Tech
"What breaks first when 100 people use this?"
Time-to-first-byte, error states, async patience, what's cached, what's optimistic, what fails loudly vs silently. Whether the trade-offs are named or accidental.
It works on the happy path. The defaults carry it.
Vercel + framework defaults. Any deviation from the demo flow surfaces unhandled errors.
Errors are caught. Performance is uninvestigated.
Try/catch blocks exist. Loading states are spinners. Nothing measured, nothing optimised.
Trade-offs are visible and intentional.
Hot paths are named. Caching is purposeful. The thing that would break at scale has a written plan.
Boring where it matters, sharp where it doesn't.
Architecture is dull on purpose. Complexity is concentrated where it earns its keep. Future-you can read it cold.
- Loading states are infinite spinners with no timeout
- Clever architecture doing simple work
- Errors swallowed silently to keep the demo green
- 1Put a millisecond budget on every interaction. Measure. Fix the worst one.
- 2Remove your most clever abstraction. Ship the boring version. See if anyone notices.
- 3Write the failure mode for 100x current load. Now design for it.