--- title: "AI Chrome extension for Coursera & edX: explain lecture transcripts and readings in place." description: "Highlight lecture transcripts, course readings, and practice quiz concepts on Coursera, edX, or MIT OCW and get instant Socratic explanations in place." canonical_url: "https://vezz.me/for/coursera-and-edx/" type: "pseo-for" category: "for" topic: "Platform Guide — Coursera, edX & MIT OCW" intent: "Best for online learners taking university courses on Coursera, edX, MIT OpenCourseWare, and Khan Academy." --- # AI Chrome extension for Coursera & edX: explain lecture transcripts and readings in place. Highlight lecture transcripts, course readings, and practice quiz concepts on Coursera, edX, or MIT OCW and get instant Socratic explanations in place. > **Best fit:** Best for online learners taking university courses on Coursera, edX, MIT OpenCourseWare, and Khan Academy. ## Live Webpage Walkthrough - **Source:** Coursera Machine Learning Specialization — Interactive Lecture Transcript (`coursera.org/learn/machine-learning/lecture`) - **Recommended Context Mode:** Basic **Highlighted Passage:** > When we apply L2 regularization (ridge regression), we add lambda over 2m times the sum of squared weights w_j^2 to the cost function. Notice that by convention we do not regularize the bias parameter b, because regularizing b would penalize the vertical intercept of our model. **User Question (Shortcut):** "Why does squaring w_j penalize large weights differently than L1 (absolute value), and why does L1 produce exact zeros?" **Vezz me On-Screen Answer:** 1. **L2 ($w_j^2$) Proportional Decay**: The gradient pull toward zero of $w_j^2$ is **$2w_j$**. When $w_j$ is large ($10$), the pull is strong ($20$); as $w_j$ gets tiny ($0.01$), the pull shrinks to near-zero ($0.02$), so weights approach zero asymptotically without hitting exact `0`. 2. **L1 ($|w_j|$) Exact Zeros (Sparsity)**: The derivative of $|w_j|$ is a **constant $\pm 1$** even when $w_j = 0.001$—so it pushes small, noisy weights all the way to **exact `0`**. **Suggested Follow-Ups:** - "Turn this L1 vs L2 gradient difference into 2 Anki flashcards." - "What does Elastic Net regularization do?" ## Step-by-step workflow 1. **Highlight the interactive transcript sentence below the lecture video**: Coursera and edX render full, clickable text transcripts directly underneath every video player. Highlight the 2–4 sentences where the instructor lost you. 2. **Use Feynman Coach or Math Notation Translator**: Press your shortcut and run the Feynman Technique Coach preset for conceptual intuition or the Math Notation Translator preset for equations. 3. **Convert module readings into Anki cards**: Before moving to the next module, highlight key definitions and generate atomic flashcards. ## Why highlighting the transcript beats 'whole-video summarizers' You don't need a generic 3-bullet summary of a 15-minute university lecture—you need an explanation of the exact 30-second derivation at `08:14` where the professor skipped two steps. - Focuses 100% of the explanation on the exact transcript sentences you highlight - Never injects overlays into `