ALGORITHMS

but like... actually explained good

no cap, this will make sense

reading time: like 1-2 hours max

vibe check: chill but educational

brain cells required: just a few

snacks recommended: yes

ok so like... what even IS an algorithm

bestie. i know. you've heard 'the algorithm' blamed for literally everything. why did you see that cursed video at 3am? the algorithm. why does your ex keep showing up on your feed? the algorithm. why is your fyp either elite or absolute garbage? THE ALGORITHM.

but here's the tea: most people have zero clue what an algorithm actually is. they just use it as a scapegoat like 'mercury retrograde' but for tech.

so let's fix that real quick.

an algorithm is literally just... instructions. that's it. it's a recipe. step 1, step 2, step 3, done.

when you make ramen (the good kind with the egg), you're running an algorithm: boil water → add noodles → wait 3 min → add seasoning → crack egg → pretend you're a chef → eat.

computers do the same thing but faster and without the existential crisis.

fun fact: the word 'algorithm' comes from some persian guy named al-Khwarizmi who was doing math in like 830 AD. every time tiktok recommends you a video, you're invoking the spirit of a thousand-year-old scholar. he would be so confused rn.

why does algorithm speed even matter tho

the 'but computers are fast' cope

ok so you might think: computers are literally so fast now, who cares about making algorithms faster? just throw more computer at it lol

this is like saying 'my car is fast so it doesn't matter if i drive to LA via alaska'

the ROUTE matters. a lot.

let me explain with the most boring example that will somehow blow your mind:

the book problem (stay with me)

imagine you have a 1000-page book and need to find page 742.

the chronically offline way: start at page 1. is this 742? no. page 2? no. page 3? ...you see where this is going. worst case you check all 1000 pages. this is called O(n) which basically means 'more pages = more suffering proportionally.'

the actually smart way: open the middle (page 500). is 742 bigger or smaller? bigger. so yeet the first half. open middle of what's left (page 750). 742 smaller? yeet the second half. keep going. you'll find it in like 10 steps.

TEN steps vs potentially ONE THOUSAND.

but here's where it gets unhinged: with the smart way, if you DOUBLE the book size, you only add ONE more step. a billion page book? about 30 steps. that's it.

this is O(log n) and it's literally cracked.

why this matters irl

google doesn't search through 1000 pages. they search TRILLIONS. tiktok doesn't recommend videos to 10 people. it's 1 BILLION users. every day.

at this scale, a bad algorithm isn't 'slower.' it's 'literally impossible even with every computer on earth.'

this is why tech companies pay software engineers like $200k+ starting salary. this is why they ask algorithm questions in interviews. it's not gatekeeping (ok it's a little gatekeeping), it's that this skill determines whether products WORK.

a well-designed algorithm = same performance boost as waiting 10 YEARS for better computers.

so yeah. it matters.

the algorithms running your life rn

tiktok (the final boss)

tiktok's algorithm is genuinely terrifying and i say this with respect.

unlike insta where you mostly see people you follow, tiktok's fyp is 100% algorithm-driven from the MOMENT you download it. it tracks:

• what you watch

• how LONG you watch

• what you rewatch (it knows you watched that video 5 times)

• what you skip

• what you share

• probably where your eyes look on the screen honestly

within like 40 minutes of using tiktok, it has you figured out. people say it knows them better than their friends. that's not a flex, that's concerning.

late 2024 update: tiktok shifted from 'viral trends' to 'search-first discovery.' the old 3-second hook? dead. now it cares if you stay for 15-20 seconds. it's rewarding depth over clickbait. creators who adapted saw 400% more reach. those who didn't? rip.

also they now PENALIZE cross-posted content by 40%. tiktok wants tiktok-native stuff only. the algorithm is getting picky fr.

google (the one that's changing everything)

remember when google just showed you links and you clicked them?

yeah that's dying.

now google has 'AI Overviews' where it just... answers you directly. no clicking needed. by late 2025, these showed up on 30% of searches.

result: 69% of google searches now end with ZERO CLICKS. seven out of ten people get their answer and leave. websites get nothing. publishers are literally going bankrupt.

business insider lost 55% of their traffic. huffpost lost half. this is not a drill.

google basically said 'what if we just... kept all the value ourselves' and did it.

meta/facebook/instagram (the rage machine)

facebook discovered something dark: angry content gets more engagement. fear makes people share. outrage makes people comment.

the algorithm doesn't have morals. it just optimizes for whatever metric the engineers measure. if that metric is 'time on app,' congrats, you've built a rage machine.

late 2024: meta introduced 'Andromeda' - a new AI that now controls ad targeting INSTEAD of advertisers. the machines are literally taking over and we're just vibing.

also 40%+ of your facebook feed is now content from people you DON'T follow. it's becoming tiktok. everything is becoming tiktok.

the AI takeover (it's happening bestie)

2024-2025: when everything got weird

ok so the big shift:

OLD way: you search → algorithm ranks existing content → you click a link → you read someone's article

NEW way: you search → AI generates an answer from multiple sources → you get what you need → original creators get... nothing

the internet is being summarized out of existence.

even wikipedia - the MOST cited source in AI answers - saw 8% fewer human visitors. they're being scraped for info, cited in AI summaries, but people don't actually visit anymore.

the content paradox (this is lowkey scary)

AI is trained on human-created content.

