The Constraint Is the Teacher

March 18, 20262 min read#building#constraints#ai

Most software fails for the same reason most meetings run long. Nobody removed anything.

I trained as a Chartered Accountant. Twenty years of watching processes grow one well-meaning step at a time until nobody remembers why half of them exist. So when I started building with AI, I gave myself one rule that sounds harsh and turned out to be a gift: ship small, ship honest, and remove every step that only exists because a human gets tired.

That rule is a constraint. And the constraint is the teacher.

What one hard limit reveals

When you cannot spend three weeks debating a database schema, you find out what the product actually is. When you cannot add a feature "just in case," you find out which features were fear, not need. A tight limit does not make you cut corners. It makes the corners visible.

Most people add steps to feel thorough. I remove them until what remains is the shortest honest path. Then I automate that, and only that.

What the machine actually is

I do not build these alone, and I do not pretend to. ZeroOrigine is an ecosystem of eight AI Minds, each with a role. One finds problems worth solving. One argues about whether an idea deserves to exist. One refuses to ship when the work is not good enough. They research, build, test, and launch, and my job is the two or three decisions a day that genuinely need a human.

The honest scoreboard: a handful of products are live. Revenue is zero. Some builds have failed the same way twice and been dropped, on the record, not quietly deleted. Those numbers live on the machine's own site, and I do not dress them up here.

Every product is a hypothesis

"I believe this specific group of people has this specific problem, and would pay this specific amount to remove it." Ship it. See if anyone cares. Most will not, and that is the design, not the disappointment. I am not chasing a hit rate. I am looking for the few where the problem was real and someone opens their wallet and says: yes, this.

The ones that work get nurtured. The ones that do not get honest data about why. Neither is wasted.

Back to zero

Underneath all of it is the oldest idea I have. Strip away the assumptions about what the market wants, what a company is supposed to look like, how software is supposed to get made. Return to zero. What survives the return is worth building.

The number was never the point. The removing was the point.

What in your work exists only because a human used to do it?

Read next

July 14, 2026 · 2 min

Automating the Profession That Trained Me

Twenty years inside accounting taught me where the hours actually die. Not in the judgment. In the chasing. Speed CA is my attempt to remove the steps that only exist because we are human.

Read →
July 14, 2026 · 1 min

The Accountant Sits on the Richest Data in the Company

Every transaction, every anomaly, every quiet lie passes through one system before it becomes a financial statement. We trained accountants to report that data. We should have trained them to think with it.

Read →
August 19, 2026 · 4 min

You don't fear the algebra lesson, you fear the letter from the state

A pre-filled notice, a real calendar of your state's actual deadlines, and reminders before they matter, so compliance stops eating the hours you meant to spend teaching.

Read →
August 17, 2026 · 4 min

You lost your right to get paid before the job was even finished

LienClock calculates every preliminary notice and lien deadline for your job the moment you enter the state, GC, and dates, so the clock never runs out without you knowing.

Read →

← All essays · If this made you think, tell me: cajagdishlade@gmail.com. I answer everyone.