AI and scaffolding

All knowledge workers understand that every project has a lot of scaffolding wrapping the final deliverable.

I’m a web developer, so my job is erecting scaffolding around a precious nugget of creative work. It’s about gluing a frontend to a backend, a backend to a database, with a grab bag of dependencies to be tested and debugged.

If you’re a writer, there’s a different scaffolding system: gather notes, build an outline, proofreading, editing, etc.

If you’re an attorney, you spend a lot of time reviewing case law, fitting arguments into specific templates, and figuring out precise language (that is incomprehensible to the rest of us).

The problem

There’s a problem with this scaffolding: it’s not what makes your work valuable.

As a developer, I’m paid to deliver features.

As a writer, you’re paid to write compelling prose.

As an attorney, you’re paid to protect and further a client’s interests.

We all get lost in the scaffolding. I wouldn’t be surprised if I spend 80% within the system’s scaffolding and only 20% writing the business logic – the reason the product exists.

This scaffolding is necessary. We build within a culture with norms that make projects easier to transfer.

I don’t want to write software from scratch, just like an attorney wouldn’t want to create a new contract with different language. It slows everything down.

What should we do?

The scaffolding machine

You might think this post is leading to an AI conclusion.

Spoiler: it is.

I would like to point out that I’m much more measured about AI discourse than my developer friends.

When I create a product, typing isn’t the issue. The hard work is in the membrane between the product and reality.

Defining a market, identifying the problem, designing a solution, and validating assumptions are the hard work.

That’s where the creative solution lives. Many entrepreneurs type a lot and make $0 for a reason.

They mistake typing for hard work.

Many people start typing right away because there’s a lot to do. Before you can complete your design, there’s a ton of boilerplate to write and hours of redundant work.

You need to build a sign-up form, model the database, integrate payments, etc.

Each task is high stakes but they don’t vary between projects.

Now we have AI.

Claude and ChatGPT excel at generating boilerplate. When given a task and context, these tools prepare your solution for distribution.

They’re like a toy car. Just wind it up, point it, and let go. If it’s aimed correctly, it’ll get there.

Our creative idea is easier to distribute to users.

Use a scaffolding machine

Gather the right tools

Signing up for ChatGPT is easy, but it leaves productivity on the table.

The chat interface is great for discovering new information and manipulating output, but I find it slow compared to dedicated tools.

If you’re a programmer, invest in a Cursor subscription.

If you’re a writer, check out Lex.

If you’re an attorney, find and master the latest GPT wrapper application.

These tools are tailor-made for your use case. They will keep you in flow and nudge you along more subtly than a chat conversation.

Build a habit

I forgot to use the tools when I first tried to adopt them.

When you’re at work, it’s easy to concentrate on tasks instead of finding new ways to do them.

One trick that worked for me was rewriting my to-do list to include AI prompts in the task description.

Instead of “Create button,” the task might be: “Ask Claude to add a save button that refreshes the table after succeeding.”

The verbose description reminded me that my hard work of scoping the task was done and I should let my copilot take over.

Shift your focus

If your job seems to be mostly grunt work, it’s time to start taking on more creative tasks.

For those of us who just want to make, this should be a relief. Finally, the boring parts of creation can be handled, and we can focus on creating.

Creation is what we’re here for. Part of being human is the drive to create. By letting AI handle non-creative tasks, we can focus on the human aspects of our work. We can focus on our user, understand their pain and frustration, and build a new solution.

That’s it for this issue. Thanks for reading.

Hunter