01
Knowledge Graph
Sixty-six concepts wired by two kinds of dependency. The structural foundation every other system reads from.
Inside the engine
SQLMaxx runs on four systems composed into one engine. Each does one job. Below: every system named, every system's behavior shown on screen, every number drawn from how the product actually runs.
Four foundational concepts free. No card.
The architecture
01
Sixty-six concepts wired by two kinds of dependency. The structural foundation every other system reads from.
02
A real probability per concept that moves with every answer. Decides when a concept is locked in.
03
Hard practice quietly maintains easy concepts. Compounds your work so review never piles up.
04
Schedules each mastered concept to come back the day before you would forget it.
01 / Knowledge Graph
Every concept declares the ones it builds on. INNER JOIN reads against the rules of FROM and WHERE. Aggregates read SELECT, FROM, and the rules of NULL. The graph holds these relationships explicitly, so the engine never asks you to write SQL on top of something you have not been taught.
Two edge sets run in parallel. Prerequisite edges decide what unlocks next. A second set of edges, used by Stability Growth below, decides what gets implicit credit when you solve harder problems.
02 / Mastery Confidence
Mastery Confidence is a real probability between zero and one, held per concept. A right answer raises it. A wrong answer drops it harder. The number on screen is what the engine actually stores about you.
A concept locks in only when the probability crosses the mastery line: 0.95 for foundations, 0.98 for advanced. Hours studied do not enter the calculation. Evidence is the only input.
03 / Stability Growth
Every JOIN you solve exercises SELECT, FROM, and WHERE underneath. The engine watches that and gives those concepts partial review credit, on a per-edge weight defined in the graph.
The further you get, the less explicit review you owe. Advanced practice keeps the foundations sharp on its own. Your workload compounds in your favor.
04 / Review Stability
Once a concept is locked in, the engine starts tracking how stable that memory is. Each successful review pushes the next one further out. The intervals stretch as the memory hardens.
You will not get to month four and find that week-one concepts have evaporated. The schedule maintains them on the curve each concept actually needs.
How you actually use it
SQLMaxx does not auto-decide your session for you. From the dashboard, you choose Learn or Review. Each mode pulls on the systems above to do its job.
Move forward through the curriculum.
Open-ended. You stop when you want.
Solve new problems on the next-up concept until Mastery Confidence locks it in. The engine picks the concept; you pick when to leave.
Keep mastered concepts sharp.
Either the full overdue queue, or a fixed batch you set (5, 10, or 20).
Work through the queue. Each correct answer pushes the concept further out on its review schedule. You decide when you are done.
The first four concepts run on the same engine the rest of the platform does. No card.
When you get one wrong
Maxx is the only AI in SQLMaxx. Mastery Confidence, Stability Growth, Review Stability, and the Knowledge Graph are deterministic. The engine that decides whether you have learned something runs on math.
Maxx does one thing: explains a wrong answer in three sentences, grounded in the exact query you submitted. Each sentence does a specific job.
Anatomy of one Maxx response
What your query did
You filtered orders by status before joining customers, which excludes customers who have only cancelled orders entirely from the result.
The rule that produced the result
INNER JOIN keeps a row only when the join condition matches on both sides. Rows that fail the WHERE filter never reach the join.
The pattern the problem was looking for
Move the status filter into the JOIN condition with AND, or switch to a LEFT JOIN so customers without matching orders are still returned.
What it isn't
What it actually does
What it isn't
A course you complete.
What it actually does
A concept locks in only when Mastery Confidence crosses the line. There is no Done button.
What it isn't
A question bank you grind.
What it actually does
Learn picks the next problem against your real Mastery Confidence, on the next concept the graph has unlocked. Random grinding is the thing it replaces.
What it isn't
Ten YouTube channels stitched together.
What it actually does
One curriculum, written in dependency order, by someone who actually does analyst work. Six hundred fifteen practice problems, all graded against real PostgreSQL.