Full-stack study scatters easily: frameworks, languages, databases, deploy, auth. Each rabbit hole is deep. What I care about now is simpler: get one complete path working.
A minimal loop
For example:
User opens a page → request goes out → server handles it → data is read or written → response returns → UI updates
If that path works, swapping the middle pieces later is less scary. If you only stack tutorials and never wire it yourself, integration bugs feel like fog.
What I deliberately write down
- Where the boundary is: what the client owns vs the server — don’t smear them together
- How failure looks: timeouts, empty data, auth errors — what the page shows and what the logs say
- What to reuse next time: folder layout, API conventions, how env vars are managed
Agent projects are the same story. Models look flashy; the hard part is keeping tools, state, and edge cases steady.
Ideas belong here too
Besides study notes, I keep product / tech ideas — even immature ones. Publishing them forces plain language, and makes restarting easier later.
If you’re learning full-stack or building agents, use these notes as a mirror, not a syllabus.
Discussion
评论
用 GitHub 账号登录即可留言。想法、纠错、补充都欢迎。