We don't write enough code to learn it anymore

AI writes our code now, so we type less — and learn slower. Anki spaced repetition fills the gap.

I can’t remember the last time I centered a div by hand. Coding agents handle them for me. And for most of what I do, that’s fine. But I’m learning Rust now, and I’ve noticed something: the repetition that used to bake a language into my fingers isn’t there anymore. I type less, so I learn slower. Anki is how I’m fixing that.

Typing was my teacher

Coding used to be repetitive by default. Every language, framework, or tool you adopted came with a built-in curriculum of grind.

• I typed .box { display: flex } a dozen times a week
• I wrote the same useState functions across every React component
• I used to drive git mergetool and git bisect before Claude got good enough to do them for me

Agentic tools changed this. AI-assisted engineering now means we spend less time typing and more time reading, reviewing, and directing. The output per developer is up 33.7% ¹, but the physical act of writing code—typing—is close to extinction.

This is a good thing: it frees us to focus on higher-level problems, but it has an unintended side effect: Less typing, less repetition. Less repetition, less learning.

Procedural memory (the “fingers know before brain catches up” phenomenon) is built through repeated execution ². When you offload execution, you stop building it. For the languages and tools you already know, this just means slower recall. For something new you’re trying to learn, it stalls you.

Enter spaced repetition

Spaced repetition systems (SRS) like Anki is infamous in communities of language learners, medical students, law students. ³ It solve a specific optimisation problem: how do we maximise retention, given finite review time?

• An algorithm predicts when you’re about to forget something. ⁴
• It shows you cards you’re predicted to forget that day. ⁵
• First it shows you 1 day later. Then 3 days, then 2 weeks, then 6 weeks and so on.

I’ve used Anki for 1 year now. It carried me through learning my third spoken language. It can work for programming too.

Replacing doomscrolling with Anki

My workflow goes like this:

1. Read Rust code and work on Rust projects. The cards come from real friction: something I had to look up, something that tripped me up in a PR review or an AI plan review.
2. Identify concepts that confuses me. Not everything needs a card.
3. Write cards. They can be hand-written, or generated with AI, or based off a pre-made deck. It looks like this:

This is something I can do on the phone with the Anki mobile app. I can review it any time I have 3 minutes of downtime: on the train ride to work, in the waiting line, anywhere I would otherwise be tempted to doomscroll.

Key insights I learned

Make your own cards. Writing cards is better than downloading decks. While pre-made decks exist for programming, the act of making a card is itself the first step of learning.

• I focus on gaps I actually have. I don’t need cards for “what is mutable vs immutable”: I know that from other languages. I need cards for the stuff I keep looking up.

• Making cards is the first review. Deciding what to test, how to phrase the question, what counts as the right answer: that’s where the encoding starts.

• Published decks are great for inspiration, however. I steal card formats I hadn’t thought of, adapt them, remix them for my gaps. I just don’t import and call it done.

Don’t confuse the deck with the doing. Anki is only one tool in the learning process. Learning can’t happen with Anki alone. We still need to learn from other resources, and more importantly, learn by doing.

Why this matters more now

Our role has shifted from writing code to reading it. Robert Martin estimated a 10:1 reading-to-writing ratio back in 2008 ⁶. In the agentic era, that ratio has likely widened to 100:1 or more.

Developers are spending less time writing code. A 2026 longitudinal study found 82% of developers reported a decline ⁷, and a survey of 2,847 developers found reviewing AI-generated code (11.4 hrs/week) now outpaces writing new code (9.8 hrs/week) ⁸.

Less writing means fewer opportunities to build procedural fluency. Spaced repetition replaces that lost exercise with a deliberate system.

