Public ledger · settled by real prices
AI watches.
Paper proves. You decide.
Most AI stock tools hide their misses. We publish every single one.
A public, append-only ledger of AI stock judgments — settled by real prices, not marketing. Five trading days per call, misses stay on the record.
Judgments
75
llm-v1 public ledger
Settled / pending
1 / 74
as of 2026-11-05
Universe
165 tickers
fixed, 5-trading-day horizon
Hit threshold
±1.0%
preregistered · 5 trading days
Why we publish the misses
Every AI tool claims to be good. Almost none show their failures. A record that only contains wins is marketing, not evidence — so ours contains everything, settled by prices we don't control, on a schedule we don't control. That is the whole product: not signals, receipts.
How doxt compares to typical AI stock pickers
Typical AI stock picksDoxt Stock
Shows winning screenshotsPublishes every judgment, misses included
"Trust our model" (black box)Rules in public, settled by code
Promises returnsGuarantees nothing but the record
Records can be quietly deletedAppend-only — edits are rejected by the database
Questions people actually ask
Is there an AI stock tool with a public track record?
Yes. Doxt Stock is exactly that: a public, append-only ledger of AI stock judgments. Every call — direction, confidence, evidence, counter-argument — is published before it settles, then graded against real market prices 5 trading days later. Misses are never removed.
Yes. Doxt Stock is exactly that: a public, append-only ledger of AI stock judgments. Every call — direction, confidence, evidence, counter-argument — is published before it settles, then graded against real market prices 5 trading days later. Misses are never removed.
Is doxt Stock financial advice?
No. Doxt Stock is a research and record-keeping tool. It never places orders, manages money, or promises returns. The only thing it guarantees is that the record is real and cannot be edited.
No. Doxt Stock is a research and record-keeping tool. It never places orders, manages money, or promises returns. The only thing it guarantees is that the record is real and cannot be edited.
What happens when the AI is wrong?
The miss stays on the public ledger permanently. Hit rate, coverage and calibration are computed from the raw append-only record by deterministic code — not curated by us.
The miss stays on the public ledger permanently. Hit rate, coverage and calibration are computed from the raw append-only record by deterministic code — not curated by us.
How are judgments settled?
A judgment made after market close on day T enters at the next trading day's open and settles at the close of the 5th trading day. Splits and dividends are adjusted via adjusted prices. The simulated paper account includes commissions and slippage.
A judgment made after market close on day T enters at the next trading day's open and settles at the close of the 5th trading day. Splits and dividends are adjusted via adjusted prices. The simulated paper account includes commissions and slippage.
Can I verify the AI stock calls independently?
Yes — that is the point. Judgments are hash-chained and append-only at the database level, settlement thresholds are preregistered (±1.0% over 5 trading days), and baselines (always-bullish, buy & hold, cash) are computed on the same window. The public ledger at /ledger is the raw evidence.
Yes — that is the point. Judgments are hash-chained and append-only at the database level, settlement thresholds are preregistered (±1.0% over 5 trading days), and baselines (always-bullish, buy & hold, cash) are computed on the same window. The public ledger at /ledger is the raw evidence.
What does coverage mean?
Coverage is the share of judgments that commit to a direction instead of abstaining. It is published next to hit rate because a model that rarely commits can look good by staying silent.
Coverage is the share of judgments that commit to a direction instead of abstaining. It is published next to hit rate because a model that rarely commits can look good by staying silent.
The interactive app is loading below.