Do not build apps!

Many developers have more ideas than free time. AI can now write a first version of a tool in one afternoon. This makes it tempting to skip thinking and go straight to a new repository.

But writing code was never the expensive part. The expensive part comes later:the dependencies, the bugs, and the unfinished project that keeps waiting for your attention.

So before you start, ask two questions.

1. What problem am I solving?#

Answer in one plain sentence. "A tool that does X" is not a problem. "I want to log in without typing a password" is a problem. Writing the problem down often shows you that someone already solved it, or that it was not a real problem.

2. What already solves it, and where does it fail for me?#

Most people skip this question. Search for existing solutions. For each one, name the specific reason it does not work for you. If you cannot name a real failure, you probably do not need to build anything.

An example from my own experience#

I wanted to use my fingerprint reader for passkeys in Firefox on Linux. The idea seemed small, but the problem underneath was simple: I wanted to log in without typing a password.

First idea: extend pass with WebAuthn. Firefox on Linux can't use a fingerprint reader as a passkey authenticator directly. The workaround was a software authenticator built on my pass store, which meant implementing CTAP2, a kernel HID interface, and CBOR encoding. That was far too much work for this problem, so I dropped it.

Second idea: pass-nostr. I wanted to store passwords on the Nostr network. Before writing any design, I asked whether something already existed. That's when I found ngit, which extends git to work with Nostr. I already use it in my stack, and pass already runs on git, so the problem was already covered. I dropped that idea too.

These were two different ideas, left for two different reasons:

WebAuthn for pass: the cost was too high for the problem, and a hardware security key already solves it.
pass-nostr: the solution already existed in my own stack.

The lesson is simple: asking the question first saved me from the design work entirely.

Neither idea was bad. Both are interesting. They just didn't solve my problem better than what I already had.

The habit#

Next time you have an idea, spend ten minutes on the two questions before you create the repository. If the problem is clear, and no existing solution fails in a way you can name, go build it. You will probably build something better, and you will finish more of what you start.

References#

Password and passkey tools

Passkeys and hardware keys

Linux fingerprint support

Nostr and git

1. What problem am I solving?#

2. What already solves it, and where does it fail for me?#

An example from my own experience#

The habit#

References#

Emre Yilmaz

Comments & Discussions (0)