back

0% Completed

Contents
Codebasics Icon
Slide Makeover

Prev Next
Slide Makeover
ART OF TEACHING @ CODEBASICS · CHAPTER 5
Slide Makeover
A real death-by-PowerPoint slide is on the table. Four decisions stand between it and a Codebasics-grade lesson. Operate carefully.
Codebasics INC
How the surgery works
1
Meet the patient — one genuinely terrible (and terribly common) slide.
2
Make 4 makeover decisions. Pick what a Codebasics instructor would do.
3
Watch the slide transform after every correct call — then download your surgery report (PDF).
THE PATIENT
Machine Learning — Introduction
Submitted by a (fictional) new instructor. Take a good look. Feel the pain of the back row.
THE SLIDE, AS SUBMITTED
Machine Learning — Introduction, Types, Algorithms, Applications & Future Scope 🤖
• Machine learning is a subset of artificial intelligence that enables systems to learn from data
• Types: supervised learning, unsupervised learning, semi-supervised, reinforcement learning
• Algorithms: linear regression, logistic regression, decision trees, random forest, SVM, KNN, K-means...
• Our internal churn model achieved 87% accuracy on the holdout set (see appendix 4B)
• Applications: recommendations, fraud detection, medical imaging, self-driving cars, chatbots
• Future scope: AGI, edge ML, AutoML, quantum machine learning, neuromorphic computing
• Prerequisites: statistics, linear algebra, calculus, Python, SQL, cloud fundamentals
• NOTE: please hold all questions until the end of the section!!
🤖🧠⚡
Eight bullets · 7.5px text · three accent colors · an underlined red title · emoji clipart · and a “hold all questions” threat. This slide breaks at least five Codebasics rules at once. Let’s operate.
DECISION 1 OF 4
Slide Makeover
Eight bullets, three colors, a robot emoji. What is the FIRST surgery?
THE PATIENT (as submitted)
Machine Learning — Introduction, Types, Algorithms, Applications & Future Scope 🤖
• Machine learning is a subset of artificial intelligence that enables systems to learn from data
• Types: supervised learning, unsupervised learning, semi-supervised, reinforcement learning
• Algorithms: linear regression, logistic regression, decision trees, random forest, SVM, KNN, K-means...
• Our internal churn model achieved 87% accuracy on the holdout set (see appendix 4B)
• Applications: recommendations, fraud detection, medical imaging, self-driving cars, chatbots
• Future scope: AGI, edge ML, AutoML, quantum machine learning, neuromorphic computing
• Prerequisites: statistics, linear algebra, calculus, Python, SQL, cloud fundamentals
• NOTE: please hold all questions until the end of the section!!
🤖🧠⚡
A.  Make the fonts bigger so the back row can finally read all eight bullets.
B.  Split it: one idea per slide. This slide keeps only the core message.
C.  Add a professional background image behind the text to make it less boring.
Reveal the correct surgery
Principle: One idea per slide.

Bigger fonts on eight bullets is a louder document; a background image is lipstick on a document. The surgery is subtraction: this slide earns ONE job - the core message - and every other bullet gets its own slide or gets cut.
AFTER DECISION 1
Machine learning finds patterns in data — so you don’t have to write the rules.
Codebasics INC
DECISION 2 OF 4
Slide Makeover
The 87% accuracy stat was buried in bullet 4. Where does it live now?
CURRENT STATE
Machine learning finds patterns in data — so you don’t have to write the rules.
Codebasics INC
A.  Bold it inside a bullet list on the next slide.
B.  Put it in the slide footer as a citation.
C.  Give it its own billboard slide: 87% huge, one quiet line under it.
Reveal the correct surgery
Principle: Big fonts — a slide is a billboard.

A number that proves your whole story deserves a slide, not a hiding place. 46-80pt, green, alone. The learner reads it in one second and remembers it next week - that is what slides are FOR.
AFTER DECISION 2
87%
accuracy — our churn model, on data it had never seen
Codebasics INC
DECISION 3 OF 4
Slide Makeover
The robot emoji collage is gone. What imagery replaces it?
CURRENT STATE
87%
accuracy — our churn model, on data it had never seen
Codebasics INC
A.  One real photo from daily life, with a single caption connecting it to ML.
B.  A tasteful set of flat icons: robot, brain, gears, lightning.
C.  No imagery at all - imagery is decoration and decoration distracts.
Reveal the correct surgery
Principle: Photos over clipart — one image, one feeling.

Icon sets are clipart in a suit. A real photo of a real moment (your commute, your kitchen) gives the concept a body the learner already owns. And 'no imagery ever' throws away the fastest highway to memory - the rule is fewer, realer images, not zero.
AFTER DECISION 3
Where you already met ML today
Your morning commute: maps reroute, music queues, fraud checks — all ML.
Codebasics INC
DECISION 4 OF 4
Slide Makeover
You still must teach the 4 types of ML. How do they appear?
CURRENT STATE
Where you already met ML today
Your morning commute: maps reroute, music queues, fraud checks — all ML.
Codebasics INC
A.  A clean 2x2 grid - all four visible so learners see the full map at once.
B.  One type per click: the first appears, gets taught, then the next.
C.  Skip the types - taxonomy is boring, jump straight to a demo.
Reveal the correct surgery
Principle: Reveal one thing at a time.

Show all four at once and learners read type 4 while you teach type 1 - the grid steals your own audience. Click-to-appear keeps the whole class on the same idea at the same moment. (And skipping structure entirely isn't simplicity, it's abdication.)
AFTER DECISION 4
The 4 types of machine learning
1 · Supervised — learning with an answer key
types appear one per click as you teach
Codebasics INC
The transformation
BEFORE — 1 slide, 8 ideas, 0 attention
Machine Learning — Introduction, Types, Algorithms, Applications & Future Scope 🤖
• Machine learning is a subset of artificial intelligence that enables systems to learn from data
• Types: supervised learning, unsupervised learning, semi-supervised, reinforcement learning
• Algorithms: linear regression, logistic regression, decision trees, random forest, SVM, KNN, K-means...
• Our internal churn model achieved 87% accuracy on the holdout set (see appendix 4B)
• Applications: recommendations, fraud detection, medical imaging, self-driving cars, chatbots
• Future scope: AGI, edge ML, AutoML, quantum machine learning, neuromorphic computing
• Prerequisites: statistics, linear algebra, calculus, Python, SQL, cloud fundamentals
• NOTE: please hold all questions until the end of the section!!
🤖🧠⚡
AFTER — 4 slides, 1 idea each
Machine learning finds patterns in data — so you don’t have to write the rules.
Codebasics INC
87%
accuracy — our churn model, on data it had never seen
Codebasics INC
Where you already met ML today
Your morning commute: maps reroute, music queues, fraud checks — all ML.
Codebasics INC
The 4 types of machine learning
1 · Supervised — learning with an answer key
types appear one per click as you teach
Codebasics INC
YOUR RESULT
– / 4
Complete the steps above
(If the buttons are not interactive in your LMS view, score yourself using the reveals above.)
Every slide you save, saves a classroom.
Now open the deck for your own baseline video. Find its worst slide. Operate with the same four moves — and be as ruthless with your slides as you just were with this one.
Codebasics INC · Art of Teaching Crash Course · Chapter 5 — Slide Makeover