2025 YouTube Video Roadmap & Accountability Dashboard

North Star Goal 🚀 — Publish 50 videos in 2025 (⏱ 1 video / week). 10 are already live → 40 to go.

Why it matters: Teaching sharpens my own understanding, helps busy engineers, and builds a public body of work that compounds over time.


⏳ Live Progress (checked every Sunday evening, updated in the table)

Updated every Sunday evening

Metric Target Current Delta
Videos published 50 15 –35
Week streak n/a 7  

🎯 Outcome KPIs

Priority KPI Verification Method
P1 1 video / week YouTube Studio publish date
P2 40 new videos by 31-Dec-2025 Roadmap table shows 50 total rows marked Published

🔄 Weekly Workflow

  1. Sunday PM (30 min)
    • Update progress tables
    • Fill metrics
    • Choose next “Focus Episode” (see algorithm below)
  2. Mon-Sun
    • Script ✍️
    • Slides 🎨
    • Record 🎥
    • Edit 🖥️
    • Publish 🚀

Note: Skipped steps require a public blameless post-mortem in the next Sunday entry.


🧭 Focus-Episode Selection Algorithm

When in doubt, follow this deterministic queue to prevent decision-fatigue from sabotaging momentum.

IF any item in **WIP** has Status = "drafting" OR "editing" → pick that.
ELSE pull the top card from **Backlog** (highest Priority number).

🗂 Kanban

📺 Published Videos

# Title Pub Date Link Category
1 2017 MoE 2025-06-01 ▶︎ ML / RecSys
2 Divide and Conquer — Leetcode, Productivity, Fitness, Power 2025-05-25 ▶︎ Thoughts
3 Understanding Performer 2025-05-18 ▶︎ ML / RecSys
4 Self-Communication 2025-05-11 ▶︎ Thoughts
4 Understanding 1991 MOE 2025-05-04 ▶︎ ML / RecSys
5 Transformer and Its Use in RecSys 2025-04-27 ▶︎ ML / RecSys
6 Understanding Linformer 2025-04-20 ▶︎ ML / RecSys
7 Let’s Chat about Dune 2025-03-09 ▶︎ Book Review
8 Attention Part 1 2025-03-09 ▶︎ ML / RecSys
9 PyTorch Dataset Memory Duplicate 2025-02-16 ▶︎ ML / RecSys
10 Understanding Low GPU Usage 2025-02-09 ▶︎ ML / RecSys
11 LLM Gestalt: From ChatGPT to Silicon 2025-02-02 ▶︎ ML / RecSys
12 PyTorch Part 1 – Bare-Min Text Class 2025-01-20 ▶︎ ML / RecSys
13 Creating GPU VM in AWS 2025-01-05 ▶︎ ML / RecSys
14 Debugging my 2024 Self 2025-01-25 ▶︎ Reflection

🛠 Work in Progress (1-2 max)

Title Status Target Pub Date Next Micro-Task
Understanding git scripting 2025-06-08 Research git internals
K8s basics scripting 2025-06-15 Research k8s internals

📥 Backlog and ideas

Rank Title Category Notes
1 Automating every-day decisions Thoughts Decision fatigue and automation
2 What can you do with github? Dev Tools Beyond code hosting
3 Understanding Huggingface/Github business model Business Platform economics
4 Terminal tools Dev Tools CLI productivity boosters
5 Keystrokes that make you throw your mouse away Dev Tools Keyboard-driven workflows
6 On scaling Thoughts Systems and personal growth
7 Parallel: Human As Machine Thoughts Optimization mindset
8 How to measure and improve experiment throughput? ML / RecSys Research velocity metrics
9 Reading Design-Data-Intensive Applications Book Review System design fundamentals
10 DIN ML / RecSys Launch mini-series Attention in RecSys
11 DCN v2 ML / RecSys Contrast with Wide & Deep
12 SASRec ML / RecSys Bridge to Sequential models
13 Ideas about Scaling Thoughts Babysit until they can fly themselves
14 Me as a machine Thoughts Helpful to think of myself as a machine
15 Map Reduce/ Parameter Server ML / RecSys Distributed training systems

Add new ideas at the bottom; reorder every quarter.


🏆 Reward System

Milestone Trigger Reward Ideas
Micro Publish any video that week • Treat at a new café or boba run 🔋
• 30-min skate-park session 🛹
• Digital comic or Kindle short story 📚
Streak 4 videos in 4 weeks • ≤ $150 production upgrade 🎙️
• New vintage Nike/Adidas piece 👕
• Board-game night with friends 🎲
Quarterly 12 videos in a quarter • Weekend micro-trip 🚗
• Speed-cubing timer + premium cube ⏱️
• 1-hr coaching call with YouTuber 🧑‍🏫
Year-End Hit North-Star Goal (50/50) • 3-day ski/surf getaway with vlog 📹🏄‍♀️
• 1 TB iPhone upgrade 📱
• $500 STEM education donation 🤝

Consequences for missed week: Publish a short “lessons learned” video or community post outlining the slip & fix. No dining treat that week.


🔑 Knowledge Pillars ↔ Tracks

(unchanged, serves as syllabus reference)

Track A · Transformer Basics

  1. Attention Is All You Need (done)
  2. Linformer (done)
  3. Performer (WIP)

Track B · Mixture-of-Experts

  1. 1991 Adaptive MoE
  2. 2017 MoE
  3. Switch Transformer

Track C · Classic RecSys Models

  1. Wide & Deep
  2. DCN-v2
  3. DLRM + Two-Tower

Track D · Attention in RecSys

  1. DIN
  2. DIEN
  3. DSIN
  4. BST
  5. SIM
  6. ETA
  7. TWIN

Track E · Sequential RecSys

  1. SASRec
  2. HSTU

Track F · Retrieval Tricks

  1. Locality-Sensitive Hashing
  2. Vector Quantization

Track G · Generative Retrieval

  1. Google TIGER
  2. Meta HSTU

📝 Changelog

  • 2025-05-31 — Updated 2017 MoE video.
  • 2025-05-25 — Updated divide and conquer and understanding performer videos.
  • 2025-05-11 — Updated self-communication video.
  • 2025-05-04 — Updated understainding 1991 MoE video.
  • 2025-05-03 — Started tracking week streak.