Run Fast, With Your Shoes Tied

👻 Recreation notice. The original ran as KOHO’s EOW send on April 10, 2023, and lived on Notion until that page was deleted. This version was reconstructed in November 2025 from what the piece actually argued, not a saved draft, then lightly revised for this canonical copy. Call it a faithful cover, not the original recording.

There’s a seductive lie we tell ourselves in tech: we’ll clean it up later.

Later, when we have more time. Later, when the pressure’s off. Later, when we’ve shipped this one thing. The problem is, “later” is a debt collector with compound interest, and eventually it comes knocking with a baseball bat.

I’ve spent two decades watching teams fall into this trap. I’ve fallen into it myself. The irony is that the shortcuts we take to go faster almost always make us slower. Not immediately. Not obviously. But inevitably.

The Untied Shoelace Problem

Picture a sprinter who doesn’t bother tying their shoes because it takes 10 seconds. For the first 50 metres, they’re flying. Then they’re face-down on the track, watching everyone else pass them.

That’s what it looks like when you:

• Skip writing tests because “we need to ship”
• Hardcode that config because “we’ll parameterize it later”
• Deploy without monitoring because “we’ll add observability next sprint”
• Bypass code review because “it’s a small change”

Each of these decisions feels rational in isolation. Each one is a shoelace waiting to trip you.

What Tied Shoes Actually Look Like

At KOHO, we’ve reduced model deployment time from 15 days to 1 day. We’ve maintained 100% uptime for six consecutive quarters. We’ve dropped costs from $280K peaks to 35¢ per transaction.

None of that happened by moving fast and breaking things. It happened by investing in the foundations that let us move fast:

• MLOps platforms that make deployment boring and repeatable
• Observability that tells us what’s wrong before customers do
• Automation that removes humans from the loop where humans add risk
• Documentation that means we’re not held hostage by tribal knowledge

These aren’t constraints on speed. They are speed. Just not the kind you can see in a single sprint.

The Perpetual Motion Machine

The goal isn’t to go fast once. It’s to build a system that goes fast forever, or at least until you intentionally change direction.

Ask yourself: will this decision make us faster next month, or just today?

If it only makes you faster today, you’re borrowing from your future self. And your future self is going to be pissed.

Practical Shoelace-Tying

1. Invest in CI/CD early. The teams that “don’t have time” for proper pipelines spend more time on deployment than teams that do. Every single time.
2. Write the test. Not because tests are virtuous, but because debugging in production at 2 AM is a terrible use of your finite time on Earth.
3. Automate the second time. First time you do something manually? Fine. Second time? That’s a script waiting to be written.
4. Make the right thing the easy thing. If doing it correctly is harder than doing it wrong, fix the system, not the people.
5. Pay down debt deliberately. Schedule it. Budget for it. Treat it like the investment it is.

The Counterintuitive Truth

The fastest teams I’ve ever worked with are also the most disciplined. They’re not fast despite their rigor. They’re fast because of it.

They’ve tied their shoes so many times it’s automatic. They’ve built systems where quality is the path of least resistance. They’ve learned that sustainable speed beats unsustainable speed every time, because sustainable speed is the only kind that actually compounds.

So tie your shoes. Take the extra 10 seconds. Then run like hell.