It was a real piece of engineering that made my GitHub graph look busier than my work was. Both of those things are true, and only one of them matters to a reviewer.
I built a tool that paints patterns onto a GitHub contribution graph. It models a design grid as a versioned schema, maps it to backdated commit workflows, and triggers the whole thing in under two minutes. As a piece of engineering it is fine.
It was also pinned to the top of my GitHub profile for a year, and I have taken it down.
What a reviewer sees
A green contribution graph is supposed to be evidence. The moment you ship a tool whose output is a green contribution graph, you have not just built something clever — you have told every reviewer that your own graph is not evidence any more.
That is a bad trade. It costs me the credibility of a real signal to keep a novelty project on the shelf, and no hiring manager reads it the generous way.
Keeping it honest
I have not deleted it, and I have not hidden it. It still has its case study on this site and it still has a row in the work archive, labelled as archived. Pretending it never existed would be the same kind of dishonesty in the other direction.
What it no longer does is front the portfolio. Featured work should be the things I want to be judged on, and this is not one of them.
The general rule
Consistency is worth showing. A streak is not the point; showing up and shipping is. But the proof has to be something a sceptical reader can check — a live pipeline, a merged change, a number that came out the other side of a real system. Anything that only looks like proof is worse than nothing, because it teaches people to discount the rest.
Read next
The month AI wrote 99.77% of my code
In October 2025 AI wrote none of my line changes. In August 2026 it wrote 99.77% of them. The engineering did not get easier — it moved.
The spec is the deliverable now
If an agent writes the implementation, the specification stops being paperwork. It becomes the part of the work with your name on it.