The call stays visible after the outcome arrives.
A permanent record of what I expected, what happened and what changed. Wins, misses and revisions remain together so hindsight cannot quietly rewrite the setup.
THE METHOD
One format. Four checkpoints.
The score belongs to the published claim—not to the later share-price move or portfolio return.
- 01Publish the setup
State the question and mechanism before the result.
- 02Define the evidence
Record what would confirm, weaken or reject it.
- 03Observe the outcome
Separate business evidence from the market reaction.
- 04Score the call
Keep the original claim and the revised model together.
SCORING KEY
What each label means.
- CONFIRMED
- Evidence matched both the direction and the mechanism.
- MIXED
- Some expected evidence arrived; material parts did not.
- REFINED
- The evidence improved or narrowed the original model.
- WRONG
- The observed outcome rejected the published call.
- OPEN
- There is not enough evidence to score the thesis yet.
PUBLIC SCORECARDS
The record, including the misses.
05 OPEN
24–30 AUG 2026
Is the August portfolio genuinely diversified, or is AI infrastructure one common factor wearing different labels?
The first look-through already weakened the sector-count story: SPY, direct platforms, hardware and physical infrastructure share exposure to the same AI capital-spending cycle. Nvidia then confirmed that demand remains strong, but one positive report is not a downside stress test.
Exposure map refined; diversification verdict still open. 04 REFINED
17–20 AUG 2026
Were consumers broadly pulling back, or was July simply a soft month?
The five retailers did not support a clean strong-or-broken answer. Housing demand stayed uneven, discretionary results diverged and Walmart showed slower comparable-sales growth alongside strong digital and operating-profit growth. Consumers were still spending, but more selectively and through different channels.
The binary frame was too crude: demand was category-, value- and channel-dependent. 03 WRONG
10–12 AUG 2026
Would July’s oil spike create a hot backward-looking CPI print?
No. CPI came in broadly as expected and the seasonally adjusted gasoline index fell 2.9%. The crude-to-pump transmission was slower and more complicated than the directional thesis allowed.
Wrong forecast; zero portfolio cost because no trade depended on it. 02 REFINED
27–31 JUL 2026
Would the market reject the next wave of AI capital spending?
Microsoft and Amazon were rewarded despite enormous spending. Meta and Apple fell. The amount spent was not the dividing line; visible revenue, margins and cash conversion were.
The original rule was too broad: spending needs receipts. 01 MIXED
20–23 JUL 2026
Would the semiconductor damage spread into the broader AI trade?
The pressure spread to a hyperscaler, but not because AI demand failed. Alphabet passed the operating test while capital intensity and free cash flow drove the repricing.
Direction partly right; mechanism refined.