The hidden cost of 'we'll add it later'

I used to think technical debt was sloppy code. It's not — it's postponed decisions. Every "we'll add it later" is a loan, and the interest is paid in complexity.

Where my 'laters' compounded

"Users can change their email later." Except "later" meant every other account flow — password reset, notifications, export — had to pretend email is permanent. When the change-email feature finally got built, it cost ten times more because five systems had baked in the assumption.

"No auth needed for version one." True and fine — until adding auth meant retrofitting identity into every table, endpoint, and screen that had honestly, wrongly assumed anonymity forever.

"We'll pick the payment provider later." Meanwhile the pricing model, invoices, and refunds all grew around the assumption that money is simple. It is not, and neither was the retrofit.

What I do differently

For every "later," I ask: what assumption does building now write into the system? If the answer is "nothing much," later is fine — later is a prioritisation tool. If the answer is "the data model will assume it's impossible," then this is a one-way door wearing a two-way door's clothes.

Postpone work. Never postpone decisions about what's true.