Brightlab
brightlab.Made for curious mindsbrightlab.Made for curious mindsBrightlab
The curriculum explorer
Three substantial lessons for every year. Follow the sequence or find the next idea your class is ready to explore.
Share or bookmark this address to keep your filters. Preparation times are planning estimates. Every classroom pack also includes a 45- and 60-minute focus plan.
39 investigations · Every core lesson runs locally in a browser.
Notice & make
Instructions · Digital Technologies
How can we help a rover that does exactly what we say?
Start with: Point forward, then turn your body without moving your feet.
Classification · Mathematics
Can the same shape belong in different groups?
Start with: Compare two objects and say which is bigger.
Privacy · Digital Technologies
What does a weather helper really need to know?
Start with: Choose a coat for a rainy day.
Notice & make
Patterns · Mathematics
Can two different rules explain the same beginning?
Start with: Continue a repeating two-symbol pattern.
Algorithms · Digital Technologies
When should a rule take a different path?
Start with: Follow a two-step instruction and answer a yes/no question.
Sensors · Science
Can the same object become invisible to a sensor?
Start with: Explain the difference between an object and a picture of it.
Notice & make
Classification · Digital Technologies
How does a machine name a creature it has never seen?
Start with: Compare lengths and count spots using two features.
Training data · Science
Is a bigger collection always a better collection?
Start with: Count examples inside and outside a drawn boundary.
Privacy · Mathematics
Can harmless-looking clues fit together to identify someone?
Start with: Group fictional cards by a visible feature.
Train & test
Representation · Digital Technologies
What does a picture lose when we turn it into numbers?
Start with: Count filled squares in a small grid.
Evaluation · Mathematics
How do we find out whether a machine learned beyond its practice?
Start with: Separate examples used to choose a rule from examples used to test it.
Automation · Digital Technologies
Should one uncertain signal be allowed to trigger an action?
Start with: Read a yes/no sensor result and follow a simple rule.
Train & test
Evaluation · Science
When is asking a person better than choosing a bin?
Start with: Count missed targets separately from false alarms.
Training data · Mathematics
Who disappears inside an average?
Start with: Compare two groups and describe a fair test.
Robustness · Science
Why can an unchanged model stop working?
Start with: Keep a test rule fixed while changing one environmental condition.
Train & test
Similarity · Mathematics
Who decides what a good match looks like?
Start with: Measure distance on a feature map.
Feedback loops · Digital Technologies
Can a recommendation create the popularity it measures?
Start with: Describe how collected examples affect a later choice.
Objectives · English
What happens when we reward the easy thing to count?
Start with: Distinguish a measured score from the purpose it represents.
Look inside
Generation · English
Does a likely next word make a true sentence?
Start with: Read a probability bar and distinguish a guess from evidence.
Context · English
What happens when the fact you need falls out of view?
Start with: Identify a fact and a condition in a short instruction.
Evidence · English
What must a source actually say to support a claim?
Start with: Separate an assertion from the passage offered as evidence.
Look inside
Evaluation · Digital Technologies
Is a brilliant score hiding information from the future?
Start with: Separate training and testing and order events by time.
Evaluation · Mathematics
Which mistakes should an alert system try hardest to avoid?
Start with: Count false positives and false negatives.
Systems · Digital Technologies
Can correct code still pass the wrong numbers to a model?
Start with: Identify a value, its units and its source.
Look inside
Neural networks · Mathematics
What can one artificial neuron actually compute?
Start with: Substitute values into a weighted sum.
Evaluation · Mathematics
When does a better fit become a worse prediction?
Start with: Read a coordinate plot and distinguish training from testing.
Provenance · English
What does an intact record prove about a picture’s caption?
Start with: Explain the difference between evidence and a source claim.
Build systems
Optimisation · Mathematics
Why can trying harder move a model further from its goal?
Start with: Evaluate a quadratic and follow a repeated numeric update.
Unsupervised learning · Mathematics
Do clusters discover categories, or help us invent them?
Start with: Compute or compare distances between points.
Uncertainty · Mathematics
When a model says 90%, how often should it be right?
Start with: Interpret frequencies, probabilities and sample sizes.
Build systems
Retrieval · Digital Technologies
Can a cut in the wrong place change what an assistant finds?
Start with: Find a claim and its qualifier in a passage.
Retrieval · English
When should an assistant answer, and when should it stop?
Start with: Match claims to evidence and inspect retrieved passages.
Security · Digital Technologies
What stops a source from becoming an instruction?
Start with: Distinguish data from instructions and identify a permission boundary.
Design & defend
Deployment · Digital Technologies
What is the best system when quality is only one constraint?
Start with: Interpret latency distributions and compare cost and quality trade-offs.
Security · Digital Technologies
What must be true before a proposed action is allowed?
Start with: Read JSON and apply type, range and permission checks.
Monitoring · Mathematics
How will we notice a deployed model is no longer behaving as expected?
Start with: Compare a reference distribution with recent observations.
Design & defend
Governance · HASS
Can a popular pilot still be unacceptable to the people it affects?
Start with: Separate technical performance, consent, representation and accountability.
Evaluation · Science
How can we tell which part of a system actually helped?
Start with: Keep cases paired across treatments and interpret variation.
Governance · Digital Technologies
What would make a system’s public claims defensible?
Start with: Distinguish a claim, relevant evidence, an assumption and a defeater.