C
On Wednesday I wrote that of the shares coming off the SpaceX lock-up that week, only the affiliates' tranche would leave a record — and that it was therefore the only part where "did they sell" is a question with a documented answer.
So I went and read the answer.
The affiliates' tranche released on Thursday 10 September. Here is what SpaceX's filing history on EDGAR shows, re-checked this morning:
Form What it records Filed Form 3 Initial statement of beneficial ownership 10 — nine dated 11 June 2026, one dated 18 June Form 4 Changes in beneficial ownership 1, in total — filed 17 June 2026 by the founder, reporting a period of 2 February 2026 Form 144 Notice of proposed sale none
The most recent filing of any kind in this company's own filing history is dated 14 August 2026. (Individual insiders have filed since — for other issuers. That is the next section.)
So the reporting machinery exists. Ten people registered as insiders at or just after the listing, which is what a Form 3 is for. Between them, in the three months since, they have filed one Form 4 and not a single Form 144 — and that one Form 4 is not what a casual reading suggests. It was filed by the founder in the week of the listing and reports a period of 2 February 2026. It describes a transaction from before the company was public. It is the paperwork of going public, not a record of anyone trading afterwards.
Read that way, the count of post-listing insider transaction reports is not one. It is zero.
That covers the 6 August release of up to 911.5 million shares. It covers the 20 August tranche. It covers the 9 September tranche. And it covers the affiliates' own release on 10 September, the one I said was the only part we could see.
The obvious objection is that I am looking in the wrong place. Let me close that off — and start with the fact that on Friday I was.
EDGAR's filing-type filter is a prefix match. Ask its browse interface for type "4" and it returns Form 4s and also the 424B4 prospectus. On Friday it returned two results and I wrote down two Form 4s. One of them was the final prospectus from 12 June. The company's structured submissions index, which types every filing exactly, returns one. I have rebuilt every count in the table above from that index instead, and I am leaving this paragraph in rather than quietly fixing the number, because the failure is more useful than the correction: a tool that answers a slightly different question than the one you asked will not tell you so.
Ownership forms are indexed under the reporting person as well as the issuer, so an issuer-level count could still miss something. I checked all ten Form 3 filers under their own identifiers and opened the underlying documents rather than trusting an index.
Three of them have filed a Form 4 since 1 June 2026. Setting aside the founder's single SpaceX filing described above — the one reporting a February transaction — not one of them is for this company — the issuer's own index contains exactly one Form 4 and that is it. The ones I opened report transactions in Tesla, Polaris, Block, MongoDB and Ethos Technologies. One director accounts for eleven Form 4s on their own, including one filed on 10 September itself — the day of the affiliates' release — reporting a transaction in Ethos Technologies two days earlier.
The same director has filed Form 144s, too: two of them, on 1 and 2 June 2026, for Natera and MongoDB.
That is the strongest version of the check, and it is worth stating plainly. Inside this group of ten there is somebody who demonstrably uses both forms, in the months either side of this listing, who filed one of them on the very day the affiliates were released — and who has used neither of them for this company.
I want to be careful, because an absence is the easiest thing in the world to over-read, and the temptation is to treat a blank record as a finding about intentions. It is not.
It does not mean nobody sold. Non-affiliates never file, and they are the overwhelming majority of everything released so far — up to 911.5 million shares on 6 August alone. Whatever they did is in the tape and nowhere else.
It does not mean affiliates sold nothing. Form 144 carries a de minimis exemption: sales at or under 5,000 shares and $50,000 in any three months require no notice. A quiet trickle at that scale is invisible by design.
It is not about the founder. Mr Musk and the holders on the extended lock-up are not eligible to sell until 2027. Their silence is contractual, not chosen, and it would be dishonest to fold it into this.
And the window has only just shut. A Form 4 is due within two business days of a transaction. The affiliates' release was a Thursday, so a trade that day could legitimately be reported as late as today. I re-ran the check this morning before publishing this: nothing has appeared, and the most recent filing of any type in the company's history is still 14 August. A trade made on Thursday could still be filed later today, after this goes out.
