Internal generation console. Dial in the run below — every control has an ⓘ with a full explanation. Sane defaults are set, so you can also just hit Generate.
01 Scope & content
Define what these cards cover and where their facts come from — the broad exam subject, an optional finer category, which California code books to draw from, and how many cards this run should aim to produce. These choices set the card headers and decide which source material is searched.
02 Topics & retrieval
Decide which topics get cards and how their source material is gathered. Paste your own topics or leave it blank to let the app propose them, then tune how many code passages are pulled per topic, how many cards each topic can yield, and how strictly weak source matches are filtered out before Claude writes.
03 Voice & reading
Shape how the cards sound and read — the reading level for your audience, the brand tone, the output language (including Spanish and bilingual for California's trades), and how aggressively each card surfaces exam tips and the specific numbers most likely to be tested.
04 Depth & length
Set how deep and how long each card goes — from a single key rule for beginners to fully unpacked edge cases and exceptions for advanced study. Use the quick length preset for a rough size, or the Exact length slider when cards must hit a specific character count. Claude never pads weak content to reach a target.
05 Card types & mix
Choose the blend of card formats this run produces: standard concept cards (the normal anatomy), True/False breakdowns that simplify dense code sections into quick-check statements, and step-by-step scenario math drills (board feet, concrete volume, load calcs). The three percentages are balanced to 100% automatically.
06 Content boosters
Optional extras that make cards stickier and more exam-ready — real-world jobsite examples, memory hooks and mnemonics, alternate terms the exam may swap in, common-mistake "trap" callouts, and bolding every tested number. There's also a box for anything you want deliberately kept out of this run.
07 Custom Claude prompt
Free-text instructions Claude reads on every card in this run, layered on top of your house style rules. Use it for anything the controls above don't cover — a special focus, a particular framing, a one-off formatting tweak, or a reminder. These take priority over the style presets while the core format and grounding rules still hold.
08 Output & files
Decide what comes out and how it's organized — Word (.docx, the fully formatted main file), Markdown, and/or plain text; one combined document or a separate file per category; and whether related cards link to each other with "See Card #N" references. Files download to your computer when the run finishes.
09 Generate & results
Pick the AI model, review the estimated cost, and launch the batch. Live progress, file downloads, the activity log, and the QA scorecard all appear here as the run works through your topics — and finished cards auto-save so a hiccup never loses your work.
—