TL;DR
The author argues that large language models (LLMs) are not required to learn programming; curiosity, hands-on practice, and community resources are sufficient. They emphasize working through problems, sharing knowledge with others, and using open-source materials and projects as primary learning tools.
What happened
In a personal essay published January 14, 2026, the author lays out their approach to learning programming and teaching others. Responsible for educating several neurodivergent learners (and not a formally trained teacher), they describe relying on freely available resources—open-source code, online books, tutorials, forums, meetups—and on iterative, experimental learning: form hypotheses, test them, and revise. The piece acknowledges that LLMs can summarize large codebases and are convenient and nonjudgmental conversation partners, but the author cautions that such tools do not replace the deep understanding built by doing the work yourself. Retention, they argue, comes from practicing skills until you can solve novel problems unaided and from explaining ideas to other people. The author lists concrete project-style entry points—courses and projects like Programming Language Foundations in Agda, NAND to Tetris, reading or hacking Postgres source, restoring vintage hardware, and building small systems—as ways to learn by doing and to connect with the wider programming community.
Why it matters
- Reinforces that accessible, low-cost learning pathways (open source, tutorials, meetups) still exist without relying on LLMs.
- Challenges the idea that AI tools can shortcut the sustained practice required for durable understanding.
- Highlights the pedagogical value of teaching others and engaging with human communities to solidify knowledge.
- Signals potential risks if learners substitute interaction with people and hands-on experiments with passive LLM queries.
Key facts
- The author believes LLMs are not necessary to learn programming.
- They have taught many people about programming and system design and support neurodivergent learners, though they are not a trained teacher.
- Programmers commonly share knowledge openly—source code, books, tutorials, presentations, forums, and meetups are widely available.
- Learning, in the author's view, requires curiosity, reading, downloading code, forming hypotheses, running experiments, and iterating.
- The author argues that solving exercises and practicing until you can tackle novel problems leads to lasting retention.
- LLMs are described as always-available, nonjudgmental helpers, but relying on them can reduce practice explaining and teaching to others.
- Examples of hands-on learning paths the author cites include Programming Language Foundations in Agda, NAND to Tetris, studying Postgres source, building a text editor, and implementing a hash table in C.
- The author notes that large public codebases and documentation (for example, Linux kernel materials) are present in the corpus LLMs draw on, which can make AI summaries possible.
What to watch next
- Whether learners increasingly default to LLMs instead of community resources and hands-on projects (not confirmed in the source).
- Studies or reports comparing long-term retention from LLM-assisted learning versus traditional practice-based learning (not confirmed in the source).
Quick glossary
- Large language model (LLM): A machine-learning system trained on large amounts of text to generate or summarize human-like language and answer questions.
- Open source: Software and materials whose source code or content are freely available for inspection, modification, and sharing.
- Compiler: A program that translates source code written in a high-level language into machine code or another lower-level form.
- Operating system (OS): System software that manages hardware resources and provides services for computer programs.
Reader FAQ
Do you need LLMs to learn programming?
The author asserts no; they recommend curiosity, hands-on practice, and community resources instead.
Are LLMs useful at all according to the author?
They can be convenient and provide nonjudgmental, always-available help, but the author warns they shouldn't replace active practice and human interaction.
What learning activities does the author recommend?
Working through projects, reading and compiling real source code, forming hypotheses, running experiments, and teaching others.
Is formal teaching experience required to help others learn programming?
The author says they are not a trained teacher but have still taught many people; formal training is not presented as a necessary condition.
Posted on January 14, 2026 I don’t think you need LLMs to learn programming. I recognize that people need different strategies and tools to learn new skills and information. I…
Sources
- How I learned everything I know about programming
- 🕵️♂️ The Art of Self-Learning: How to Teach Yourself …
- How to Learn to Code from Zero
- How to make interest & curiosity while learning a …
Related posts
- Best alternatives to Spotify for streaming and listening to music
- How I Learned Everything I Know About Programming Without LLMs
- Aerial Photos Reveal Scale of China’s Massive Wind and Solar Buildout