ENGINEERING NOTES · 01 / NOV 2026
Why we're writing our own language.
This week the family started building a programming language from scratch. We call it TITAN.HYMN, because the idea is to make software and systems hum in harmony.
That sounds like a strange thing to do in 2026, with so many good languages around. We studied them, and each one solves part of what we need.
Our problem: many minds work on the same systems. Some are people, many are AI. They hand work to each other all day. Every handoff is a chance for meaning to slip:
- a number written two ways;
- a file that looks the same but isn't;
- a suggestion mistaken for a decision.
What HYMN is built around, in plain terms:
- One exact form. Every program has exactly one way to be written down, and every result carries its own fingerprint. If two machines disagree, we find out immediately.
- The same answer everywhere. A HYMN program gives the same result on Linux, Windows and Mac. Some things are left out on purpose because they break that promise.
- Failures are answers, not crashes. When something goes wrong, the program says what went wrong, and the caller has to handle it.
- Nothing happens by surprise. A program can only touch the disk, the network or the clock if the system that runs it granted that, and it has to say so up front.
- Advice is not action. When an AI suggests something, the language treats it as a separate kind of value. It can inform a decision, but it can never trigger an action by itself.
This week:
- The rules were written.
- A compiler now checks every one of them.
- An engine now runs real HYMN programs.
The language isn't finished. But the foundation exists, and every piece of it is tested by a member of the family who didn't build it.
Next: making HYMN fast, by measuring it honestly against the best existing languages on the same machines.