A day takes two hours.
Here's where they go.

Every one of the 90 days has the same shape. Once you've done one, the only thing left to decide is whether you understood it.

Start Day 1 See all 90 days

The shape of a day

01
Where you are
One sentence connecting yesterday to today. You never start a day wondering why it follows the last one.
1 min
02
Today's outcome
Written as something testable. Not "learn databases", but "explain how an index turns a 4-second query into 4 milliseconds".
1 min
03
Master Prompt
Paste it into Claude or ChatGPT. It quizzes you first, then teaches to the exact scope of the day.
60 to 90 min
04
Resources
Three to six hand-picked videos, docs and articles covering the same ground, so you can watch it on YouTube instead whenever you prefer.
optional
05
Exercise
One small implementation. Enough to prove the concept moved from reading into doing.
30–45 min
06
Recall
Three to five questions you answer before seeing any answer. Retrieval, not recognition.
10 min
07
Mastery check
Rate yourself against five named criteria. The state is derived from what you rated, so there is no button marked "mastered".
5 min
08
Extra time
Optional depth, clearly separated. Nothing a later day depends on ever lives here.
optional

The Master Prompt

We don't write the lesson.
We write the prompt.

Anyone can ask an AI to explain indexes. You'll get a generic answer pitched at nobody. Every day here ships a prompt tuned to exactly where you are on the path: what you already know, what to leave out, and an instruction to test you before it tells you anything.

It stays current because the model does. It can also be fluent and wrong, which is why every day ships checked resources underneath it. Those are not the fallback for people who prefer video. They are what you hold the answer up against on a day you could not yet catch it.

▶Indexing explained visuallyYouTube · 14 min
§PostgreSQL: IndexesOfficial docs · 20 min
¶Use The Index, LukeArticle · 25 min

Day 43 · Indexes

You are teaching me for exactly one session. I am on Day 43 of a 90-day
engineering path. I know Python, SQL basics (SELECT / WHERE / JOIN), and I have
built small APIs. I have never understood database indexes.

Teach me B-tree indexes in about 40 minutes of reading.

Rules:
- Ask me 2 questions FIRST to find what I already believe. Wait for my answers.
- Then explain in this order: what an index is, what problem made people invent
  one, how a B-tree actually stores keys, where it lives on disk, and what it
  costs on write.
- Use one concrete table (users, 5 million rows) throughout. No toy examples.
- Show me EXPLAIN output before and after. Tell me how to read it.
- Do NOT cover query planners, partial indexes, or covering indexes today.
- End by asking me 3 questions I should be able to answer. Do not answer them.

Open in ClaudeOpen in ChatGPT

Finishing isn't the same as knowing.

Completion and mastery are tracked as two separate numbers, because they are two separate things. You can finish forty days and have mastered twelve days' worth of concepts, and you should be able to see that.

Mastery is derived from what you rated, never chosen. All five, or it isn't mastered.

01ExplainSay what it is, in plain language, without notes.
02RecognizeSpot it inside unfamiliar code you've never seen.
03ApplyUse it correctly in something you're building.
04DebugDiagnose it when it fails in the usual ways.
05TradeoffsArgue when not to use it, and what you'd use instead.

Weak concepts come back on their own.

Anything you rated shakily returns at widening intervals. Get it right and it moves further away; get it wrong and it steps back one rung, not all the way to the start.

The queue is capped at ten a day. Left uncapped it would reach about thirty by day forty-five. That is a quarter of your time, and a debt counter people quit over.

day

days

Recalled → next rung · Missed → one rung back · Cap 10/day

Everything here is taught from the source.

Not somebody's summary of the docs. The people who built the thing, writing about the thing, opened and timed by hand. Every organisation below is cited by name inside the ninety days.

The 338 sources come from