How we prove it
A track record you don’t have to take on trust.
Most track records are shown after the fact, by the people they flatter. Ours is built so it can’t be edited quietly and can’t be cherry-picked. Here is exactly how, and how to check it yourself.
There are two kinds of record. Ideas are research: the desk holds no position in the security when an idea is published. Desk calls are the desk’s own trades: they are sealed on the ledger and revealed at close.
1. Published ideas are sealed at publication
When we publish an idea, we hash it with SHA-256 and publish the hash with it, on the site and in the newsletter. The hash covers everything that defines the call: number, title, publication time, security, view (long or short), horizon, reference price, benchmark, the desk’s position at publication (always none), the falsifiers, and the full text.
Precisely, the sealed record is this JSON string, with the keys in this order, encoded as UTF-8:
{"number":…,"title":…,"published_at":"YYYY-MM-DDTHH:MM:SS.sssZ","ticker":…,
"view":…,"horizon":…,"reference_price":…,"benchmark":…,"desk_position_at_publication":"none",
"falsifiers":[…],"body":"<the idea text, trimmed>"}hash = SHA-256( UTF-8 bytes of that string )Every idea page shows the exact sealed record and has a Verify this hash button that recomputes it in your browser. Change one character, a price, a date or a word, and the hash no longer matches. The site’s own build refuses to publish an idea that was changed after it was sealed. Corrections are added as dated updates below the idea, outside the sealed record, so you can see what changed and when.
Results (close date, return, benchmark return) are added when the idea closes. They are not sealed, because they don’t exist yet at publication, but they can be checked against public prices for the stated dates.
2. Every desk call is on the sealed ledger
Ideas are research; the desk’s own trades are a separate record. Every call the desk makes is logged on theledger as a salted hash at the moment it commits, with no ticker, direction or text, so an open position is never disclosed. When the position closes, we reveal the thesis and the salt. Every closed position is revealed, including losses.
hash = SHA-256( thesis_json_bytes || salt_bytes )thesis_json_bytesis the exact thesis JSON string, encoded as UTF-8.salt_bytesare 32 random bytes, published hex-encoded (64 characters) at reveal.||means the salt bytes are appended directly after the thesis bytes.
The salt is why a locked hash gives nothing away: without it, someone could hash likely guesses and compare.
3. Everything is scored at close
Every published idea is closed and scored against its benchmark on the track record, losses included. Nothing is deleted. When we were wrong, the idea says so.
Check it yourself
You don’t need to trust our Verify button. In Python 3, for a published idea, paste its sealed record:
import hashlib
record = r'''PASTE THE SEALED RECORD HERE'''
print(hashlib.sha256(record.encode("utf-8")).hexdigest())For a revealed ledger entry, paste the thesis and the salt:
import hashlib
thesis = r'''PASTE THE THESIS JSON STRING HERE'''
salt = "PASTE THE 64-CHARACTER SALT HERE"
print(hashlib.sha256(thesis.encode("utf-8") + bytes.fromhex(salt)).hexdigest())To be sure a hash itself wasn’t swapped later, compare it with a copy published outside this site at the time: the newsletter email, our social posts, or a web archive snapshot.
What we never publish
Position sizes, open positions (until they close and are revealed on the ledger), thresholds, analog sets, prompts, the desk’s code, or details of its infrastructure.
See also the disclosure policy.