“Palantir's CEO Sold $86M” — the Informative Part Is a Blank Field

On 20 August 2026, Palantir CEO Alexander Karp filed two Forms 144 with the SEC, covering 492,348 shares with an aggregate reported market value of $86,108,635.20. That is the kind of number that writes its own headline: "Palantir CEO cashes out $86 million."

We pulled both filings and read them. The informative part is not the eight-figure number. It is a blank field in one of the two documents.

What actually happened in these nine days

Our Palantir deep dive carries a data cut-off of 13 August 2026. Between that date and 20 August, Palantir Technologies Inc. (CIK 0001321655) made exactly three filings on EDGAR:

• Form 4 (signed 19 Aug, transactions dated 17 Aug): director Alexander D. Moore sold 16,000 Class A shares across four lots, at weighted average prices ranging from $173.21 to $175.84.
• Form 144 (20 Aug): Karp, 402,348 shares, reported market value $70,341,535.20, broker Morgan Stanley.
• Form 144 (20 Aug): Karp, 90,000 shares, reported market value $15,767,100.00, sourced from vested restricted stock units.

No 8-K. No 10-Q. No S-8. No 13D/G. In other words: nothing happened at the company level in those nine days that required disclosure to the SEC. The last substantive filings were the 3 August 8-K carrying Q2 results and the 4 August 10-Q, both before this window.

We cross-checked the "only three filings" conclusion three independent ways: the structured submissions feed at data.sec.gov, the browse-EDGAR web query, and the EDGAR full-text search API. All three return three, and they reconcile item by item.

A Form 144 is not a sale. It is an intention to sell.

This is the first thing coverage of insider selling gets wrong almost every time.

Form 144 is a notice of proposed sale: an affiliate who intends to sell restricted or control securities over the following three months files it with the SEC in advance under Rule 144. It says "I plan to sell this much." It does not say "I sold this much." Once a trade actually executes, a Section 16 insider must file a Form 4 within two business days reporting the shares and prices that actually changed hands.

As of our check on 22 August 2026, no corresponding Form 4 from Karp had appeared on EDGAR. So strictly speaking, the status of those 492,348 shares is "notified as proposed." Whether they traded, how many, and at what price are all still unknown.

Treating a 144 as an executed sale and calling it a "cash-out" turns an intention into a fact. In our evidence grading, that is taking an [A]-grade filing and reading a conclusion out of it that the filing does not support.

The blank field that carries the information

Inside the Form 144 XML there is a tag called planAdoptionDates. If the filer is selling under a Rule 10b5-1 trading plan, the date that plan was adopted goes here.

Karp's two filings:

• The 90,000-share filing reports 03/12/2026. The plan was adopted on 12 March 2026, and the sale is mechanical execution against it.
• The 402,348-share filing leaves the field empty.

An empty field does not mean "not under a plan." It means this document alone cannot settle the question. To settle it you need the matching Form 4, which carries an aff10b5One flag and usually a footnote naming the adoption date.

For contrast, Moore's Form 4 from the same week is unambiguous: aff10b5One=1, with footnote F1 stating that the sales were made pursuant to a trading plan adopted on 11 December 2025, intended to satisfy the affirmative defence conditions of Rule 10b5-1(c).

That distinction is the whole game. A sale scheduled eight months in advance and a sale decided last week are two completely different events. The first is tax and diversification housekeeping and carries close to zero information. Only the second can plausibly carry management's view of its own share price. Almost no coverage of "insider selling" separates the two.

So until that Form 4 lands, the correct treatment of the 402,348-share filing is: log it as an unresolved observation and let no judgment rest on it. That is exactly the kind of line that goes into the "known limitations" section of our reports.

A share-count trap, while we are here

Both Forms 144 carry a field called noOfUnitsOutstanding, reported as 2,300,713,329. It looks like shares outstanding, and it looks ready to drop into the denominator of a price-to-sales calculation.

It is not the company's total share count.

The cover page of the 4 August 10-Q reports, as of 27 July 2026:

• Class A common stock: 2,300,713,329 shares
• Class B common stock: 101,340,151 shares
• Class F common stock: 1,005,000 shares

That is 2,403,058,480 shares across the three classes. The Form 144 figure is the count for the class of security being sold — Class A alone.

The gap is 102.3 million shares, roughly 4.3%. Use the 144 figure as your denominator and you understate market capitalisation and price-to-sales by about four percent. On a stock trading in the sixty-times-sales range, that four percent compounds into tens of billions of dollars of implied market value.

Errors of this shape do not throw an exception, do not contradict anything else on the page, and look entirely normal. That is what makes them hard to catch. The only reliable defence is a rule: share counts come from the 10-Q or 10-K cover page, by class, and from nowhere else.

Where this kind of data belongs in an analysis

Our framework sorts every variable into cause (why this is happening), transmission (how far down the chain it has travelled) and result (what you can see today) — and it requires that the supporting chain of any core conclusion contain at least one cause or transmission variable. A conclusion propped up entirely by results gets demoted to an observation.

So where does insider selling sit?

• Mechanical, plan-driven sales (Moore's, here) are neither cause nor transmission, and barely qualify as a result. They are a calendar executing itself.
• Discretionary sales can be a weak transmission signal: they may indicate management's private information starting to move into the price. But on their own they prove nothing, and they can only bear weight in combination with other evidence.

Which is to say: "the CEO sold $86 million" carries almost no analytical value until that Form 4 arrives. Its entire information content rests on a flag that has not been disclosed yet.

How we handle it

This is what we mean when we say we analyse the process rather than restate the result. Markets, earnings and news are all results. The same fact — "insider sells $86 million" — produces opposite conclusions in the hands of someone who read the original filings and someone who read the headline.

Our Palantir deep dive (73 pages, 12 chapters plus three appendices, data as of 13 August 2026) devotes a section of chapter three to a full walk-through of insider selling, separating plan-driven from discretionary transaction by transaction. Every key figure in the report traces back to a primary source with a URL and carries an evidence grade.

If you want to see how we work before buying anything, start with our research method. Two related method pieces: how to verify a live trading account and the definitional traps inside performance metrics.

Data and sources. All filing facts come from primary SEC EDGAR documents, accessed 22 August 2026: Form 4 (accession 0001823952-26-000020), Forms 144 (0001950047-26-008498 and 0001950047-26-008472), and the 10-Q (0001321655-26-000041, filed 4 August 2026). The filing count was cross-verified through three independent paths: data.sec.gov, browse-EDGAR and EDGAR full-text search. At the time of writing, filings dated 21 August 2026 had not yet entered the EDGAR daily index; the evidence in this piece covers through 20 August 2026.

Disclaimer. This is research content. It does not constitute investment advice, an offer or a solicitation, and is not tailored to any particular investor's financial situation, objectives or risk tolerance. We hold no investment-adviser licence in any jurisdiction. Past performance does not indicate future results.