Opinionated Shovel Makers vs. Opinionated Diggers

Last updated: April 14, 2026

Lukas Kawerau’s post, "Opinionated shovel makers vs. opinionated diggers", gave language to something I’ve been feeling for a while.

AI tooling creates a strong pull toward shovel-making: orchestrators, frameworks, meta-tools, abstractions over abstractions. Some of that is genuinely useful. I use those tools every day.

But most days I care more about the dig site than the shovel. I care about the product, the domain, and whether the software is actually useful to someone.


My read

I’m mostly an opinionated digger.

I did not build bosun and the surrounding harness because I wanted to spend my life tuning orchestrators. I built it because my actual work kept breaking in predictable ways: context would evaporate between sessions, quality would drift, and I would keep re-explaining the same constraints.

So I treat shovel-making as infrastructure work in service of product work. Not the destination.


The two failure modes

There are two easy ways to get this wrong:

  • never build tools, then pay the same friction tax forever;
  • only build tools, and quietly stop shipping domain outcomes.

The middle path is usually better: build the minimum shovel that removes recurring pain, then go back to digging.

That is basically Start Manual, Automate Later stated in different words.


The filter I use

Before I build new harness infrastructure, I ask three questions:

  1. Is this pain recurring?
  2. If fixed, will it compound across sessions/agents?
  3. Can I timebox the implementation?

If the answer is yes to all three, I build the shovel. If not, I keep digging.

This one filter prevents a lot of tool-builder rabbit holes.


Where this lands

I agree with Lukas on the core point: AI increases leverage for builders who already know what they want to build.

For me the practical stance is:

  • stay anchored to product/domain outcomes,
  • invest in harness pieces only when they clearly compound,
  • avoid identity lock-in around either camp.

So the short version is: build shovels in service of digging, not instead of it.

See also: