Where the phrase comes from
"Always be launching" is Andrew Chen's. His argument, made in late 2024, is that the single big announcement belongs to an era when front pages were scarce, and that in a world of feeds a company should talk about what it is building constantly, from the founder's own account, and treat every week as a small launch. The line that carries it: "Quantity is a quality all on its own."
Most people who repeat the phrase agree with it on instinct. We can do something else with it. The launch directory holds every AI launch on X we have been able to find, with the views, saves and reach against following each one earned.
If launching often works, it should show up there as a number. It does, and the reason it shows up is more useful than the slogan.
What can be tested
The claim has two halves. The first is that companies which launch repeatedly do better than companies which launch once. The second is the mechanism: that each launch warms the ground for the next, so an audience accumulates and later launches travel further than earlier ones.
The directory can check both. It knows which account posted each launch, so launches group by company.
It has the date, so they order. And it has the views, so a company's first launch in our window can be set against its later ones.
Measured on 14 September 2026, the directory held 592 launches from 535 companies. Forty-two of those companies launched two or more times, eight of them three or more, and between them they account for 99 launches and 36 percent of every view we have measured.
One limit to state up front. "First" here means first inside our window, which opens in March 2026. A company's first launch in the directory is rarely its first launch ever.
What the data compares is a company's earlier and later launches over roughly six months, which is the span Chen's argument is about.
Companies that launch again
| Companies | Launches | Median views per launch |
|---|---|---|
| Launched once | 493 | 231,359 |
| Launched two or more times | 99 | 568,056 |
The first half of the claim holds, and by a wide margin. A launch from a company that launches repeatedly does about two and a half times the views of a launch from a company that has launched once. That is per launch, before counting that the repeat launchers have several of them.
Part of this is who they are. The repeat launchers post from larger accounts, a median of 15,174 followers against 4,945 for the one-time launchers, and they are companies with enough shipping to have something to announce every few weeks. The median gap between one launch and the next, across all 42, is 25 days.
Selection explains some of the difference. It does not explain a factor of two and a half on its own, because the median launch reaches many times the following of the account that posted it, and account size is a weak predictor of where a launch ends up.
What did not compound
The second half of the claim is where the data parts from the slogan. If launches warmed the ground for each other, later launches would outperform earlier ones. They do not.
| Among the 42 repeat launchers | |
|---|---|
| Median views, first launch in our window | 568,056 |
| Median views, every later launch | 567,274 |
| Companies whose biggest launch came first | 21 |
| Companies whose biggest launch came later | 21 |
A coin flip. The sequences behind that number make it concrete.
Higgsfield launched eight times between April and September. The first, a brand film about the history of moving images, did 14.8 million views. The next four, each a product release two to three weeks apart, landed between 1.4 and 2 million.
Two September releases did 98,401 and 258,977. The eighth, an After Effects integration, did 2.7 million. Eight launches, no trend, and the biggest was the first.
Runway went the other way. A research post on 31 August did 713,802. A second three days later did 301,320.
On 11 September a 30-second video from an employee's account, which we wrote about at the time, passed two million views inside three days and has since passed ten million. The third launch did more than fourteen times the first, and the two before it had nothing to do with why.
Enrich Labs, posting from an account with 5,010 followers, did 3.7 million in March, 2.2 million in August and 66,868 in September. Chatbase did 515,585, then 849,078, then 28,201, then 709,037.
Every one of those launches was judged on its own. What it led with, what it showed, who carried it in the first hour. The launch before it was not in the room.
Why launching often still wins
Put the two findings together and the mechanism is plain. Launching often does not make each launch bigger. It makes more launches, and each one is a fresh draw from a distribution that is very wide.
Across the directory the top tenth of launches take most of the views, which means the outcome that matters is the one in ten, and the way to get one in ten is to take ten.
Higgsfield's eight launches did 24.5 million views between them. If the company had saved everything for one announcement it would have needed that single post to do what its best one did, with no second chance.
Runway's ten-million-view post would not exist if the company had decided in early September that it had already launched that month.
This is also why the September dip in Higgsfield's numbers did not matter. A 98,401-view launch followed by a 2.7-million one is the same company on two different days with two different packages. The cadence is what made the second day available.
Where the advice and the data part ways
Chen's post is about a founder's own account: post daily, talk like a person, skip the gatekeepers and the agencies. On the daily habit the data has nothing to add. That is a relationship with an audience and it is yours to keep up.
A launch is a different unit. It earns almost all of its reach from outside the account that posted it, from people who did not follow it that morning, and it earns that reach or fails to inside the first hour.
A founder posting alone has one account's first hour. Runway's biggest moment came from one employee's post, which is a point for Chen, and it came with the packaging that post had, a $9 price and a ten-minute claim a viewer could check against the video.
The same company's amplified creator push two months earlier, with a weaker opening, reached a fifth as many people from fourteen times the following. Who carries a launch matters less than what they are handed to carry, and both matter more than what the account did last week.
So the honest version of the slogan for a company, as opposed to a founder, is: keep the drumbeat, and treat each launch as its own event with its own package and its own first hour. The habit does not do the launch's work for it.
What to take from it
- Launch every time you have something. A company in the directory that launched more than once averaged a launch every 25 days. Each one was a separate chance at the top of the distribution, and the companies with the most chances hold most of the views.
- Stop expecting the last launch to help this one. It will not, and planning as if it will is how a release goes out with a thin package because the audience is "already warm".
- Package each one for a stranger. Most of the people a launch reaches did not follow you that morning. The opening line has to work on someone with no context, which is the test Runway's September post passed and its July push failed.
- Keep the founder's account separate from the launch plan. The daily post is a habit and it compounds slowly into an audience. The launch is an event and it is graded in an hour. Running the second like the first is the mistake the slogan invites.
- Check your own sequence. If you have launched more than once, your launches are in the directory. Look at whether your later ones beat your earlier ones and, if they did not, at what the biggest one led with.
Questions we get asked
Does launching more often make each launch bigger?
No, on the evidence we have. Among the 42 companies in our directory that launched two or more times, the median first launch did 568,056 views and the median later launch did 567,274.
Half had their biggest launch first and half had it later. What launching often does is give a company more launches, each a separate chance at the top of a very wide distribution.
Do companies that launch repeatedly do better overall?
Yes, by a lot. Their launches did a median of 568,056 views against 231,359 for companies that launched once, and 42 companies account for 36 percent of every view in the directory. Some of that is account size and having more to ship, and the rest is that each launch is another draw.
Is "always be launching" the same as founder-led posting?
They overlap and they are not the same thing. Andrew Chen's original argument is about a founder's own account and the daily habit of talking about the work. That habit builds an audience slowly.
A product launch is graded in its first hour, earns most of its reach from people who do not follow the account, and depends on what it leads with. Keep the habit, and plan each launch as its own event.
How far apart should launches be?
The repeat launchers in our directory averaged 25 days between launches, with Higgsfield running two to three weeks apart for four releases in a row. There is no evidence in the data that spacing them further apart helps a launch, and the companies that launched most often hold the most views. Launch when the package is ready.