Shipping Measure Once Exposed the System I Need to Fix

On 25 August 2026, I built Measure Once, the fifth tool in my plan to create 100 simple online tools for myedgelife.com. I shipped it after several iterations, but I did not feel that it was my best work.

That was exactly why publishing it mattered. Consistency and honesty felt more useful than maintaining the appearance that every tool arrived fully resolved.

Measure Once also exposed a larger issue. I had started creating a system for generating and building tools, but the system itself already needed work. The idea-generation instructions needed improvement, the Markdown reference files needed to be updated, and my way of organising the work would not hold up as the number of tools increased.

The fifth tool became less of a finished answer and more of a test of how I was building.

Why did I create a ChatGPT project for 100 tools?#

I created the ChatGPT project to give the 100-tools plan a shared context. Instead of explaining the same goals and preferences every time I needed a new idea, I wanted one place that understood what I was trying to build for myedgelife.com.

I then wrote instructions for generating ideas. Those instructions included my requirements and preferences for the kinds of tools I wanted to create.

Setting up the project made the process more repeatable, but it did not make the instructions complete. Once I saw the ideas it produced, I felt that the guidance still needed substantial improvement.

That was the first warning: writing instructions once is not the same as having a reliable idea-generation system. The output can reveal gaps in the instructions that were difficult to see while writing them.

Why did I choose Measure Once?#

I chose Measure Once from the second idea generated by the project. The starting problem was the friction around remembering body measurements and using them when buying clothes online.

I often forget my measurements even after taking them. Online size information may be shown in centimetres or inches, which adds another point of confusion. Being measured again by a tailor also takes time and can feel uncomfortable.

I gave Claude the following starting point for an idea-grilling session:

Problem: It is difficult to get the right fit online, and I do not remember my exact measurements. I keep forgetting them even after taking them. While browsing online, I often see sizes in centimetres or inches, and it is confusing to select one. When going to a tailor, they have to measure me every time, which takes time, and I often feel uncomfortable.

Solution: A web app that guides a user through taking their own body measurements, either alone or with help from someone they trust. It would create a file or message that could be shared through WhatsApp or saved in a notes app, store measurements in both centimetres and inches, ask how the user prefers clothes to fit, and recommend sizes for formal, casual, and leisure clothing.

This was a starting proposal, not a finished specification. It combined a real problem with several possible features, and those features still needed to be questioned before development.

How did the idea become a product brief?#

The idea-grilling session pushed the initial prompt into a deeper discussion of the problem and the proposed solution. Through that back-and-forth, the idea evolved and was eventually defined in a product requirements document, or PRD.

The PRD gave the build a more structured input than the original paragraph. I then used it to create the web app in Claude.

This sequence is becoming part of my working method:

1. Generate a possible tool idea.
2. Challenge the first version of the idea.
3. Turn the resulting decisions into a written PRD.
4. Use the PRD to build the app with AI.
5. Review and iterate on what was produced.

The sequence helped me move from an idea to an app, but Measure Once showed that a sequence alone is not enough. The quality of every stage depends on the instructions and reference material feeding it.

Why did I ship something I had mixed feelings about?#

I shipped Measure Once because waiting until every tool felt like my best work would make it easy to hide the uneven parts of the process. After multiple iterations, I reached a version that felt acceptable to publish, even though I still had mixed feelings about it.

Publishing it does not mean I consider every decision settled. It means this is the version I was able to make with the process I currently have.

There is a tradeoff here. Shipping supports consistency and creates a real record of the work. Waiting longer might improve a particular tool, but it can also turn “not perfect yet” into a reason never to finish.

For a building-in-public project, presenting only the work I feel completely confident about would leave out an important part of the story. The uncertainty is part of the record too.

What did Measure Once expose about my process?#

Measure Once exposed that my building system needs maintenance, not just repetition. The ChatGPT project, idea-generation instructions, Markdown references, PRD, and build process are connected. Weakness in one part can affect everything that follows.

Three gaps became clear.

First, the idea-generation instructions need to improve. I provided my preferences, but the resulting ideas showed me that the instructions do not yet express those preferences well enough.

Second, the Markdown reference files need to change as I learn. If they remain fixed while my standards and process evolve, future builds may repeat decisions I already know need reconsideration.

Third, I need a clearer way to organise the growing set of ideas, prompts, instructions, reference files, PRDs, builds, and iteration notes. Managing this informally may work for a few tools. It is unlikely to remain manageable across 100.

The complexity is not only in the apps. It is also in the material required to produce them consistently.

What needs to change next?#

The next step is to improve the system using what this build revealed. That means revising the idea-generation instructions, updating the Markdown references that guide development, and creating a more deliberate structure for organising the work.

I have not solved those problems yet. Measure Once gave me evidence that they exist and a concrete build I can use to examine them.

The fifth tool may not be the one I feel best about. It has still been useful because it changed what I think needs attention. I began with a plan for generating more ideas. I ended with a clearer understanding that the system behind the tools has to be built and iterated too.