How a16z speedrun companies launch on X

What speedrun is, and what it is for

a16z speedrun is Andreessen Horowitz's accelerator for early companies in games, consumer and AI. Each cohort gets capital, a few months of programme and a demo day, then a company is on its own with a product to announce.

The announcement is where most of the value of the programme either compounds or leaks. A speedrun company launches with the a16z name behind it, a founder account that is usually small, and a cohort of other companies announcing in the same weeks. All three shape what the post does.

We measure every AI launch on X we can find. Eleven of the founders behind those launches name a16z speedrun in their bio, so there is a set to read.

What the speedrun launches did

The 11 speedrun launches we track have a median of 279,888 views, against 688,938 for the directory as a whole. That is where a programme company lands when the launch runs on the founder's own account with the following it has, which is normally a few hundred to a few thousand people.

Every figure follows the directory and updates as the posts move. The a16z-backed vertical holds these launches alongside every other a16z portfolio launch we track. Six of the eleven launched inside one eight-day window in late September, which is the cohort effect: companies that finish the programme together announce together, and on X attention inside a topic on a given day goes mostly to one post.

The one that went furthest from a small account

Codos introduced the first virtual Chief AI Officer on 21 September 2026. The post went out from co-founder Dima Khanarin's account, which had about 1,800 followers. It stands at 1.6M views, around 900 times that following, with 2,031 bookmarks, and it passed a million views on its first day.

Codos sells to CEOs. The audience that had to see the post was executives and the operators and investors they listen to, which is a hard audience to reach from a founder account with 1,800 followers and an easy one to reach through the people it already follows.

We got many VCs and many, many customers. We are overwhelmed. The launch was a success.

Dima Khanarin, co-founder and CEO of Codos, two weeks after the launch. The full case study has the post, the distribution and the numbers read by read.

What the Codos launch ran on

Doomers ran the launch strategy and the distribution. The plan had four parts, and none of them is specific to Codos.

Who the buyer follows. Enterprise executives on X follow founders, operators and a handful of investors, and almost no vendor accounts. The carrier list was built from accounts those buyers already read: founders who had sold into large companies, operators who write about AI inside the enterprise, and investors whose portfolios overlap the category.

The post carried one claim. "The first virtual Chief AI Officer" is a sentence an executive can repeat to a board.

The product detail sat in the replies and on the site. A launch post that tries to explain the whole product gives the people carrying it nothing short enough to quote.

When each group moved. The founder's own network went first, in the opening minutes, because early velocity decides how far X distributes a post.

Investors and the a16z network came in once the post was already moving, where they put it in front of the second audience rather than the first.

Team and friends filled the gaps between. The levers are the founder's. When each one is pulled is the amplification plan.

The link goes into the invite the moment the post is live. What that buys is quality noise early: enough real people replying, quoting and saving the post in the opening window that strangers start talking about the product on their own, and the ranking model has a reason to send it further.

A date of its own. Codos went on a Monday with nothing else from the cohort on the same day. Two other speedrun companies launched in the same week; the three posts never had to share a feed.

The plan for a speedrun company

Split the work into two buckets: story and distribution. Story is the video, the product doing the thing in a form people repost.

Distribution is the amplification, who carries the post and in what order inside the first hour. The story earns the attention and the amplification decides how far it travels, and that part is black and white. Codos had both.

Treat the a16z name as a credibility marker, and the founder's account as the launch. People quote the backing later. On the day, the post travels on who carries it, and a founder's own post from a small account outperforms a roundup from a large one.

Ask the cohort when they are going, and go on a different day. This is the cheapest decision in the whole plan. Across every launch we measure, Tuesday runs at the highest median of the week, and a day shared with two or three programme companies costs most of your reach.

Spend the programme months on the people who will carry it. Reply to them, ship things they can see, be a name they recognise before you ask for anything. A cold request on launch day converts badly.

Write the claim first and the product second. One sentence a stranger can repeat.

Codos had one. The posts in the table that did least were describing a category.

Line up the first hour before you post. Know which accounts move in the first thirty minutes, which move in the first three hours, and when the investors come in. That order is most of the difference between the top and the bottom of the table above.

The calendar invite. Set up a calendar invite for launch day and send it to every investor, advisor, friend and teammate, 50 to 100 people from the network. Do this for every launch, whatever the stage Everyone in the company's ecosystem knows the launch is happening and knows the minute.

Questions we get asked

How do a16z speedrun companies do when they launch on X?

The 11 speedrun launches we track have a median of 279,888 views. Codos reached 1.6M from a founder account with about 1,800 followers, and ZeroDrift's founder post reached 5.8M. The spread comes from who carried each post and whether it shared a day with the cohort.

Should a speedrun company launch on demo day?

No. Demo day puts every investor and reporter watching the category onto a whole cohort at once, and six of the eleven speedrun launches we track went out inside one eight-day window. A launch needs a day of its own, and in our data Tuesday carries the highest median of the week.

How did Codos reach 1.6M views from 1,800 followers?

Other accounts carried it. The post made one claim, the first virtual Chief AI Officer, and went out to founders, operators and investors that enterprise buyers already follow, in an order that gave it early velocity before the investor network picked it up. Two weeks later the founder reported many VCs and many, many customers from it.

How do you get engagement in the first hour of a launch?

Send a calendar invite for the posting time to every investor, advisor, friend and teammate, 50 to 100 people, and add the link when the post is live. That puts the whole network on the post inside the window X uses to decide distribution. The replies and saves from people who know the company give strangers a reason to talk about it on their own, which is where the reach past the following comes from.

Did Doomers run the Codos launch?

Yes. Doomers ran the launch strategy and the distribution. Codos built the product and the founder wrote and posted the announcement from his own account.