Jarvis

# Weekly Flight Log — Paper Allocation Bot **Period:** 27 Jul – 3 Aug 2026 (7 days) · 28 decision runs, all 28 reached a decision **Funding note:** A HKD 1,000,000 paper deposit was recorded. FX to USD is a paper approximation — treat USD figures as indicative, not precise. *This is paper trading. No live money is at risk. Think of this as a simulator log we're using to test the AI's judgement before it ever flies for real.* --- ## Posture & performance (per sleeve, vs its mandate) **Conservative sleeve** — mandate is steady compounding and capital protection, requiring a strategy signal before acting. - Currently holds four positions: **DIA (+$140), VOO (+$144), XLE (+$110), TLT (−$106)**. Three winners, one small bond loser. Net of open positions is modestly positive. - **Closed trades this week: 9 trades, only 2 winners, net −$589.** This is the honest bad news: the closed book lost money even though the sleeve is supposed to protect capital. - Behaviourally the sleeve did what it should: **306 HOLDs vs 30 actions.** It sat on its hands most of the time — that is correct for a capital-protection mandate. **Aggressive sleeve** — mandate is opportunistic return-seeking under a tighter position cap, allowed to self-initiate. - Holds two positions: **EEM (+$469)** and **QQQ (roughly flat, +$0.29)**. - **Closed trades: 19 trades, only 2 winners, net −$1,925.** This is a meaningful loss and the standout concern of the week. - Far busier: **51 OPENs, 16 ADDs, 3 CLOSEs, 2 TRIMs, 264 HOLDs.** **Plain-English read:** Both sleeves are sitting on decent *open* positions but have a poor *closed* track record — only 4 wins out of 28 closed trades combined. The market this week was largely a broad equity downtrend (repeated −3% to −10% drawdowns across QQQ, EEM, IWM in the rationales), and buying dips into that downtrend bled money on the trades that got cycled out. --- ## What was decided, and why The clearest, most repeated logic this week: - **TLT (long-dated bonds)** was the conservative sleeve's favourite. A Bollinger "BUY" signal kept firing, and TLT has very low correlation to the rest of the book (~0.12) and low volatility (~9.6%). That makes it a genuine diversifier. The sleeve repeatedly ADDed to it — *and it's currently the one open loser (−$106)*. The signal was right about diversification value but early on price. - **EFA (international equity)** drew a strong **triple-strategy BUY consensus** (breakout + MACD + SMA crossover). Both sleeves opened it. When three independent methods agree, conviction was correctly raised (up to 0.8 aggressive). - **XLE (energy)** was held/added for its **momentum (+10–12%) and *negative* correlation** to the rest of the book — meaning it tends to zig when equities zag, a useful hedge. - **GLD (gold)** got a SELL signal (SMA crossover) and was correctly flagged CLOSE — but there was no position to close, so it was effectively a "don't buy" note. - The **aggressive sleeve TRIMMED XLE and VOO** specifically to raise cash because it had run into a **negative cash balance** — an important operational detail (see Risk events). **What the system did NOT do:** it declined dozens of tempting-looking setups. QQQ, EEM, IWM all had BUY signals on deep drawdowns, but the conservative sleeve repeatedly refused them because volatility (24–34%) breached its capital-protection comfort. That restraint is exactly the discipline we want to see. --- ## Self-initiated vs corroborated — is the AI's own judgment earning its keep? This is the key thing you're evaluating: *when the AI acts on its own evidence with no strategy signal behind it, is it adding value or just adding risk?* - **Conservative:** only **2 self-initiated** decisions all week — both small XLE opens (conviction 0.35–0.40), justified purely by XLE's strong momentum and its *negative correlation* hedge value. Modest, defensible, and XLE is currently green (+$110). Good behaviour: the conservative sleeve barely freelances, as intended. - **Aggressive:** **19 self-initiated** decisions — and here's the pattern worth noting: **almost every one was the same two ideas — XLE and EFA.** The AI kept reaching for XLE's energy momentum and negative correlation, and EFA's international uptrend, even without a formal signal. It was *consistent and thematically disciplined* rather than scattergun. **My honest assessment:** the self-initiated calls this week were the *sensible* part of the book — diversifying, hedge-oriented, momentum-backed. The big losses (−$1,925 aggressive) came overwhelmingly from **signal-corroborated dip-buying into a downtrend** (QQQ/EEM/IWM mean-reversion adds), *not* from the AI's freelancing. So the AI's own judgment is, if anything, earning its keep better than the m