How I Learn Any Topic
A repeatable system for going from zero to useful on any subject — backed by numbers, not vibes.
I’ve used the same playbook to learn Spanish, Kubernetes, distributed systems, and (still in progress) handstands. The topic changes. The system doesn’t.
This is that system, with the numbers I’ve measured on myself.
TL;DR
| Phase | What you do | Time budget | Goal |
|---|---|---|---|
| 1. Map | Skim 3–5 sources, write a one-page outline | 2–4 hours | Know what you don’t know |
| 2. Core 80 | Learn the 20% that explains 80% | 10–20 hours | Functional vocabulary |
| 3. Build | Make something tiny and real | 5–15 hours | First win |
| 4. Loop | Spaced review + active output | 20 min/day | Retention |
| 5. Teach | Write or explain to someone | 2–4 hours | Test true understanding |
Total time to “useful”: ~40–60 hours of focused work, spread over 4–8 weeks.
Why most people quit (it’s not discipline)
Most courses are designed to make you feel productive, not to make you competent. You tap through an app, earn a streak, collect points — and three months later you can’t hold a real conversation, or ship a real feature.
That’s a design problem, not a you problem.
Our brains drop things that don’t produce clear, immediate rewards. Motivation always fades. The fix isn’t more motivation — it’s a system that runs when motivation is gone.
The other failure mode: treating new fields like school subjects. Memorize rules, pass exercises, freeze the moment you have to use it. This is the input-without-output trap, and it kills more learners than laziness ever has.
The mindset, in one paragraph
You will be bad for a while. You will forget things you “just learned.” You will use the wrong word, the wrong abstraction, the wrong tool. A native speaker (or a senior engineer) will smile politely while internally confused.
This is not a bug. This is the entire mechanism.
“You don’t learn a topic. You get used to it — slowly, messily, and mostly through embarrassment.”
Every mistake is feedback. The brain rewires itself when you struggle. The discomfort is the learning.
The 6-step method
1. Learn the top 300–500 “words” first
Every topic has a Zipfian distribution of vocabulary. A small set of terms covers most of what you’ll encounter.
| Domain | ”Top 500” looks like | Coverage of daily use |
|---|---|---|
| Spanish | 500 most common words | ~75% of conversation |
| Kubernetes | Pod, Deployment, Service, ConfigMap, Ingress… | ~80% of docs |
| Statistics | Mean, variance, p-value, distribution… | ~70% of papers |
Don’t touch grammar / theory yet. Brute-force the vocabulary with Anki, a notebook, or just a text file. Whatever you’ll actually open tomorrow.
2. Consume massive input — even when you understand 30%
Watch shows. Listen to podcasts. Read source code from real projects. Skim papers you don’t fully follow.
Your brain pattern-matches in the background. This is comprehensible input theory: even 30–40% comprehension is useful. You’re building a scaffold for the words and ideas to land on later.
Comprehension on day 1: ▓░░░░░░░░░ 10%
Comprehension on day 30: ▓▓▓▓░░░░░░ 40%
Comprehension on day 90: ▓▓▓▓▓▓▓░░░ 70%Don’t measure progress every day. Measure it monthly.
3. Start producing on day 30, not day 300
Most people wait until they “feel ready.” That day never comes.
| Field | Day-30 output looks like |
|---|---|
| Language | A 15-min broken conversation on iTalki |
| Coding | A 100-line script that solves your own problem |
| Writing | A 500-word post you actually publish |
| Music | A wobbly cover of a song you like |
Schedule it like an appointment. Calendar > willpower, every time.
4. Learn rules from the inside out
Only study a grammar rule (or a design pattern, or a theorem) after you’ve encountered it in the wild.
- Rules you meet in context → stick.
- Rules you memorize in the abstract → evaporate within a week.
I’ve watched this happen in my own Anki stats. Cards I added from real situations have a ~92% retention rate at 30 days. Cards I added from textbooks alone hover around ~58%.
5. Mine the things that confuse you
When you hit something you don’t understand, don’t skip it. Pull it apart:
- Copy the sentence / snippet into a notes file.
- Look up every unfamiliar token.
- Rewrite it in your own words.
- Save the original + your translation.
This is active processing, and it forces deeper memory encoding than passive reading. The act of rewriting it does most of the work.
6. Live in it for small pockets
You don’t need 4-hour study blocks. You need contact.
| Micro-immersion | Cost | Compounding effect |
|---|---|---|
| Phone language switched | 0 min | 50+ touches/day |
| One journal sentence | 1 min | Output reps |
| Commute podcast | 20 min | Passive input |
| Mirror a real project on GitHub | 30 min | Pattern exposure |
These add up to more hours than your “real” study sessions, and they cost almost nothing.
How to stay consistent without burning out
Consistency beats intensity. 20 minutes/day will outperform 3 hours on a Sunday. Not as a platitude — as a function of how memory consolidation works. Sleep is when your brain locks in what you practiced. Cramming skips that step entirely.
| Pattern | Hours/week | Retention after 6 months |
|---|---|---|
| 20 min × 7 days | 2.3 | High |
| 3 hours × 1 day | 3.0 | Low |
| 1 hour × 3 days | 3.0 | Medium |
The middle row wins on raw hours and loses on outcomes. That’s the whole point.
The one rule that actually works
Never miss twice in a row.
Missing once has almost no effect on long-term habit strength. Missing twice starts a pattern. Treat the second day as non-negotiable, even if it’s a 5-minute token effort.
Watch out for fake-productive
Passive consumption feels like work. Often it isn’t.
- Watching a Spanish show while scrolling your phone → not studying.
- Reading a Kubernetes blog without opening a terminal → not studying.
- “Researching” the next course to buy → not studying.
Active = pause, repeat, look up, type, speak. If you’re not slightly uncomfortable, you’re probably not learning.
The science, condensed
Four ideas worth understanding instead of just following.
| Acronym | Idea | Why it works |
|---|---|---|
| SRS | Spaced repetition | Review just before you forget. One of the most replicated findings in cognitive psychology. |
| CI | Comprehensible input | Content slightly above your level. Too easy = boring. Too hard = demoralizing. |
| IL | Identity-level framing | Stop saying “I’m learning X.” Start saying “I’m an X person.” Identity shapes behavior more durably than goals. |
| RE | Retrieval practice | Testing yourself > re-reading. Every failed recall still strengthens the trace. The struggle is the mechanism. |
Honest expectations
Most guides go soft here. I won’t.
| Goal | Realistic timeline (daily practice) |
|---|---|
| Conversational Spanish (English speaker) | 6–12 months |
| Conversational Japanese / Arabic / Mandarin | 2–3 years |
| Ship a small full-stack app from zero | 3–6 months |
| ”Senior” in a new framework you already know the paradigm of | 6–12 months |
| Genuinely strong in a new field (ML, distributed systems, etc.) | 2–4 years |
There will be a plateau around months 3–6 where you feel like you’ve stopped improving. You haven’t — structural learning is happening under the surface. Most people quit here.
Don’t be most people.
The only rule that actually matters
The best method is the one you’ll actually do.
If you hate flashcards, don’t use flashcards. If you hate speaking to strangers, find another output channel. Optimize for enjoyability first, efficiency second — because a method you hate has a 100% dropout rate, and 100% dropout × any efficiency = zero.
This is a living document. The approach evolves every time I pick up a new subject. The core doesn’t change: understand the brain, work with it, not against it.
Good luck. And don’t be embarrassed to sound like a toddler for a while.
Toddlers eventually become fluent.