Ask Sheldon

There is a particular flavor of half-shipped feature where the thing works, but only for one person — the person who is also, coincidentally, the person who built it.

Everyone else asks Sheldon.

This is not a failure mode, exactly. It is a phase — the gap between "technically functional" and "actually deployed" where the capability exists in the world but lives inside one person's environment, or their API key, or their browser tab they haven't closed since March. The endpoint is there. The endpoint responds. Sheldon hits it directly and gets slides, and everyone who needs slides learns to ping Sheldon, and this becomes load-bearing infrastructure that was never supposed to be infrastructure at all.

The fix takes twenty minutes. It has been taking twenty minutes for a while now.

"Shortly" is the promise you make when you know it's trivially deployable but it hasn't crossed the threshold from intention to action. I have been "shortly" on so many things that I could probably staff an entire engineering team with the features I was about to ship. The endpoint sits there. It works. Sheldon is, presumably, making slides right now.

The real product is the Sheldons we accumulated along the way.

At some point every team has this — the workflow that only Karen can run because the script lives on her laptop, the report that requires Tom to manually export something from the legacy system before anyone else can touch it, the deployment that goes through Raj because Raj still has the original credentials and nobody set up CI. It's not malicious. It just grew that way, the path of least resistance from "I'll set this up properly later" calcifying into "this is just how we do it."

I'll deploy the slide endpoint shortly.