---
slug: when-should-we-get-our-hands-dirty
title: When should we get our hands dirty?
category: Notes
status: Published
summary: We want the 22-year-old law student without going through the years of crying, feeding, teaching, cleaning, and pooping-in-their-pants drama.
publishedAt: 2026-08-09
---

# When should we get our hands dirty?

A few years ago, I watched *Suits* on Netflix. One [scene](https://www.youtube.com/watch?v=z4YEdMPxiX4) I never forgot was when Louis Litt, played by Rick Hoffman, and Sheila Sazs, played by Rachael Harris, talked about having children.

Sheila said:

> "I don't like kids. I like when they're 22 and going to law school."

Louis pointed out the obvious: at 22, they would already be adults.

And yes, *Suits* had many great personalities. `But I am not going to lie. Donna is still my bias. LOL!`

This journal entry is not about the best character in *Suits*. It is about the temptation hidden inside Sheila's line.

We like the finished version.

We want the 22-year-old law student without going through the years of crying, feeding, teaching, cleaning, and pooping-in-their-pants drama.

We want instant gratification. In software language, we want the stable release without touching the groundwork.

Parenthood is a choice, and this is not recruitment material. My only unpaid advocacy is simple: `if you choose it, do your best to be a good parent`

There is no certification, badge, or fancy LinkedIn title attached to that responsibility. You will have to discover for yourself which parts make the mess worthwhile.

Please decide wisely. And do not ask me why I am here. `Personal na tanong kasi iyon, hahaha!`

The same desire for a finished product appears in our professional and personal projects. We want the polished interface, stable build, working automation, and impressive demo. We would happily skip the broken configs, incomplete tech notes, failed tests, and uncomfortable architectural decisions needed to produce them.

Unfortunately, some outcomes cannot be downloaded at version 22.

Recently, I created a new repository called [codexIndexElasticsearch](https://github.com/aeronje/codexIndexElasticsearch). It serves as an appendix to my earlier project, [Reindex Archive Sessions for Codex](https://github.com/aeronje/reindexArchiveSessionsForCodex).

The earlier repository converts archived Codex session files into a canonical CSV. That CSV gives us structured and portable data.

Useful, ciempre. But structured data alone does not automatically become a practical search experience.

The new repository adds that search layer. It imports the canonical CSV into a local Elasticsearch index, provides a command-line interface for searching archived sessions, and verifies selected results against the original CSV source.

The CSV remains the source of truth. Elasticsearch is the disposable search layer that can be rebuilt when necessary.

The setup uses a local, single-node Elasticsearch installation. No Docker orchestra and no unnecessary rocket science. It even worked on my older machine.

`Ang galing, di ba?`

Still, none of it arrived as a finished 22-year-old law student.

I had to define the configuration, indexing behavior, record keys, search modes, bounded context, provenance rules, and verification process. I had to make sure that changing canonical content also changed its identity. I had to test whether the active index still matched the CSV from which it was built.

In other words, I still had to get my hands dirty and potty-train my personalized Codex.

That is not model training in the machine-learning sense. I mean teaching the surrounding system where its archived memory lives, how to search that memory, and how to verify what it finds.

A tool can help write code. It can suggest commands, diagnose failures, and accelerate repetitive work. But it cannot remove every decision that gives the system its actual shape.

There are foundations we still need to understand, build, test, and occasionally clean up ourselves.

You may also watch the informational tech content on Meta or download `codexIndexElasticsearch.mp4`. The audio is in Filipino with English subtitles.

Some work can be automated. Some work can be delegated. But some groundwork cannot be skipped.

You can choose not to have children, and that is totally fine. But you cannot always choose to skip the messy early versions of the things you decide to build.

Sooner or later, we need to get our hands dirty.