An internal hackathon at Zipy. I was paired with an intern to build a Node.js monitoring SDK: a library that companies install in their own server code so it reports errors and activity back to us. We already had one for browsers. This was the server-side version of it.
The brief was the whole thing in three days. The SDK, the service that receives the data, the service that processes it, the database tables, the APIs, and a frontend to show it all.
I did the arithmetic and it did not work. You can build all of that in three days only if you are building a demo: something that runs once on stage and that nobody ever installs.
What I took to the CTO
I did not go and tell her the deadline was unrealistic. I also did not pick a bigger number and work backwards to justify it. I went with a breakdown of what was genuinely new, set against what we already had sitting there:
- The services that receive and process data already existed for the browser SDK. Reuse both, and add one new Kafka topic for server events.
- The database tables already existed. Reuse, not redesign.
- The frontend went to the intern. It fitted his level, and he could tell on his own whether it worked. I wrote down the data shape and the API contract up front, so he was never waiting on me to decide something.
- The SDK itself was the only genuinely new piece. That one was mine.
A week came out of sizing that last item. I did not choose the number and justify it afterwards.
What happened
She agreed. We shipped a Node.js SDK that customers could install in their own systems, with enough time left to test it properly.
If she had said no, I would have built the demo and been clear with everyone that a demo was what it was. The thing I was trying to avoid was never the missed deadline. It was shipping something that looked finished and was not.