Generated here, given to one person, good once. The code is random — a chosen code is a guessable one. Note is for who it went to; a list of codes with no record of that is unusable within a week.

% off
Applies to
Either plan Individual only Practice only Expires
Note — who is this for?
Generate

Funds/ETFs a real portfolio held that the app had no look-through data for — the demand side, versus the security-master panels below which show what's actually been learned. Status reflects the shared corpus, not merely whether a resolve was cached: In corpus = learned and served · Resolved, not yet published = resolve worked, will publish next time someone holds it (no action needed) · Rejected = the quality gate refused it, reason shown · Needs review = no automated path works, so this is the list worth a manual HOLDINGS_DB entry.

Individual holdings the auto-resolve pipeline identified inside a fund's SEC filing but couldn't identify via OpenFIGI at all (as opposed to identified-but-not-yet-sector-classified, which self-heals on its own) — the genuine residual tail worth a periodic look.

When a fund nobody has seen before turns up in a portfolio, it is resolved from SEC filings once, checked by a quality gate, then stored so every later visitor gets it instantly instead of re-resolving. This is where you confirm that loop is actually running: what it has learned, when, and what the gate rejected. An empty table here is normal until someone uploads a portfolio holding an unrecognised fund.

Green = accepted into the shared corpus · Red = rejected by the quality gate

Companies the app had never seen before, discovered inside a fund's holdings and classified automatically. These become available to every user.

Opening a stored portfolio, opening an account's report, changing an account's type, name, firm, plan or plan date, deleting an account (and a by-hand Stripe confirmation before a deletion), turning the payment switch on or off, and administrator one-time-code attempts, accepted or refused. Not recorded: the accounts list, an account's profile and tab history, who used a given tab, discount codes, and edits to a phone number or how-heard answer. An account change is logged after it is saved, so a failed log write leaves the change with no row. An adviser's report opened before a client is chosen gets a row naming no client, though their client list was shown. The privacy policy promises this record exists, so it is the one page that has to be able to show it. Reading it is not itself recorded.

What the scheduled jobs write about themselves — the macro baseline refresh and the trailing-returns refresh. They are kept out of every member figure in this console, which is why they need a panel of their own: a feed that fails behind a working fallback is invisible otherwise, because the baked constants cover for it perfectly. A run that got everything writes nothing here.

Raw append-only record of every publish attempt. Failures are persisted here rather than logged to a console, because a Worker's console output isn't durable.