retrieval practice for engineers

Know your systems cold.

You read a codebase once and lose it. Cold Recall turns the systems you work in into memory that doesn't decay: rebuild them from scratch, graded against reality, spaced so they stick.

seed code · rebuild from memory · graded, not guessed

TRACE1 of 16 due

Reconstruct what happens when a payment webhook arrives, from the request to the order being marked paid.

the flow — first to last
webhook receivedverify signaturecreateOrder()mark paid
retry queuesend receiptrefund path
lowmedhigh
Check

the problem

Reading is not remembering.

You've read the auth flow, the webhook handler, the payment path. Could you rebuild them right now, from memory, with the file closed? Understanding you can't reconstruct is understanding you don't have. It fades the moment you switch tabs.

how it works

Three steps. No flashcards.

Cold Recall isn't trivia about your code. It makes you rebuild how the system behaves, then brings it back before you forget.

01

Seed a system

Paste a codebase, a service, or a topic. Cold Recall reads it and authors challenges that force reconstruction — never “what does line 12 do.”

02

Rebuild from memory

Trace the flow. Fix the bug. Predict the failure. Explain the why. No hints, no code in front of you. A cold start, on purpose.

03

Spaced so it sticks

An expanding ladder resurfaces each system at 1, 3, 7, 14, 30, 60 days — right before you'd forget. Knowledge that lasts, not a cram.

how you rebuild

Four ways to prove you know it.

Every challenge is reconstructive. You rebuild behavior, not recall words on a screen.

Trace

Rebuild the flow

Reconstruct a system's data and control flow as a graph, from memory — order and causality, not a vague sketch.

Fix

Find the bug

Diagnose and repair a realistic bug injected into real code, from the symptom alone. No “// BUG here” labels.

Predict

Call the failure

A fault gets toggled. Predict exactly what breaks — throws, hangs, wrong value — before it runs.

Explain

Articulate the why

Explain the load-bearing decisions and what breaks without them, graded point-by-point against a rubric.

not vibes

Graded against reality, not a model's opinion.

Fixes run against real tests. Predictions are checked by deterministic execution. Your trace is scored against what the code actually does. Cold Recall knows the difference between “I think I get it” and “I got it.”

cold-recall rate

Sharper than last week?

Track how often you rebuild a system with zero hints — and whether that rate is climbing.

confidently wrong

Find your blind spots

The most dangerous gaps are the ones you're sure about. Cold Recall surfaces where confidence and correctness diverge.

the ladder

Nothing decays silently

Every system has a next-review date. What's slipping comes back up before it's gone.

the method

Two ideas that actually work.

Cold Recall isn't a new trick. It's the two most-replicated findings in learning science, pointed at software.

active reconstruction

Retrieval builds memory

You don't reread — you regenerate. The effort of pulling a system back from memory is exactly what makes it durable. Rereading feels productive and isn't.

spaced repetition

Reviews land before you forget

Each system returns at a widening interval, timed to the edge of forgetting. You keep more, for less time, and it compounds.

Know it cold.

Stop rereading the same systems. Start knowing them.

Free to start · your code stays yours