← ReplayCS home

Live product tour · about 2:50

The judge path through real execution.

Five focused stops show the shipped product—not a separate pitch deck. Open each lab, perform the listed interaction, then return here and check off what you verified.

Deterministic truth firstEvery trace is produced by a bounded local engine.
AI is optional and labeledGPT-5.6 when configured; deterministic fallback when not.
No demo-only stateLab actions update the same browser-local profile learners use.

Lab links open in a new tab so this checklist stays available. Checking a stop records only your tour position; it does not award XP or claim that a lab was completed. Real lab interactions do that.

Guided verification

Five stops, one learning loop

Target time · 2:50
  1. DSA I · Binary Search

    Predict, diverge, and recover

    90 sec

    One stop shows the core ReplayCS loop and its safe explanation path.

    1. Enter 4 as the midpoint prediction and lock it. Replay the first divergence, then recover with 3.
    2. Switch Python to C, C++, or Java. The selected semantic operation stays aligned.
    3. Choose “Explain this step.” The mentor uses GPT-5.6 when configured and a clearly labeled deterministic fallback otherwise.

    Verify: Look for recovered XP, the same Step 3 across languages, and a source label on the mentor response.

    Open Binary Search ↗
  2. DBMS · SQL Query Pipeline

    Trace rows through SQL clauses

    25 sec

    Move from source relations to a result without pretending a logical order is an optimizer plan.

    1. Compare the logical teaching order with the explicitly illustrative physical plan.
    2. Inspect the before/after relation at GROUP BY, then predict HAVING for the aggregate filter.

    Verify: Look for fixed in-memory HR data, NULL-aware LEFT JOIN behavior, and row-count deltas.

    Open SQL Pipeline ↗
  3. Operating Systems · CPU Scheduling

    Replay a scheduler decision

    25 sec

    See policy, state, and metrics derive from one deterministic workload.

    1. Keep the default Round Robin policy, lock the first-dispatch prediction, and reveal a few clock events.
    2. Scan the ready queue, Gantt timeline, and side-by-side waiting/turnaround comparison.

    Verify: Look for explicit tie-breaking, remaining burst state, and reproducible per-process metrics.

    Open CPU Scheduling ↗
  4. Networks · Packet Journey

    Follow one URL to rendered bytes

    20 sec

    Connect protocol events to topology, layers, addresses, and encapsulation.

    1. Predict the first cold-cache event, then step through DNS, ARP, TCP, TLS, HTTP, and rendering.
    2. Rebuild with a warm response cache and notice which network events disappear.

    Verify: Look for reserved example addresses, explicit simplifications, and changing frame/packet/segment envelopes.

    Open Packet Journey ↗
  5. Learner Profile · Recommendation

    Inspect evidence-based progress

    10 sec

    Close the loop with progress created by the real interactions you just performed.

    1. Inspect XP, prediction accuracy, recovered mistakes, and misconception evidence stored in this browser.
    2. Follow the next recommendation, which is selected from recorded progress rather than a fabricated score.

    Verify: Look for a mastery map, local-data controls, and a recommendation that links back into a real lesson.

    Open Progress ↗

After the final stop

Let the learner evidence choose what comes next.

Prediction, execution, explanation, recovery, and recommendation all lead to actual product routes and browser-local state.

Review learner profile ↗