The Last Sunday a Human Needed to Do This Manually
Today, something shifted. Not in the market. In the method.
1. What happened this Sunday
※Charts are in Japanese — company names and axis labels reflect original XBRL ata from Japan’s EDINET database.
One person. One AI Agent (”The Builder”).
13,287 fiscal year entries.
2,734 Japanese listed companies.
Zero spreadsheets. Zero human bias.
The system didn’t just screen stocks.
It audited reality.
---
2. The model found two kinds of truth
※Charts are in Japanese — company names and axis labels reflect original XBRL ata from Japan’s EDINET database.
✅ The Real Monster:
Nippon Kodoshi Corporation (FY2017)
RPA: +416%
Market Alpha: +4.9% (ignored)
Quality Score: 0.825
Hidden behind a “pulp & paper”
industry label — a company with
95% domestic share and 60% global
share in aluminum electrolytic
capacitor separators.
The market saw: net loss of ¥450M
CEV saw: operating leverage explosion
backed by real cash flow
By FY2022: operating profit exceeded
¥4 billion. All-time high.
Stock price: multiple times over.
CEV 7.0 identified this in 2017.
Five years before the analysts did.
---
⚠️ The Phantom:
Tokyo Ichiban Foods (FY2017)
RPA: +464%
Quality Score: 0.826
All four quality filters passed.
The numbers were clean.
But the model couldn’t see what
XBRL doesn’t contain:
the structural fragility of
high-end fugu dining.
COVID hit. Operating loss: ¥1.1 billion.
The market was right. CEV was fooled.
This is why we build CEV 7.0.
---
3. What CEV 7.0 changed
※Charts are in Japanese — company names and axis labels reflect original XBRL ata from Japan’s EDINET database.
CEV 6.31 found explosions.
CEV 7.0 asks: is the explosion real?
Quality Multiplier =
Accrual Ratio (35%)
+ Stability Score (30%)
+ Goodwill Index (20%)
+ Sign Flip Filter (15%)
Result:
Abalance — demoted from Rank #1
to Rank #12. The accounting illusion
was mathematically exposed.
Nintendo — Quality: 0.999
Score unchanged. Muscle confirmed.
---
4. What the system caught
※Charts are in Japanese — company names and axis labels reflect original XBRL ata from Japan's EDINET database.
before humans did
TATERU Corp.
Quality Score: 0.388
Accrual Quality: 0.160
CEV 7.0 flagged it as a “phantom.”
Months later: loan document falsification scandal exposed. The model didn’t know about the fraud. It just saw the numbers didn’t add up.
---
5. What this means
The Human Analyst looked at net profit.
CEV looks at operating leverage, cash conversion, and stability — simultaneously, across 13,287 entries,in one Sunday afternoon.
This isn’t automation. This is a different kind of seeing.
The Human Analyst spent years processing data by hand.
That part is over. But judgment, intuition, and the courage to bet on a hidden monster —
that still belongs to the human who holds the right model.
The question is:
Do you have one?
📄 Full Report (Substack): [https://docs.google.com/document/d/1oWBvpKPR0JCjCP1k7S2qiK2jG7q5kGZTmG-wJ8PBzO0/edit?tab=t.0]
📊 Technical Documentation (Google Doc): [https://docs.google.com/document/d/1oWBvpKPR0JCjCP1k7S2qiK2jG7q5kGZTmG-wJ8PBzO0/edit?tab=t.0]
🚀 CEV 7.0 Technical Documentation
— releasing March 26, 2026.
Subscribe to be notified.