Practice run

A run is a set of scenarios plus a length. Pick a preset, pick what it draws from, and answer. Afterwards you get percent-correct per domain and per task statement, and the name of the distractor family behind every wrong option you chose.

No scaled score, ever. The 720 cut score on the 100–1000 scale is fixed by a standard-setting study against the real item pool. A percentage on original practice items cannot be converted into it, so this product will not pretend otherwise. What a run tells you is which task statements you get wrong and which wrong-answer habits you have — which is the part you can act on.

1 · Preset
2 · Scenarios

Pick the scenarios this run should draw from. Counts are what the bank can currently supply.

3 · Start
Scenarios
—
Items
—
Clock
—
Blueprint spanned
—

Start run

A seed fixes which items you get and the order of their options, and every run shows its seed on the results page. It does not fix the scenario draw: a full simulation always draws its scenarios fresh, so no seed can be used to pick or share a known set. Leave it blank unless you are deliberately re-sitting the same items.

Why the simulation draws its scenarios instead of letting you pick

A real sitting presents 4 of the 6 published scenarios, so there are 15 possible draws. 14 of those 15 exercise all 5 domains — the exceptions miss D4, which is reachable only through S5 and S6. You cannot know which draw you get, so the only rational preparation covers every scenario. That is why this preset exposes a randomize flag and no scenario picker.

Draws reaching each domain, out of 15
Domain Blueprint weight Reachable via Draws reaching it
D1 Agentic Architecture & Orchestration 27% S1 S3 S4 15 / 15
D2 Tool Design & MCP Integration 18% S1 S3 S4 15 / 15
D3 Claude Code Configuration & Workflows 20% S2 S4 S5 15 / 15
D4 Prompt Engineering & Structured Output 20% S5 S6 14 / 15
D5 Context Management & Reliability 15% S1 S2 S3 S6 15 / 15

What the bank can currently supply

120 original items, each written against one of the 30 published task statements. 30 of 30 statements have at least one item; 30 are at the 3-item floor. A run says exactly how short it is rather than quietly serving a thinner paper.

5 domains, 6 published scenarios, 4 presented on a real sitting.

Runs on this device

Attempts are kept in this browser only, in the same shape the database uses, so nothing is lost when they later sync. An unfinished run shows as abandoned — which is itself worth seeing.

No runs recorded in this browser yet.