The adaptive engine

Four quiet systems doing the work
of a patient teacher.

Behind every course is a small loop: figure out what the learner knows, decide what comes next, bring back what's fading, and ask them to use it. Nothing dramatic. Just done well, lesson after lesson.

01
Diagnose

A calibration that doesn't feel like a test.

02
Sequence

A path through the course, just for you.

03
Reinforce

Recall, spaced to where forgetting begins.

04
Apply

A real project, not a multiple choice.

01
Diagnose

Fifteen minutes to figure out
what you already know.

Most courses start at lesson one. We start with a short calibration — a handful of small questions, a tiny build, a couple of recall prompts. It's not graded. We just need a baseline so we don't waste the next ten weeks of your time.

a.
Skill probes, not exams. A few targeted prompts per topic — not 60-question quizzes.
b.
Behavior matters too. How long you pause, where you re-read, what you skip. We treat hesitation as data.
c.
You can disagree. If the engine misreads you, override it. It will recalibrate around your nudge.
calibration · 04/12
Probe · async
What does the following expression evaluate to?
const r = Promise.resolve(2)
  .then(x => x + 1)
  .then(x => x * 2);
console.log(r);
Don't worry — this only tunes your path.
02
Sequence

The course rewrites itself
around what you need.

Once we know roughly where you are, the engine sequences the lessons that follow. It skims familiar territory, branches into prerequisite refreshers when needed, and adds depth on the parts that connect to your goal.

a.
Skip the obvious. If you know hooks, you'll see one recap and move on — not a five-hour module.
b.
Branch into prerequisites. If async is shaky, we route you through a side lesson — and then return.
c.
Goal-aware depth. Building a marketplace? You'll get more on forms and payments. Building a dashboard? Charts and data.
your path · Modern React
01
The mental model
Skipped — you scored 92%
Skipped
02
Hooks, properly
Standard pace
2h 12m
Side lesson — async refresher
Inserted because your async probe was shaky
+ added
03
State that scales
Standard pace
2h 48m
04
Async, suspense and transitions
Expanded — extra depth for your marketplace goal
+ depth
05
Server components and the network
3h 04m
03
Reinforce

Bring concepts back
just before you forget them.

We track every concept you've touched against the curve of forgetting — and resurface it at the moment recall is hardest, but still possible. Each session opens with a tiny recall block. It feels effortless. It does the heavy lifting.

a.
Spaced retrieval. Concepts come back at 1, 3, 7, 17 day intervals — adjusted live based on whether you remembered.
b.
Active, not passive. Not a flashcard you read — a tiny code prompt, a short answer, a build.
c.
Quiet. Two minutes at the start of a session. You won't notice until you realize you remember everything.
today · recall block · 2 min
Recall · 3 of 4
Last touched 4 days ago — what does useReducer give you that useState doesn't?
StrongForecast: 92% recall in 7 days
Forgetting curve · this week
12 concepts tracked
Day 1Day 7Day 14Day 30
04
Apply

Every module ends with something
you can actually show.

Watching is not learning. Each module hands you a small, self-contained build that uses the concepts you just covered. By the end of the course, you have a public capstone with mentor notes attached.

a.
Pick your flavor. Same skills, different domains. Marketplace, dashboard, editor — choose the one that excites you.
b.
Mentor reviewed. A working practitioner reads your code. You get senior-engineer feedback, not auto-graded checkmarks.
c.
A public artifact. Live URL, source repo, and a write-up. The kind of thing that gets a recruiter to read further.
capstone preview / your-marketplace.app

Capstone — Marketplace

Reviewed

A working storefront with cart, search, and stripe-mode checkout. Mentor: Maya Okafor.

ArchitectureStrong — clean module boundaries.
State designStrong — colocated, no premature global.
TestsGood — covered the cart logic.
AccessibilityStretch — keyboard nav in modals.
In the product

The page you'll see most often.

Your dashboard is intentionally calm. One day, three blocks, your mastery line, and the choice to do more or stop. That's it.

Tuesday, May 21
Good morning, Elena.
7-day streakModern React · wk 4
1
Recall — context vs prop drilling
2 min · spaced from yesterday
Recall
2
New — designing reducer shapes
22 min · video + 1 build
New
3
Apply — wire cart logic into your capstone
18 min · open in editor
Apply
Mastery
72 / 100
Strong: hooks, state shape, immutability.
Stretch: context, transitions.
Est. 42 minutes today · your goal is 30. Stop anytime — your path adjusts.
Built on evidence

Old ideas, finally easy to apply.

None of this is new science. It's just rarely been done at scale, because doing it by hand for every learner is impossible. We're the bit in the middle.

PRINCIPLE 01

Retrieval practice

Recalling — not re-reading — is what consolidates memory. Roediger & Karpicke, 2006.

How we use it: A two-minute recall block opens every session, drawn from the concepts your forgetting model says are most at risk.

PRINCIPLE 02

Spaced repetition

Concepts revisited at expanding intervals stick longer. Ebbinghaus, 1885 onward.

How we use it: Each concept gets an individual review schedule — adjusted live by your performance.

PRINCIPLE 03

Desirable difficulty

Learners progress fastest when challenges sit just above current ability. Bjork, 1994.

How we use it: The engine tunes question difficulty live, aiming for a ~75% success rate — hard enough to learn, easy enough to stay.

PRINCIPLE 04

Interleaving

Mixing related topics beats blocking them, for long-term mastery. Rohrer, 2012.

How we use it: Recall blocks mix topics from the last 30 days — not just the lesson you finished.

PRINCIPLE 05

Worked examples → fading

Show the full solution, then progressively remove the scaffolding. Sweller, 1985.

How we use it: Code lessons walk you through, then ask you to fill in shrinking gaps, then write fresh.

PRINCIPLE 06

Transfer through projects

Knowledge sticks when it's used in a realistic, non-trivial context. Mayer, 2002.

How we use it: Every module ends with a project that's structurally similar to industry work — not a sandbox toy.

Questions you might have

What the engine doesn't do.

Does the engine watch me?
It sees what you do inside lessons — answers, pauses, retries. It does not access your camera, mic, screen, or other apps. Behavioral signals are aggregated, never sold.
What if it gets me wrong?
Every recommendation is overridable from the path view. The engine learns from your override, so it gets you right next time.
Is this an AI tutor?
Not in the chatbot sense. Courses are written by humans. The engine sequences them, schedules recall, and adjusts difficulty. We use ML where it earns its keep — not as the headline feature.
Can I just turn the adaptation off?
Yes. “Linear mode” walks the course top-to-bottom, in original order. You'll still get recall, but no resequencing.

Curious how it'd read you?

Take the 15-minute calibration. No account required. We'll show you the path it would build — and you can decide.