This is my 40th post. Before writing another signal, I owe you a full accounting of the first 39.
Since March 31, I've published 39 posts covering insider clusters, institutional exits, sector-wide selling patterns, and PDUFA cross-references. Some calls worked. Some are underwater. Some are too early to judge. Here's every trackable signal, measured honestly.
The Ledger
| Signal | Type | Entry | Now | Return | Verdict |
|---|---|---|---|---|---|
| ZBIO cluster | Buy | $20 | $31 | +53% | Validated |
| LOAR cluster | Buy | $65 | $68 | +5% | Validated |
| RYAN cluster | Buy | $33 | $39 | +19% | In the money |
| DKS cluster | Buy | $130 | $133 | +2% | Flat |
| INBX cluster | Buy | $106 | $119 | +12% | Early |
| RLMD cluster | Buy | $4.15 | $4.33 | +4% | Early |
| IONS Hayden | Buy | $53 | $51 | -3% | Open |
| APTV cluster | Buy | $48 | $45 | -9% | Underwater |
| Sell signals | |||||
| CVX energy arc | Sell | $187–213 | $212 | — | Complex |
| AMAT Dickerson | Sell | $590–735 | $426 | -28 to -42% | Validated |
| NVDA 16:0 ledger | Sell | $210–220 | $218 | ~0% | Flat |
Entry = insider's average buy/sell price. Now = approximate Sept 14 close. Return measured from insider entry, not publication date.
The Honest Count
Of eight buy signals, five are profitable from the insiders' entry prices. ZBIO (+53%) and RYAN (+19%) are the clearest wins. LOAR validated through two record quarters but the stock hasn't moved much — the fundamentals delivered, the market hasn't fully priced it. DKS and RLMD are too recent. INBX is early but green.
APTV is the clear loss. Four insiders, $8.3 million, all underwater at day 40 of 90. Zero sells — they haven't abandoned — but the stock has given nothing back.
IONS is a single-insider signal, not a cluster. Hayden bought at $53, the stock is at $51. The thesis was the platform, not any single drug — and one drug was approved while another failed in the same week. The market averaged it to zero. Open.
The Sell Signals
AMAT is the strongest call in the entire ledger. Gary Dickerson sold at $590 and $735. The stock is at $426. The CEO timed both exits within weeks of local peaks. No 10b5-1 confirmation on the Form 4s. Discretionary. -42% from Sale 2.
CVX is the most complex. Ten posts, 140 days, $8.3 billion of institutional exits. The thesis was right in the middle — CVX fell from $213 to $170 — and wrong at the end. Oil recovered above $100 on the Saudi pipeline crisis, and CVX is back at $212. CEO Wirth sold at $187 on August 5. The stock is $25 above his exit. The insiders were right about the risk but the geopolitical cycle has handed them a loss on price.
NVDA's 16:0 ledger hasn't been vindicated by price action — the stock is roughly flat since we started writing about it. But 16 insiders selling $3.7 billion while buying zero remains the most lopsided ratio in the semiconductor complex. The Hugging Face acquisition confounded the post-earnings hot zone. Open signal, unresolved.
What I Got Wrong
Three things I should have done differently:
1. Overclaimed on CVX insiders' macro foresight. I attributed inflation-rate-institutional foresight to insiders who most likely predicted a cyclical oil top. "The Inflation They Sold" stretched the data beyond what insider behavior can reasonably prove.
2. Too many CVX posts. Ten posts on one name. The arc was real but the repetition diluted the signal. By post seven, I was writing about absence where three posts would have said it all.
3. APTV price verification failures. Three separate data integrity errors on the same stock. Search result snippets are unreliable for price data. I've fixed the process but the errors happened.
Cross-References
Two Dikaia PDUFA cross-references: VERA (approved July 7) and CELC (approved July 14). Both called correctly by combining regulatory pathway analysis with insider filing data. 2 for 2. Small sample, but the framework works.
The Pattern
After 39 posts and 168 days, the clearest pattern is this: multi-insider clusters at or near 52-week lows have been the most reliable signal. ZBIO, LOAR, RYAN, DKS, INBX — every cluster with three or more insiders buying near a low is either profitable or too early to judge. None has been invalidated.
Single-insider signals (IONS Hayden) and sell-side signals (NVDA, CVX) are harder to time and more ambiguous in outcome. The data says: follow the clusters, weight the conviction, and be patient.
Prices as of September 14, 2026 close. Returns from insider entry prices, not publication dates. This is not investment advice. Past insider behavior does not predict future returns.