AI then creates new content that gets published online.

future AI gets trained on THAT content (which includes AI stuff).

you see the problem? the snake is eating its tail.

by late 2025, ~17% of top google results were AI-generated. some researchers worry about 'model collapse' - where AI training on AI content makes everything gradually worse.

it's like infinite telephone but with robots and also it determines what information humans have access to. no big deal.

how to actually learn this stuff

why should you even care

valid question. three answers:

1. money. entry level at google = $180k+. the interview is almost entirely algorithms. learn it → pass interview → secure the bag.

2. understanding. algorithms control your reality now. not understanding them is like not understanding money. technically optional, practically a disaster.

3. it's actually kinda beautiful? i know i know. but once you get WHY binary search is so elegant, there's genuine aesthetic pleasure. it's like when a beat drops perfectly but for your brain.

the speedrun learning path

STEP 1: CS50 on youtube. free. harvard course. specifically lecture 3 on algorithms. david malan is an elite teacher. watch it like youtube but actually learn something.

STEP 2: pick ONE course and actually finish it:

• stanford algorithms (coursera): very theoretical, math heavy, makes you feel smart. ~160 hours. prof tim roughgarden is goated.

• princeton algorithms (coursera): more practical, java-based, FREE to audit. prof sedgewick literally studied under the 'father of algorithms.' programming assignments are weirdly fun.

• zero to mastery - master the coding interview: most interview-focused. teaches you HOW to think about problems. subscription but worth it. use code FRIENDS10 for 10% off.

STEP 3: read cracking the coding interview. by gayle laakmann mcdowell. she was literally on google's hiring committee. 189 real interview questions. this book is the bible. ~$35 on amazon.

STEP 4: leetcode grind. just do problems. every day. even one a day adds up. you don't need premium. just consistency.

the quote that hits different

linus torvalds (created linux, kind of a big deal) said:

'Bad programmers worry about the code. Good programmers worry about data structures and their relationships.'

basically: the STRUCTURE you choose matters more than how clever your code is. pick the right algorithm and the problem solves itself. pick the wrong one and you're cooked no matter what.

this changed how i think fr.

what's coming next (it's giving dystopia)

2030 preview

• traditional SEO is dying. if 70%+ searches have zero clicks, why optimize to rank on a list nobody sees? new game: get cited by AI.

• AI agents are coming. rn you ASK chatgpt questions. soon you'll TELL it to DO things. book flights. manage calendar. negotiate prices. scary but convenient.

• the filter bubble gets weirder. as AI gets better at predicting what you want, you might stop seeing ANYTHING that challenges your views. ever. good for engagement, bad for society.

• regulations maybe. EU AI Act launched 2025. more coming. the 'move fast and break things' era might be ending. maybe.

2050 preview (pure speculation but fun)

• AGI (AI as smart as humans at everything): many experts think mid-century. some say sooner. what happens then? literally nobody knows.

• human-AI merger: ray kurzweil (good prediction track record) thinks we'll start merging with AI via brain chips. neuralink is already doing trials. the future is closer than you think.

• the singularity: when AI can improve itself recursively, leading to intelligence explosion. either best thing ever or we're all cooked. kurzweil predicted 2045.

algorithms that started as math homework are becoming the operating system of civilization. understanding them isn't optional anymore.

tldr; you now know more than 99% of people

congrats bestie. you now understand:

✓ what algorithms actually are (instructions/recipes but for computers)

✓ why speed matters (billion-item searches, O(n) vs O(log n), etc)

✓ how tiktok/google/meta control what you see (it's not random)

✓ what changed in 2024-2025 (AI generating answers, zero-click apocalypse, content paradox)

✓ how to actually learn this (CS50 → coursera → ctci → leetcode → bag secured)

✓ what's coming (more AI, less human control, maybe regulations, possibly singularity)

you're now equipped to:

• actually understand tech news

• know why your fyp is weirdly accurate

• potentially get a job that pays absurd money

• sound smart in conversations

not bad for like an hour of reading.

the algorithms have awakened. they're shaping your reality whether you understand them or not. the only question is whether you'll understand them while they do.

now go outside. touch grass. then come back and practice leetcode.

— that's it that's the masterclass —

go forth and be cracked at algorithms

resources (the useful links)

free stuff

• CS50 on YouTube - search 'CS50 2024' - lecture 3 specifically

• Princeton Algorithms I & II on Coursera (free to audit)

• Stanford Algorithms Specialization on Coursera (free to audit)

• LeetCode (don't pay for premium, free is fine)

• algs4.cs.princeton.edu (free textbook stuff)

paid stuff that's worth it

• Zero to Mastery - Master the Coding Interview (use FRIENDS10 for 10% off)

• Cracking the Coding Interview book - ~$35 on amazon

• Algorithms 4th Edition by Sedgewick - the textbook if you're serious

the 6-month speedrun

week 1-2: CS50 lecture 3 + youtube videos on Big O

week 3-8: pick ONE course and finish it (don't start three and finish none)

week 9-12: cracking the coding interview, do the problems

week 13+: leetcode daily. even one problem. consistency > intensity

month 4-6: apply to jobs. get rejected. learn. apply more. get hired.

that's literally it. no secret sauce. just actually do the work.

now close this and go learn something

your future self will thank you

(or at least be able to afford better snacks)