Case study · Portfolio candidate · learner study not run
Praxeos
A beginner’s place to see the structure of a choice: who acts, what they seek, what they use, and what they give up.
The public path starts with the first lesson, then the Choice Machine. Market Without a Manager stays beside that path as an 8–10 minute demonstration. Notes stay in the browser, and a learner study has not been run.
The first action is the first lesson. Market Without a Manager stays beside it as an 8–10 minute demonstration with five people and five goods.
01 · The learner
The problem
A beginner can meet praxeology as a list of conclusions. The path on the site is one ordinary situation, a response the learner controls, a visible consequence, and a private note of what changed.
Help a beginner practice praxeology through purposeful action: a person chooses means under constraints and gives something up. The first action is the first lesson. The Choice Machine is the next step. The market lab remains easy to find.
The product record does not include a private motive story. What it does record is the learner problem and the decisions below.
02 · The path
What I built
The product’s own account says I conceived and directed Praxeos: the progression from individual choice to exchange, entrepreneurial discovery, and money across time; the market lab as a demonstration rather than the front door; and the split between checklist checks and optional guidance. Codex assisted implementation, fixtures, documentation, and verification. Claude is the optional in-product guide.
Eleven lessons
Each lesson adds one distinction. Lesson 1 uses a skipped workout: forty minutes of short videos instead of the planned exercise. The learner picks a reading. The distinctions arrive after that answer, and the page says the comparison is not a grade. A sketch asks for the end, the means, the alternative given up, and any inferred motive, kept separate from what was observed.
Practice
Practice cases use ordinary situations, such as a status purchase or a rent cap. A learner can answer in their own words, open a worked example, or continue without answering. None of those controls grades the case.
Four labs
The lab path is a choice, then a market, then discovery, then money across time. Market Without a Manager is the featured demonstration: five people, five goods, and no director. It is not the first action on the homepage.
Sources and a glossary
Sources name a claim, a stable link, and where to look. The glossary says its definitions are starting points, not substitutes for the primary sources. A counterargument sits beside the claim inside the lab.
On lesson 1 I chose a reading of the skipped workout. The distinctions appear afterward, and the page says the comparison is not a grade.
The Choice Machine asks what one choice reveals and what it leaves unseen. It has six stages. The concept name stays hidden until the learner writes an initial interpretation, or asks to see it.
I walked a guided session. The afternoon starts with four hours, $30, and four feasible paths: help at a community event, take a paid shift, study, or rest. Predicting the paid shift does not apply it. The page said the model had not applied that guess. Choosing the shift then produced the observation “more income and less open time,” and described the other futures as visible but unrealized.
The counters beside that observation still showed 4 hours, $30, four feasible paths, and zero constraint changes. I am reporting that as I saw it.
Choosing the paid shift produces the observation “more income and less open time.” The concept stays unnamed until a first interpretation. The counters beside that sentence still showed 4 hours, $30, and zero constraint changes.
The Notebook keeps the in-progress Choice Machine in this browser and offers Resume. There is no account and no score. Clearing site data deletes the notes. Share links are not where the writing lives.
The Choice Machine can be resumed in this browser. There is no account and no score. Clearing site data deletes the notes.
Guidance is optional
Claude can ask one question from a fixed list of cited sources. It does not score a conclusion. I did not call the live guide. The build report linked from the product says ten live Claude cases were skipped because no API key was supplied.
Checks are not a learner study
The build report dated 14 July 2026 records 43 automated tests and 201 browser checks on the four-lab build. I did not rerun that suite. Those checks can show that models replay and that writing stays out of a share link. They cannot show that a beginner understood praxeology.
v1.0 is still waiting
Five real beginners still have to complete every lab without me explaining the interface. That study has not been run. Until it has, this remains a portfolio candidate.