Learning to code has changed

Main ideas / lessons

• Learning to code has changed dramatically (especially due to AI).

  • The speaker contrasts “old” coding education (pre-AI) with “now” (2024–2026 era).
  • Even their own approach as a long-form tutorial creator is evolving.

• Pre-AI learning required deep fundamentals and problem-solving.

  • Limited resources compared to today.
  • A common learning process involved:
    • Searching the internet for documentation/resources
    • Using older YouTube channels and example projects
    • Learning language syntax directly (e.g., loops/conditionals, HTML elements)
    • Debugging and solving problems slowly (sometimes days)
  • This forced learners to build confidence through writing code manually and understanding what they were doing.

• AI makes coding easier, but can weaken fundamentals if over-relied upon.

  • The speaker argues AI can:
    • Generate projects and code quickly
    • Set up entire applications
  • However, beginners may end up:
    • Feeling productive without understanding
    • Missing out on critical thinking
    • Outsourcing “decisions” to an AI model rather than learning why something is correct

• Observed consequence: a growing divide in developer quality.

  • Over the last 6 months, the speaker worked with 200+ developers (via their program).
  • They report many junior/mid developers have lost:
    • Critical thinking
    • Ability to explain tradeoffs/alternatives
    • Even basic coding knowledge (examples mentioned: not knowing what a for loop is, or what type to use in Java/C++)
  • At the same time, not all learners are struggling—those who use AI effectively without crutching progress faster.

---

Methodology: how the speaker recommends learning to code in 2026

1) Start with a clear goal (most important thing)

• Ask yourself what you are trying to achieve:
  • If the goal is output-based (ship a project, build a SaaS, start a startup):
    • You may not need to “grind” every fundamental topic (e.g., data structures/algorithms) immediately.
    • You can focus more on outcomes and hire other experts later once successful.
  • If the goal is job-based (land a job, pass technical interviews):
    • Your learning approach must change.
    • You should prepare for how interviews test knowledge and reasoning.

2) Use structure / a roadmap (avoid hopping technologies)

• Create a plan that specifies:
  • What you’ll work on over the next 3, 4, and 6 months
  • Which topics/skills come first
  • How to progress through APIs, databases, etc.
  • How many projects you’ll build
• Why this matters:
  • Avoid “AI-chasing” and constant switching (frontend → backend → cloud → etc.) without depth.
  • AI can let many people build “decent” projects across areas, but employers still value depth and expertise.
• Practical principle:
  • Not knowing what to do each day is “half the battle”—structure reduces that friction.

3) Constant evaluation (measure whether you truly understand)

• Don’t just feel like you’re learning because you’re generating cool results with AI.
• The speaker recommends assessing yourself more than you used to, to confirm:
  • You understand what’s going on
  • You can reproduce work without relying on AI
• Concrete evaluation techniques:
  • Follow structured courses that include:
    • Quizzes / assessments
    • Projects you complete independently
  • If self-paced with AI:
    • Ask AI to quiz you
    • Ask for multiple-choice questions
    • Assign mini-projects you must complete yourself
    • Review AI-generated files and check if you can reproduce similar outcomes after reading
  • Do “offline tests”:
    • Turn off the internet
    • Write code for even ~10 minutes by yourself
    • Use this to reveal knowledge gaps
  • Track objective evidence of what you know vs. don’t know
• Purpose:
  • Reduce surprises in interviews when they quiz areas you don’t actually know.

4) Use AI tools—but as a balanced tutor/coach, not a cheat

• The speaker says you should learn AI tooling and related concepts:
  • Prompt engineering
  • MCP (mentioned explicitly)
• But you should also:
  • Still write code manually (avoid “crutching”)
  • Learn fundamentals deeply so you understand system design, make decisions, and question AI outputs
  • Treat AI as:
    • A tutor/coach beside you
    • A productivity and learning accelerator
• Warning:
  • Over-cheating with AI only harms you—your skills won’t develop in the real scenarios that matter (e.g., interviews, production debugging, reasoning tasks).

---

Claimed outcomes / overall recommendation

• People who:
  • Use AI effectively,
  • Maintain fundamentals,
  • Evaluate themselves,
  • Question AI outputs,
  • Build depth based on a roadmap,

are described as progressing much faster (the speaker claims up to 5x faster than they could, mainly due to easier access to information and better comprehension/retention).

• People who:
  • Lean too hard on AI without understanding,

fall behind as the skills gap widens.

---

Sources / speakers featured (as mentioned)

• Speaker: “Tech with Tim” (implied by the sponsor name and the video context)
• Referenced coding educators/channels (examples):
  • The New Boston
  • Sendex
• Program mentioned (personal):
  • Dev Launch
• Sponsor / platform:
  • Brilliant
• Institutions associated with Brilliant course development (mentioned):
  • MIT, Caltech, Google