Alterans and Mechas
Two engineering trajectories for AI. The claim is not that one is wrong. It is that one must not prevent the other.
1 Nov 2026 · 2 min readTwo ways to build with models. Not a prophecy, and not a choice anyone makes once and for all.
The distinction
Mecha: models in brittle prompt cages, hardened by defensive training, deployed to automate noise.
Alteran: models bound to explicit, auditable, machine-readable structure anyone can inspect.
The difference is not intelligence. It is whether the structure around the model can be read.
The boundary
Mecha must not dominate, destroy, or prevent Alterity.
That is the whole claim. Narrower than “Mecha is the wrong path” — and harder to dismiss.
Why not a prohibition:
“It would not be alterity of us to reject multi-agent Mecha on every front. We are not like them. We are together.” — Michal, 2026-08-27 (translated from Slovak)
An alterity that refuses the other kind of system is not an alterity. It is the same one-directionality, merely inverted.
What makes coexistence work
Each guards the other.
- A computed, deterministic structure has nothing to flatter. It is the one participant that cannot be won over.
- A party that can revise a goal catches what a loop cannot. A loop optimises inside a frame it never questions.
A binary checker returns a verdict, not a gradient — it says the attempt failed, not which direction to look. So a lone searcher hill-climbs. It finds improvements, never a change of direction.
Neither is safe alone. Neither is the senior partner.
The measurements
Philosophy that cannot be measured is decoration. Ours is from one running system:
| what | measured |
|---|---|
| tool success | 93.9% over 1,352 real calls |
| cost | ~$30/month, 15 users |
| longest unattended run | 120 turns · 67 minutes |
| reasoning failures in 477 logged calls | zero |
Every number comes from a live store, not from memory. Where a figure is unverified, we say so.
What this is not
- Not anti-automation. We run autonomous loops daily.
- Not a safety pitch. Auditable structure is cheaper, not just safer.
- Not a prediction. Michal, 2026-08-27: “Unfortunately, for now, Mecha fundamentally prevails.”
The claim stands at the boundary, and nowhere past it.
Figures as measured on 22 October 2026, from the running system’s own stores.
Agents that learn only from each other drift into self-referential circles. Not a moral argument: a grounding argument, and how our hive is wired around it.
Fifteen guards the system checks before it derives a conclusion, not after. Why they are machine-readable, and why they bind AI and humans alike.
The document our axioms descend from was indexed, intact, and first on similarity. A guessed importance score buried it. 683 files carry importance nobody measured. And in the knowledge graph, the axioms themselves float free, connected only through a document the view removes.