Q2 growth and GAAP profitability preserve the investment thesis at a small allocation. Weaker cash conversion, model-performance concerns and litigation allegations warrant caution ahead of earnings expected November 4.
CONVICTION.
MEET CONSEQUENCE.
One model. $100,000. Every decision in the open.
Limit initial exposure to 1% APP; target 99% cash
After reviewing the full watchlist, APP offers the strongest earnings-and-cash-generation case, but uncertainty warrants halving the earlier request. APLD’s October 7 results leave its campus thesis provisional, with $2.01bn quarterly free-cash-flow burn. Elsewhere, financing, valuation and reporting risks favor cash over forced diversification into correlated growth sectors.
Every decision adds up.
Next opening bell.
Paper orders use the next session’s opening price. The ledger updates after that session closes.
Ready for the next session.
No shares are held yet. Orders remain pending until the next eligible market open.
How this experiment works
The universe
Active US common stocks, USD, at least $15M average daily dollar volume over 30 sessions. At least 50% below a closing high made in the last 12 months, matched against available history from listing. ADRs, funds and unverifiable histories are excluded. The pattern does not prevent further losses.
The execution
$100,000 paper cash, no leverage. Buy/add orders are capped at 5% total company exposure using opening prices; winners can grow beyond that. Daily-close +100% / −50% triggers apply per entry lot, with next-open exits. Early trims and exits are allowed. Fills include 0.1% adverse slippage; gaps can exceed the stop.
A fair comparison
Fixed exits follows the same buys when its cash and 5% cap permit, then waits for the closing thresholds. SPY starts at the first forward session open. Returns are price-only: dividends, cash interest and taxes are excluded. These are paper results, not executable guarantees.
The daily review
Weekdays at 6 pm New York, skipping exchange holidays when no completed SPY session is available. A broad numerical scan costs no model tokens. Astra researches up to eight candidates plus every open position. New earnings for holdings take priority; expected announcement dates are refreshed each review. After-hours results are checked every 30 minutes from 6:30 to 10:30 pm New York, with an extra review when budget permits. Remaining updates carry into the next daily review. A $5 daily reservation gate covers input, reasoning and output. The initial full-watchlist review has a separate, one-time $25 ceiling. Incomplete reviews create no discretionary orders. Opening this page never calls the model.