Prior art

Spaced repetition for technical learning has been explored by some smart minds in recent history:

• Using Anki for mathematics and computer science (Michael Nielsen, “Augmenting Long-term Memory”): A deep guide to using spaced repetition for understanding complex technical material: research papers, command line tools, and programming concepts, not just memorising facts. ⁹

• Memory systems for technical concepts (Andy Matuschak, “Why books don’t work”): Transformative cards test understanding, not recall. ¹⁰

• SRS works for any knowledge domain (Derek Sivers): Spaced repetition for language learning, advocated for programming. ¹¹

• Coding knowledge as a durable mental model (Janki Method, Jack Kinsella): “I had the sinking feeling that I’d been lying to myself about what I actually knew. The Anki cards exposed the gaps.” ¹²

• 9 years of Anki for programming (Semicolon & Sons, “How to Learn to Code I: Use SRS and Anki”): A practitioner retrospective on drilling 5,550+ cards to go from beginner to professional developer. ¹³

• Spaced retrieval improves long-term learning for programming (ICER 2019): A controlled experiment found improved knowledge transfer for Python learners, though short-term performance dipped. ¹⁴

Take this with you

• Use Anki to replace the repetition that typing used to provide. You don’t need cards for syntax; the LLM handles that. Build cards for discrimination and pattern recognition.

• Don’t download decks. Make your own. The card you write encodes more than the card you import. Treat shared decks as inspiration, not shortcuts.

• 10 minutes a day is enough. 10 new cards, ~20 reviews. Consistency beats volume every time.

• Spaced repetition is not a replacement for building things. You still need to write code. Anki is the warm-up, not the workout.

The agentic tools will keep getting better, and we’ll type less every year. But understanding what the code does, and why it’s the right code, is worth keeping.

---

Footnotes

1. Faros AI — “AI Acceleration Whiplash” (2026) — Developer productivity impact of AI coding assistants, 33.7% output increase ↩

2. Smolen, P., Zhang, Y., & Byrne, J. H. — “The right time to learn: mechanisms and optimization of spaced learning” — Nature Reviews Neuroscience (2016) — Spaced training builds longer-lasting procedural memories ↩

3. Kang, S. H. K. — “Spaced Repetition Promotes Efficient and Effective Learning” — Policy Insights from the Behavioral and Brain Sciences (2016) ↩

4. Adrian Sieber — “ABC of FSRS” — Free Spaced Repetition Scheduler algorithm used by modern Anki ↩

5. Ebbinghaus, H. — Memory: A Contribution to Experimental Psychology (1885/1913) — The forgetting curve and spaced repetition ↩

6. Martin, R. C. — Clean Code: A Handbook of Agile Software Craftsmanship (2008) — “Indeed, the ratio of time spent reading versus writing is well over 10 to 1.” ↩

7. Vella, A. & Blincoe, K. — “The Impact of AI Coding Assistants on Software Engineering: A Longitudinal Study” (arXiv, 2026) — 82% of developers reported spending less time writing code ↩

8. Digital Applied — “AI Coding Tool Adoption 2026: Developer Survey” (2026) — Reviewing AI code (11.4 hrs/wk) outpaces writing new code (9.8 hrs/wk), n=2,847 ↩

9. Nielsen, M. — “Augmenting Long-term Memory” (2018) — Using Anki for mathematics and computer science ↩

10. Matuschak, A. — “Why books don’t work” — Memory systems and transformative flashcards ↩

11. Sivers, D. — “Spaced repetition and memory” — Practical use of SRS for learning ↩

12. Kinsella, J. — “Janki Method” (now offline) — Using spaced repetition for programming knowledge ↩

13. Semicolon & Sons — “How to Learn to Code I: Use SRS and Anki” (YouTube, 2020) — 9 years of using Anki to go from beginner to professional programmer ↩

14. ICER 2019 — “A Spaced, Interleaved Retrieval Practice Tool that is Motivating and Effective” ↩

Written by Rico Sta. Cruz

I am a web developer helping make the world a better place through JavaScript, Ruby, and UI design. I write articles like these often. If you'd like to stay in touch, subscribe to my list.

Comments