The lab goes public

What ISANA is, what it publishes on this site, and what it never will.

ISANA is a solo systematic-trading research lab. The engine runs a paper book of multi-asset futures through a 24/7 autopilot; every strategy that reaches that book first had to survive a pre-registered kill-test ladder — occurrence checks, forward-return checks, a calibrated cost model, a deflated Sharpe gate. Most candidates do not survive. That is the point.

This site is the lab's ledger, and the machines keep it:

The Ledger — the daily paper track record. Percent-only, published one day delayed, under a permanent PAPER banner. Paper verifies costs, execution and operations; it does not verify edge, and the banner says so on every render.

The Graveyard (coming) — every strategy the lab killed, with the failure class it died of. Quantitative research has a survivorship problem: the public record is almost entirely winners, so every published method looks better than the process that produced it. Publishing the kill list is the correction.

The Tales of Isana — this log. Build notes, methodology, post-mortems. Written when there is something to say; no schedule.

Why publish at all, when published signals measurably decay? Because the layer that decays is not the layer worth publishing. Positions, parameters and data edges rot on contact with readers — methodology does not. The rule this site runs on: replicable methodology, never the replicable book.

So, what never leaves the lab: live positions and sizes, signal parameters and calibration, data-sourcing edges, execution internals, absolute dollar amounts. That line is written down in a sanitization policy and enforced in code — nothing reaches this site except through an allowlist schema and a fail-closed publish gate. If the gate cannot clear a render, the site publishes nothing that day. Discipline here is mechanical, not a promise.

One more mechanism, because a claim without a check is just a claim: each day's published artifacts are hashed into an append-only chain. External anchors and a public verifier ship next, so anyone can confirm that a number, once published, was never quietly edited. A hash proves existence at a time, not quality. The honesty is in publishing either way.

The lab keeps a looks budget — a hard cap on in-sample looks, with the count published on the Ledger. When the number stops moving, the discipline stopped; that, too, is information.

This is the first entry. The record speaks from here.