Frontier labs launches on X

Launching in this vertical

These are the launches everyone else in the directory gets measured against, usually unfairly. A frontier lab announcing a model is news in the way a funding round is news: it travels on the name before anyone has read the post, and the reach it earns says almost nothing about what a company with no name should expect.

So the first use of this page is to stop benchmarking against it. A startup launch that does a third of a lab's median has done extraordinarily well, and a founder reading the top of the hub without this filter will conclude the opposite.

What a lab launch is actually competing on

Not attention. A lab has that before it posts. What separates the launches above is how much of the reach turned into people putting their name to it, and the repost rate in the strip at the top of this page is the number to read.

A model release that everyone saw and few endorsed is a release the category has decided to wait on. One that travelled on reposts is one the category has decided to try.

The second thing to read is the account. The biggest launch in the directory went out from the company account, because a lab's name is the product.

That ordering reverses for almost everyone else here, where the founder's account travels further from a smaller following, and it is the single clearest sign that lab launches and startup launches are different objects.

What is useful in them anyway

Timing and shape. Lab launches are the most watched events on the platform and they set the week for everyone in the category.

A startup launching the same day as a model release is launching into a feed that has already decided what it is talking about, and the directory dates let you check that before choosing a day. The guide on timing works through it.

The other thing worth taking is the demo. The lab launches that travelled furthest led with something a reader could watch or try in the same session, and the ones that led with a benchmark table sit further down. That holds at every size in the directory.

Where we sit

We do not run lab launches, and this page does not pretend otherwise. It exists so the rest of the directory can be read correctly.

Every other vertical page compares itself with the overall median, and a large share of that median is set by the companies above. Knowing that changes what a good number looks like.