joeycastillo.us / research

research

Decisions about tech, made deliberately.

Every external tool, library, service, or framework that crosses the path of Pantheon or kastil-systems gets the same examination. We don't drift into adoption; we don't reflexively roll our own. We look, we ask four questions, and we decide.

This page is the index. Each entry below is a candidate technology examined under four lenses, with a verdict and the reasoning behind it.

The Four Lenses

01
Integrate

Wire it in as-is. Adopt the tool whole, work with its shape, accept its trade-offs.

What changes if we adopt it without modification?

02
Replace

Find or build an alternative. The candidate is solving a real problem; the candidate may not be the right shape to solve it.

What does our version look like, and what does that cost?

03
Reverse-engineer

Use it as a source. Study the internals, the architecture, the trade-offs they made — then build clean.

What can we learn from it that we couldn't learn ourselves in the same time?

04
Dovetail

Connect it to Pantheon / kastil-systems. Find the shape where the candidate joins the spine without distorting either side.

Where does it join? What contract does it speak? What does the join look like in code?

The Meta-Question

Roll our own?

Every candidate either gets built ourselves or comes from outside. The answer is per-candidate, per-domain. The bias is toward bare metal — but the bias is not the rule. The rule is: decide on purpose.

Entries

2026-05-17 OpenClaw — open-source personal AI agent that lives in your chat apps; reverse-engineer for Pantheon's consumer surface (the shell from them, the soul from us) Reverse-engineer