THE PROBLEM

Engineering education was built for passive watching.

Videos create the illusion of competence. Real engineering intuition comes from breaking, inspecting, and synthesizing systems.

TRADITIONAL LECTURE COURSES

The Passive Learning Trap

  • ✕40+ hours of video watched at 1.5× speed forgotten within 72 hours.
  • ✕Copy-pasting boilerplate code without developing mental models of call stacks or memory.
  • ✕Blank-screen freeze during technical interviews and architectural design reviews.
Result: High drop-off, superficial understanding, zero retention.
THE BLOOM METHOD

The Active Mental Model Loop

  • ✓Interactive Visual Sandboxes: Manipulate graph nodes, registers, and memory pointers in real time.
  • ✓Timestamp-Grounded AI Mentor: Cites precise lecture moments and code lines instead of spoiling answers.
  • ✓Production GitHub Capstones: Real distributed systems, packet sniffers, and database engines.
Result: Server-verified mastery, retention decay tracking, deep intuition.
THE BLOOM METHOD

The 8-Stage Cognitive Mastery Loop

Every concept moves through an engineered sequence of active cognitive stages: from initial discovery to shipping verifiable code.

STAGE 02•LEARN

Hyper-Concise Conceptual Sparks

Timestamped lecture moments mapped to mental models

Curated 8–12 minute video chapters from world-class engineering educators, annotated with timestamped concept definitions, mathematical formulas, and common pitfalls.

✓Grounded in real university lectures & top engineering channels
✓Clickable timestamp chips jump directly to proof moments
✓Checkpoint questions trigger at critical inflection timestamps
LECTURE STREAM • CS20412:40 / 18:20
Chapter 3: BFS Queue InvariantCHECKPOINT IMMINENT

Vertices are enqueued in strict non-decreasing order of distance from source s.

Clickable Timestamp Chips (Auto-Seek):
▶ 02:14 Call Stack Setup▶ 05:40 Queue Enqueue Logic▶ 09:12 Visited Set Check
INTERACTIVE LABS

Don't just read how a system works. Touch it.

Real computational models right inside the browser. Step through algorithms, manipulate state queues, and build mental models that stick forever.

SIMULATION BENCH

Graph Traversal Lab (BFS vs DFS)

ABCDEF
Current
Visited
In Queue
STEP 1 OF 714% COMPLETE

Start BFS: Enqueue root node A

ACTIVE DATA STRUCTURE (FIFO QUEUE):
A
VISITED SET (O(1) LOOKUP):
A
TIME COMPLEXITY: O(V + E)
SPACE COMPLEXITY: O(V)
GUARANTEES: Shortest path in unweighted graph
ADAPTIVE PATHWAYS

A roadmap that evolves with your cognitive velocity.

Your college syllabus parsed by AI, compiled into an acyclic skill graph, adapting dynamically to diagnostic gaps, quiz accuracy, and spaced retention decay.

STAGE 01 • COMPLETED100% MASTERED

Core Foundations: Memory & Pointer Arithmetic

Stack vs Heap allocation, pointer indirection, cache locality, and asymptotic notation.

RETENTION STABILITY: 28 DAYS✓ VERIFIED BY QUIZ
STAGE 02 • IN PROGRESS72% COMPLETE

Data Structures: Trees, Heaps & Graph Traversal

Adjacency lists, BFS/DFS traversal invariants, cycle detection, and Dijkstra's shortest path.

NEXT: BFS INTERACTIVE LABLaunch Traversal Lab→
STAGE 03 • PREREQUISITES REQUIREDLOCKED

Distributed Systems: Consensus & Storage Engines

Raft consensus, write-ahead logging, LSM-trees, and vector similarity indexing.

REQUIRES: Graph Traversal & Network Sockets0 / 4 UNLOCKED
ENGINEERING GAMES

Turn brutal systems concepts into games you play.

Technical challenges where scoring is validated by real database queries, packet routers, and algorithm race timers — not multiple-choice trivia.

FEATURED CHALLENGE • P0 ENGINEWASM SQLITE

SQL Detective

Investigate high-profile security leaks and unauthorized transaction alterations. Write real SQL queries executed directly against real SQLite databases in your browser.

CASE #04: THE GHOST TRANSACTION
SELECT emp.name, t.amount, t.timestamp
FROM transactions t JOIN employees emp ON t.emp_id = emp.id
WHERE t.amount > 100000 AND t.approved = 0;
Validation: Row-set comparison (not regex)STATUS: VERIFIED
DIFFICULTY: ALL LEVELSPlay SQL Detective→
NETWORKING

Packet Runner

Route IP packets through CIDR subnets, ARP tables, and NAT gateways before TTL expires.

IP
OSI LAYERS 2-4Launch →
ALGORITHMS

Algorithm Race

Predict time complexity and watch QuickSort, MergeSort, and TimSort race on adversarial inputs.

O(n)
REAL CPU BENCHMARKSLaunch →
SYSTEMS

Memory Maze

Manage stack vs heap allocations, prevent memory leaks, and manually free dangling pointers.

PTR
C / RUST MEMORY MODELLaunch →
BLOOM MENTOR

Not a generic chatbot. A compiler for your mental models.

Grounded strictly in your current lesson code, video timestamps, and misconception history. Bloom Mentor guides via Socratic hints and never spoils answers.

SOCRATIC DIAGNOSTIC STATE

Cycle Detection in Directed Graphs

Student attempted DFS cycle detection but failed on cyclic graphs with back-edges.

HINT LADDER:Level 2 of 3 (Guiding Question)
IDENTIFIED MISCONCEPTION:Premature Backtracking State Mutation
Rule: Code solution locked until Level 3 hint exhausted
cycle-detector.ts
LINE 14 ERROR DETECTED
12 | function hasCycle(node: number, state: State[]): boolean {
13 | state[node] = State.VISITING; // Gray
14 | // BUG: Missing backtrack check on visited neighbors
15 | for (const neighbor of graph[node]) {
16 | if (state[neighbor] === State.VISITING) return true;
17 | }
18 | state[node] = State.VISITED; // Black
19 | }
BLOOM MENTOR (GROUNDED @ 14:22)

Notice line 14: You correctly mark the current node as VISITING, but what happens when a neighbor is already in State.VISITED? Does your recursive call check whether unvisited neighbors must be explored recursively?

PRODUCTION LABS

Learning becomes real when you build.

Capstone portfolio systems with automated CLI verification test suites and AI code reviews of your public GitHub repositories.

ADVANCED • 25 HRS5 MILESTONES

Distributed Key-Value Store

Implement Raft consensus, write-ahead logging (WAL), and LSM-tree storage engines capable of surviving network partitions.

RaftWALLSM-Trees
Automated CI RubricView Specs →
INTERMEDIATE • 18 HRS4 MILESTONES

Packet Sniffer & Firewall

Capture Ethernet frames using raw sockets, decode TCP/IP headers, and implement stateless BPF packet filtering rules.

Raw SocketsBPF FiltersTCP FSM
PCAP Benchmark SuiteView Specs →
ADVANCED • 30 HRS6 MILESTONES

Vector Database from Scratch

Construct hierarchical navigable small-world (HNSW) graphs and SIMD cosine similarity routines for sub-millisecond retrieval.

HNSWSIMD DotSIFT Dataset
SIFT 1M Accuracy TestView Specs →

Your engineering mastery starts with one concept.

Build true intuition through active mental models, interactive experiments, and real production capstone projects.

Browse Full Curriculum
✓ Open Source Core•✓ Zero Credit Card Required•✓ Real Code Execution