# The engineering words behind reliable apps, and which ones matter at your size

Source: https://sarang.pro/engineering-words-reliable-apps/
Author: Sarang | Published: 24 October 2026

A list of 111 system design terms went viral among people building apps with AI: rate limiting, caching, load balancing, idempotency, sagas and a hundred more. The comments split into two camps. Some people planned to paste the whole list into their AI tool and ask it to handle everything. Others pointed out that most of the list solves problems a small product will not meet for years.

Both reactions miss the useful middle. The list is a map of what can go wrong. The skill is knowing which parts of the map you are standing on.

## Handle these now

These protect a product of any size from the failures that hurt most: leaked keys, lost data, frozen requests and duplicate charges.

- **Secrets management.** API keys live in environment variables or a vault, never in code, screenshots or chat.

- **Backups and restores.** A backup you have never restored is a hope, not a plan.

- **Timeouts.** Give up on a slow call after a set time, so one hung request does not freeze everything.

- **Retries with exponential backoff.** Try again after 1, 2, then 4 seconds, so a brief outage does not become a long one.

- **Idempotency.** If a payment, email or post is retried, it must not happen twice.

- **Webhooks, verified and logged.** Check that each incoming message really came from the service it claims.

- **Access rules.** Each person, token and agent gets the least access it needs.

- **The big three web holes.** SQL injection, cross-site scripting and loose CORS settings.

## Plan for these as you grow

Once you have real traffic, paying clients or a small team, a second set starts to matter:

- **Caching** to avoid fetching or computing the same thing again.

- **Message queues and background jobs** so slow work does not keep users waiting.

- **Logging, monitoring and alerts** so you hear about a failure before a customer tells you.

- **Automated tests in your deployment pipeline**, so every change is checked the same way.

- **Database indexes and migrations** that keep queries fast and changes reversible.

- **Cost tracking**, because hosting, build minutes and AI calls grow quietly.

## Know the word, skip the work

Load balancers across many servers, Kubernetes, service discovery, distributed transactions and the saga pattern solve the problems of large systems with many services and teams. It is worth knowing what they mean so you can follow a conversation. Building them into a product with a few hundred users adds cost and complexity without benefit.

## Why this matters for non-engineers

Founders and managers do not need to implement any of this, but they do need to ask the right questions. ‘What happens if the payment provider is slow?’ and ‘Can a retry charge someone twice?’ are business questions with technical answers. A short tiered list makes those conversations faster and keeps engineering effort where it pays off.