Years of programming experience and large-scale system design knowledge produce dramatically better results with AI coding tools. AI amplifies existing expertise rather than replacing it.
## Core Insight
> "Use more plan mode on Claude Code.
>
> After a year of vibe coding, I've noticed something: my years of programming and my experience designing large-scale systems at TikTok help me get much better results with AI.
>
> The main reason is that I know what a good system looks like."
> — Jiayuan (JY) Zhang
## Key Principles
1. **Knowing what good looks like**: The critical input to AI coding isn't prompt engineering — it's having a mental model of what the right output should be
2. **Experience as discriminator**: Senior engineers can evaluate, redirect, and improve AI output because they recognize quality. Juniors can't tell if AI output is good or bad
3. **Plan mode insight**: Explicit planning (telling the AI what to build before building) produces better results because it forces you to apply your design knowledge upfront
4. **Multiplication, not replacement**: AI acts as a multiplier on existing skill. 10x engineer + AI = 100x output. 0x engineer + AI = unpredictable output
## Implications
- **For senior developers**: Your experience is more valuable than ever — it's the ingredient that makes AI productive
- **For junior developers**: The path to AI effectiveness runs through building genuine expertise first
- **For hiring**: The premium on experienced engineers may increase, not decrease, as AI tools mature
- **Career strategy**: Invest in deep domain knowledge and system design intuition — these are the inputs AI amplifies
## Connections
- [[AI Role Stack Redefinition]] — The role shift that makes experience the new bottleneck
- [[Competence as Joy]] — Competence remains the controllable variable; AI amplifies it
- [[Biographical Learning Method]] — Deep learning through narrative context builds the pattern recognition AI amplifies
## Source
- [[Recent Highlights Feb 4-7 2026 - Highlights]] — JY Zhang tweet, highlighted 2026-02-05
- [Readwise](https://readwise.io/open/985168160)