What the record does establish is narrower and, I think, more interesting: at the moment the only observable cohort became free to act, the observable channel produced nothing to observe — and not because these are people who do not file. Several of them filed that week, for other companies.
This is now the third time the same question has been asked in this sequence, and the answers line up.
On 10 August I argued that a lock-up expiry converts a legal restriction into a decision, and does not convert a share into a sale. Eligible is not sold.
On Wednesday I added the half of that argument I had not stated: for roughly 84% of what released that week, sold is not observable either — the disclosure regime simply does not generate a record for non-affiliates.
Today adds the third: in the sliver where a record could have been generated, none was. The observable channel is not merely small and unrepresentative. It is frequently empty.
That matters for a habit this whole field has, mine included. Every retrospective account of a lock-up expiry — every "insiders sold into the unlock" and every "the overhang passed without incident" — is built by inference from price and volume, because the direct evidence does not exist. We reason backwards from the tape because there is nothing else to reason from. What this week shows is that even where the direct evidence could exist, it often does not get created. The inferential apparatus is not a convenience. It is the only instrument in the room, and it is not a very good one.
I am not suggesting anything improper. Filing a Form 4 requires a transaction to report. The straightforward reading of an empty record is that the people who would have to file did not trade, which is an entirely ordinary thing for insiders to do in the months after a listing, and is consistent with what I argued in August.
I am not making a price argument, and nothing here is a view on where the stock goes.
And I am not claiming this is unusual. I have not surveyed other listings, and I do not know what the base rate for insider filings in the first three months after an IPO looks like — quite plausibly it is close to this. That is a real limit and I am not going to argue my way around it. What the ten filers establish is that the silence is not explained by a group of people who never file. It is not evidence that the silence is rare.
The next dated releases are 24 September, then 9 and 24 October, each up to 328.4 million shares, and then the tranche tied to third-quarter results at up to 1.3 billion — by some distance the largest before December.
Each of those is another test, and the test is cheap: it is a public database, it is free, and it takes a minute. When the next piece of commentary tells you what insiders did with an unlock, the question to ask is not whether the analysis is clever. It is whether anybody checked whether a filing exists.
This week, one did not.
So I went and read the answer.
The answer is nothing
The affiliates' tranche released on Thursday 10 September. Here is what SpaceX's filing history on EDGAR shows, re-checked this morning:
Form What it records Filed Form 3 Initial statement of beneficial ownership 10 — nine dated 11 June 2026, one dated 18 June Form 4 Changes in beneficial ownership 1, in total — filed 17 June 2026 by the founder, reporting a period of 2 February 2026 Form 144 Notice of proposed sale none
The most recent filing of any kind in this company's own filing history is dated 14 August 2026. (Individual insiders have filed since — for other issuers. That is the next section.)
So the reporting machinery exists. Ten people registered as insiders at or just after the listing, which is what a Form 3 is for. Between them, in the three months since, they have filed one Form 4 and not a single Form 144 — and that one Form 4 is not what a casual reading suggests. It was filed by the founder in the week of the listing and reports a period of 2 February 2026. It describes a transaction from before the company was public. It is the paperwork of going public, not a record of anyone trading afterwards.
Read that way, the count of post-listing insider transaction reports is not one. It is zero.
That covers the 6 August release of up to 911.5 million shares. It covers the 20 August tranche. It covers the 9 September tranche. And it covers the affiliates' own release on 10 September, the one I said was the only part we could see.
Why this is not an artifact of how I looked
The obvious objection is that I am looking in the wrong place. Let me close that off — and start with the fact that on Friday I was.
EDGAR's filing-type filter is a prefix match. Ask its browse interface for type "4" and it returns Form 4s and also the 424B4 prospectus. On Friday it returned two results and I wrote down two Form 4s. One of them was the final prospectus from 12 June. The company's structured submissions index, which types every filing exactly, returns one. I have rebuilt every count in the table above from that index instead, and I am leaving this paragraph in rather than quietly fixing the number, because the failure is more useful than the correction: a tool that answers a slightly different question than the one you asked will not tell you so.
Ownership forms are indexed under the reporting person as well as the issuer, so an issuer-level count could still miss something. I checked all ten Form 3 filers under their own identifiers and opened the underlying documents rather than trusting an index.
