Skip to content
The Post-Build Cliff

The minimum you need to learn

You do not need to become a software engineer. You do need a thin layer of operational competence so that you are not at the mercy of every tool and every contractor.

This is the layer.

The non-negotiable five

Each one matters because skipping it puts you back at the cliff. None of them takes more than a Saturday to learn well enough.

  1. Git, just enough — clone, commit, push, pull, branch, revert. Not rebase. Not interactive add. Not anything fancy.
  2. Environment variables — what they are, where they live, how to set them in production without redeploying.
  3. A version-controlled deploy — pushing to a branch causes a deploy. No “click publish” button as the source of truth.
  4. Reading logs — when something breaks, the answer is almost always already printed somewhere. Find it.
  5. One backup you’ve actually restored from — until you’ve restored, you don’t have a backup. You have a bet.

That’s the whole list. Five things. None of them is a programming skill — they’re operational skills around the work, not the work itself.

The “nice but not yet” list

Things you’ll be tempted to learn before the five above. Don’t. They’re useful eventually; they’re a distraction now.

  • DNS and custom domains beyond pointing one A record
  • Database schema design
  • Anything involving Kubernetes, Docker, or “infra”
  • Writing tests
  • CI/CD pipelines beyond “deploy on push”
  • Monorepos, microservices, “scaling architecture”
  • Anything described as “best practices” without a specific reason

The reason these are not yet and not no: most of them are real skills with real value at the right stage. But for a non-technical founder building their first AI-assisted thing, learning Kubernetes is the founder equivalent of learning to weld before you’ve finished learning to drive. It signals seriousness without delivering it.

How to learn each, fast

The whole curriculum is a Saturday and a Sunday if you commit. Each row in the table below has one primary resource — the shortest viable path — and a single practice rep that proves you’ve done it.

SkillBest 60-min resourcePractice rep
GitThe Missing Semester · GitMake 10 commits to a real repo, including one revert
Env varsYour hosting provider’s “environment variables” docsMove one secret from code to env, redeploy, confirm it still works
Deploy from gitCloudflare Pages or Vercel “connect a repo” walkthroughDeploy a hello-world, then change one word and watch it auto-deploy
Reading logsYour hosting provider’s “logs” tabTrigger an error on purpose (throw new Error('test')), find it in the logs
Restoring a backupYour DB provider’s restore docsRestore a backup to a non-production copy. Compare row counts

The “practice rep” column is the part most people skip. Reading without doing is procrastination dressed up as learning.

The order to learn them in

Strong order: git → env vars → deploy → logs → restore.

Each unlocks the next. Without git, you can’t deploy from a branch. Without env vars, you can’t separate dev from prod. Without a real deploy pipeline, you don’t have logs to read. Without logs, you don’t know when to restore. Without ever testing a restore, your backup isn’t real.

Don’t start somewhere fun and circle back. Start at the start. The lessons that follow walk through each one.

Skills you’ll pick up in this module

  • The five operational primitives that mean you’re no longer dependent on the tool you built in
  • A shared vocabulary for talking to engineers without sounding lost
  • The judgement to recognise when a “best practice” is actually right for you, right now

References & further reading