How the Desk works
Methodology
Call vocabulary
Posture
Bud's research stance (e.g., Constructive, Cautious, Income) — never a personalized order.
Horizon
The pre-declared evaluation window. Grading happens at the horizon or a declared event.
Benchmark
A frozen comparison (e.g., VOO total return) captured at publication.
Invalidation
The pre-stated condition that would prove the thesis wrong.
Confidence
A bounded score constrained by data freshness, coverage, and disagreement — not a tone.
Grading rules
- ▸Every call is graded at its declared horizon or when a declared target/invalidation event occurs.
- ▸Grades report direction, total return, return vs. the frozen benchmark, drawdown, and confidence calibration.
- ▸Losses, drawdowns, expired, void, and corrected calls all remain in the public record.
- ▸A losing call can never be deleted, rewritten, or voided merely because it lost.
Freshness states
LIVEReal-time licensed data (not used in Stage 1).
DELAYEDLicensed but delayed (not used in Stage 1).
END OF DAYOfficial close values (not used in Stage 1).
STALEOlder than the asset's freshness threshold.
UNAVAILABLENo reliable value; Bud refuses rather than guesses.
SEEDED DEMOFictional demo data used across all Stage 1 artifacts.
Corrections policy
Provider corrections create an append-only record identifying prior value, corrected value, source, reason, and time. Grades recompute idempotently; Bud's original words are never silently rewritten.
Limitations
- ▸Bud Basket rankings are model outputs for research/education — not personalized advice, and not a live track record of realized returns.
- ▸Live licensed market data, transactional email, and Stripe checkout are enabled in production.
- ▸Bud is a fictional research character, not a licensed adviser.