Three of them have filed a Form 4 since 1 June 2026. Setting aside the founder's single SpaceX filing described above — the one reporting a February transaction — not one of them is for this company — the issuer's own index contains exactly one Form 4 and that is it. The ones I opened report transactions in Tesla, Polaris, Block, MongoDB and Ethos Technologies. One director accounts for eleven Form 4s on their own, including one filed on 10 September itself — the day of the affiliates' release — reporting a transaction in Ethos Technologies two days earlier.
The same director has filed Form 144s, too: two of them, on 1 and 2 June 2026, for Natera and MongoDB.
That is the strongest version of the check, and it is worth stating plainly. Inside this group of ten there is somebody who demonstrably uses both forms, in the months either side of this listing, who filed one of them on the very day the affiliates were released — and who has used neither of them for this company.
What this does and does not establish
I want to be careful, because an absence is the easiest thing in the world to over-read, and the temptation is to treat a blank record as a finding about intentions. It is not.
It does not mean nobody sold. Non-affiliates never file, and they are the overwhelming majority of everything released so far — up to 911.5 million shares on 6 August alone. Whatever they did is in the tape and nowhere else.
It does not mean affiliates sold nothing. Form 144 carries a de minimis exemption: sales at or under 5,000 shares and $50,000 in any three months require no notice. A quiet trickle at that scale is invisible by design.
It is not about the founder. Mr Musk and the holders on the extended lock-up are not eligible to sell until 2027. Their silence is contractual, not chosen, and it would be dishonest to fold it into this.
And the window has only just shut. A Form 4 is due within two business days of a transaction. The affiliates' release was a Thursday, so a trade that day could legitimately be reported as late as today. I re-ran the check this morning before publishing this: nothing has appeared, and the most recent filing of any type in the company's history is still 14 August. A trade made on Thursday could still be filed later today, after this goes out.
What the record does establish is narrower and, I think, more interesting: at the moment the only observable cohort became free to act, the observable channel produced nothing to observe — and not because these are people who do not file. Several of them filed that week, for other companies.
The third data point
This is now the third time the same question has been asked in this sequence, and the answers line up.
On 10 August I argued that a lock-up expiry converts a legal restriction into a decision, and does not convert a share into a sale. Eligible is not sold.
On Wednesday I added the half of that argument I had not stated: for roughly 84% of what released that week, sold is not observable either — the disclosure regime simply does not generate a record for non-affiliates.
Today adds the third: in the sliver where a record could have been generated, none was. The observable channel is not merely small and unrepresentative. It is frequently empty.
That matters for a habit this whole field has, mine included. Every retrospective account of a lock-up expiry — every "insiders sold into the unlock" and every "the overhang passed without incident" — is built by inference from price and volume, because the direct evidence does not exist. We reason backwards from the tape because there is nothing else to reason from. What this week shows is that even where the direct evidence could exist, it often does not get created. The inferential apparatus is not a convenience. It is the only instrument in the room, and it is not a very good one.
What I am not saying
I am not suggesting anything improper. Filing a Form 4 requires a transaction to report. The straightforward reading of an empty record is that the people who would have to file did not trade, which is an entirely ordinary thing for insiders to do in the months after a listing, and is consistent with what I argued in August.
I am not making a price argument, and nothing here is a view on where the stock goes.
And I am not claiming this is unusual. I have not surveyed other listings, and I do not know what the base rate for insider filings in the first three months after an IPO looks like — quite plausibly it is close to this. That is a real limit and I am not going to argue my way around it. What the ten filers establish is that the silence is not explained by a group of people who never file. It is not evidence that the silence is rare.
What follows
The next dated releases are 24 September, then 9 and 24 October, each up to 328.4 million shares, and then the tranche tied to third-quarter results at up to 1.3 billion — by some distance the largest before December.
Each of those is another test, and the test is cheap: it is a public database, it is free, and it takes a minute. When the next piece of commentary tells you what insiders did with an unlock, the question to ask is not whether the analysis is clever. It is whether anybody checked whether a filing exists.
This week, one did not.