Dejé de perseguir camiones

I Stopped Chasing Trucks

Paul Graham said do things that don't scale. He didn't say do all of them.

Here is how it used to go. A shipper posts a load on Rutazo — say, forty tons of grain that needs to move from Rosario to Buenos Aires on Thursday. No carrier bids on it. So I'd open WhatsApp, work through my contacts, find someone with a truck heading that way, and talk them into placing a bid on a platform they had never used.

The load moved. Everybody was happy. And I had spent four hours to close a transaction worth a small commission.

I did that for months. Then I stopped, and I made myself a rule:

If a transaction cannot close without me, this is not a business. It is a job with extra steps.

I still believe that rule. What I got wrong was when to run the test.

The misread

"Do Things That Don't Scale" is probably the most quoted startup essay ever written, and I think it is also one of the most misapplied. Founders read it as permission to do manual work. It isn't. It's an argument that certain manual work buys you something you cannot buy any other way.

The canonical example is the Airbnb founders photographing listings themselves. What gets remembered is the effort. What matters is the residue: after they left an apartment, that listing had good photos forever. The work was manual and it was permanent.

That is the part I missed. Not every piece of unscalable work leaves something behind.

Two kinds of unscalable work

There is work that builds an asset, and work that evaporates on contact. They look identical while you're doing them. Both are manual, both are exhausting, both feel like hustle.

Chasing a truck for a specific load evaporates. I close it, I get paid once, and on Friday morning I am exactly where I started. Nothing accumulated. I bought a transaction.

Recruiting thirty carriers who run one specific corridor does not evaporate. It is just as manual, just as unscalable, and it takes just as long. But those thirty carriers are still there next month. They start bidding on loads I never touched. I bought supply.

I quit doing both. I should have quit the first one and doubled down on the second.

The test

The question I now ask before any manual task, mine or a client's:

Does this still exist next Tuesday?

If the answer is no, it's a transaction. Do it if it pays the bills, but don't confuse it with building. If the answer is yes, it's an asset, and manual is usually the fastest way to get it.

Applied honestly, this kills a lot of work that feels productive. It also protects work that feels wasteful and isn't.

Why my test was rigged

Here is the part I only understood recently.

When I said "let's see if a transaction can close without me," I wasn't testing my product. I was testing liquidity. Those are different things and I couldn't tell them apart.

Rutazo has close to 2,000 registered users, with zero spent on advertising, and large shippers reaching out on their own. By the standards of a two-sided marketplace, that is the hard part already solved. But those users are spread thin across a whole country. A shipper posts a load in one province and the nearest interested carrier is four hundred kilometers away and doesn't know the platform exists.

No software fixes that. A marketplace without density does not close transactions on its own no matter how good the product is. So when nothing closed, I read it as "the product isn't ready." It might be. I have no idea, because I never ran the test under conditions where it could have passed.

The real test isn't can a transaction close without me. It's can a transaction close without me, on one corridor, with thirty active carriers on it. If it fails there, the product is broken and I go fix it. Today I can't tell you which one is failing, and that's a measurement problem I created myself.

What this looks like outside marketplaces

I see the same confusion in almost every operations engagement I take, wearing different clothes.

A sales team manually writes every proposal. That's evaporating work — each proposal dies the day it's sent. But the first ten proposals, written manually and then studied for what repeats, are an asset: they're the spec for the template.

A support lead personally answers every escalation. Evaporating. The same lead tagging every escalation by root cause for six weeks is an asset, because now you know what to build.

The manual work isn't the problem. The problem is doing manual work that leaves no sediment, and then concluding — after months of it — that automation won't work here.

It isn't that the work doesn't scale. It's that nothing is accumulating underneath it.

What I'm doing now

One corridor. Not national coverage — density on a single route, chosen by whichever lane already shows up most in posted loads.

Then the actual sequence: find one large shipper with recurring weekly volume on that lane, use those predictable loads to recruit carriers who run it, and only then check whether transactions close on their own.

That last step is the only one that scales. The first two are pure manual labor, and I'll be doing them myself for weeks.

The difference is that this time I know what I'm buying.

Agustin Roig is a product and technology consultant. He builds vertical platforms and helps growing companies fix the operations underneath them. Let's talk.