariate data, schemas/data dictionaries, validation, transformations and stronger uncertainty/evaluation reasoning.
AI & machine learning: Run a proper ML cycle with train/validation/test reasoning, model-family choices, bias/privacy/harm analysis and stronger AI-vs-non-AI judgment.
Systems & architecture: Increasingly choose modular architecture, APIs, persistent storage and boundaries between major responsibilities.
Technical judgment: Defend data structures, algorithms, model choices, reliability and where human judgment must override the system.
Security, privacy & human-centered computing: Evaluate dataset bias, harms, human involvement and system/user constraints.
Explanation, transfer & technical defense: Defend data structures, search strategy, evaluation evidence, AI-generated code verification and reliability trade-offs.

Open the Capability Telescope

What becomes possible next

Then Class 10 deliberately pauses the standard Band sequence

The pathway protects exam-year capacity rather than forcing a normal Band into Class 10. Prior capability remains part of the learner history. Pumori Summit begins by reactivating what matters, then accelerates into serious engineering.

See the Class 10 continuity plan

What prepared them: Ama Dablam Basecamp See the whole climb

The whole climb