← All work

Learning Systems Lab

Pick a question, change one thing, and watch the reasoning respond.

A live lab of small, authored learning experiments and Learning Toys, re-founded from the 2025 learning platform after CertGraph.

The idea

Learning Systems Lab is for curious people who learn by poking at things, and it doubles as Christian's workbench. Technical material often arrives as dense dashboards or long explanations that show more while making the reasoning harder to follow. The lab tries the opposite: one focused question at a time, where a single change, like launching a ball faster or saving for longer, visibly reshapes the model in front of you.

Two kinds of pieces share the space. Experiments are short, ordered sequences with a question, an action, and a traceable answer. Learning Toys, such as a graph of the Apollo missions, are open ground with no score and no required path. The bet is that authored interactions, where every response traces back to a rule or a source, make reasoning easier to inspect than generated explanations, and that ideas proven here can feed focused products like CertGraph.

Today the lab is live. It was released on 5 September 2026 with four authored experiments, five Learning Toys, and a public lineage page tracing the graph idea back through myDNAHealth and the AWS Glossary Graph. Nothing in it has been tested with a group of learners yet, so it makes no claim about what anyone learns from it.

How it works

  1. Source
  2. Structure
  3. Explanation
  4. Practice

Authored experiments and Learning Toys, including a typed knowledge graph with neighborhoods and search and a constraint-deconstruction experiment. Content is authored and deterministic, with no live AI, uploads, or accounts.

Where it stands today

A deployed lab, released on 5 September 2026 with a public lineage page. Its experiments are authored and deterministic; the broad product thesis has not been validated with a cohort or a single learning outcome.

Working now

  • Knowledge-graph Learning Toy with typed relationships, neighborhoods, and search
  • Constraint-deconstruction experiment
  • Guided explanation surfaces

Not there yet

  • The platform still needs one technical domain, cohort, and measurable outcome.

How it got here

  1. First version: broad learning-platform demos, including a small force graph; its AI parts were scripted
  2. CertGraph's first draft began among its drafts
  3. Re-founded as Learning Systems Lab around a shared constraint spine, borrowing CertGraph's traceable evaluator design
  4. Production release, with a public lineage page

What comes next

Next step

Document the laboratory role around CertGraph.

Later ideas, not current features

  1. Use the lab for reusable research and visualization components
  2. Run narrow experiments through CertGraph

What I checked, and what stays bounded

Dated evidence

  1. Legibility and product-separation review completedreview

Boundaries it keeps

  • Broad feature coverage is not learning efficacy
  • CertGraph remains a separate focused product
  • The 2025 version's AI features were scripted, not live

My role

Directed the learning interfaces and knowledge-visualization experiments, and decided the separation that produced CertGraph.