Summary of "If I Started Coding In 2026, I'd do this"

Core thesis

Programming isn’t dead. AI makes coding more important and changes how you learn and work. Treat modern large language models (LLMs) as your primary learning and coding assistants rather than just search engines.

Actionable roadmap (step-by-step)

Get a paid AI code-focused subscription
- Recommended: OpenAI Codex / ChatGPT code-tier and Anthropic’s Claude Code or Claude Max.
- Budget: spend ~ $20/month for a useful plan; upgrade if you can afford higher-tier models.
- Warning: avoid low-quality/free/basic models for “knowledgeable” work — they won’t teach or reason as well.
Learn to “talk to the computer”
- Install/use a desktop or CLI agent for the model so you can interact quickly.
- Practice good prompting and iterative dialogue — the model is the new primary interface.
Start with playful / “vibe” coding
- Give vague prompts (e.g., “make me a game I would enjoy”) and let the AI build a small project.
- Expect messy generated code at first — it sparks curiosity and keeps learning fun.
- Spend a couple of months (or less) on these explorations to build interest and momentum.
Alternate AI-assisted and hand-built projects
- Build some projects entirely by hand (no AI). Then use AI as a judge/reviewer to critique and improve your code.
- For other projects, use AI as a pair programmer to speed iteration, then audit and learn from the output.
Learn your stack and toolchain
- After choosing a path (web, mobile, systems, Web3, Rust/C++, Python, etc.), become familiar with stack-specific tooling and workflows.
- AI accelerates learning, but toolchains diverge per domain.
Consume structured content, free or paid
- YouTube and free tutorials are fine—many of those resources are already in AI training sets.
- Paid courses can still provide structure, unique teaching styles, or edge content not present in public training data — they’re worth it if your budget allows.
Build a personal brand and public accountability
- Share projects and learning publicly (Twitter/X, YouTube, etc.). Human creators’ brand/value is something AI cannot fully replicate.
- Public sharing creates accountability, grows an audience, and leads to opportunities (jobs, freelance work).

Practical learning tactics with AI

Role-play exercises: reverse roles where the AI quizzes you and corrects errors.
Ask targeted questions about code the model generated; request “explain like I’m five” for tricky concepts.
Use AI as a code reviewer/community replacement for beginners — it can flag many issues that would otherwise overwhelm human reviewers.
Don’t deploy AI-generated code to production without careful review and testing.

Product / model recommendations and warnings

Suggested models: OpenAI (Codex / higher-tier ChatGPT with code abilities), Anthropic Claude (Code / Max).
Use desktop/CLI integration for faster workflows.
Avoid relying only on free/basic models or lower-quality services for learning complex topics.
Beware of model training-data opacity — models know a lot, but always verify accuracy.

Mentions of content / guides

The speaker references prior videos they made (e.g., “Coding is dead” and a topic-wise roadmap video).
Advice: watch structured roadmaps or tutorials (the creator has a roadmap and many archived tutorials).
Suggestion: use both free YouTube content and paid learning tracks depending on your budget and need for structure.

Speaker / sources

Advice comes from the video’s creator/host — an experienced developer and YouTuber with ~460k subscribers. Background includes freelancing, security bug bounties, teaching (hundreds of tutorials), and building startups.
Platforms referenced: OpenAI (Codex / ChatGPT code tiers), Anthropic (Claude Code / Claude Max), and general references to YouTube, Google/Stack Overflow, and free/open-router models.

Key takeaway

Start with a good code-capable AI subscription, use playful AI-driven projects to spark curiosity, learn fundamentals and stack toolchains, alternate hand-built projects with AI critique, and build a public brand. AI is the accelerator — but fundamentals, review, and real-world practice still matter.