The shape, described
a16z calls it the barbell. Longform at one end, essays and hour-long conversations that carry a full argument.
Short form at the other, one-minute clips pulled from that longform and cut for the feed. Weight at both ends and almost nothing in the middle.
The reason it is a barbell rather than a spectrum is that the two ends do completely different jobs, and the thing that does both jobs badly is the piece sitting between them.
What the middle actually is
The middle is the eight-hundred-word company blog post. The three-minute explainer video. The carousel that takes forty seconds to swipe through.
Every company produces these, and they occupy an unfortunate position: too long to consume in a feed, too short to say anything a reader could not have guessed.
A reader scrolling gives you about a second to justify stopping. A middle-length piece asks for forty seconds and offers a summary.
Someone who genuinely wants the answer finds it thin, and someone browsing never opens it. It fails the skimmer and the specialist in one move.
What the long end does
Longform is where an argument can survive contact with someone who knows the subject. It holds the detail, the caveats and the reasoning, and it is the artefact somebody links to months later when the topic comes up again.
It is also the half that answer engines can reach. Systems that write answers pull passages rather than pages, and they need text that stands alone and can be attributed.
A clip on X is largely invisible to them, because the platform restricts automated access to its content. A full essay on your own domain is not. The full mechanics are in getting cited by AI search.
The practical consequence is that longform is the only half of the barbell that compounds. Clips decay in days. An essay that answers a question properly keeps being retrieved for years.
What the short end does
The clip does the reaching. It is the entry point for people who have never heard of you and would never have opened the essay, and its job is to make a single idea land hard enough that some fraction of viewers go looking for the rest.
The important property of a clip is that it is carryable. A creator can post a clip in their own words with their own framing, which is something they cannot do with a two-thousand-word essay.
That is why a barbell strategy and a creator network fit together so naturally: the long end is yours to write, and the short end is what other people can amplify. Creator amplification covers how that half is run.
How many clips per piece
The useful ratio is high and most teams undershoot it badly. One hour-long conversation contains somewhere between eight and fifteen genuinely distinct ideas, each of which can carry a clip. One substantial essay usually contains four or five.
Teams that produce one clip per piece are treating the clip as a trailer for the longform. It works better treated as a standalone object that happens to have somewhere to send people who want more. The clip should be worth watching by someone who never clicks through.
This is also the cheapest content in the stack, because the thinking is already done. The expensive part was having the argument in the first place. Launch video strategy covers the production side, including why a raw screen recording usually beats a produced film.
The mistake almost everyone makes
The common failure is producing clips with no longform behind them. A feed full of one-minute takes with nothing underneath reads as content marketing, because a viewer who wants to check whether you actually know the subject finds nothing to check.
The reverse failure is quieter and more common at technical companies: writing genuinely good longform and never cutting it. The argument exists, it is correct, and about forty people read it. Every idea inside it that could have travelled stayed on a page nobody visited.
Both halves are cheap once the other exists. The order matters, and it runs long to short, because the clip is an extract rather than a source.
Questions we get asked
What is a barbell content strategy?
Publishing at two extremes and skipping the middle. Longform at one end, essays and hour-long conversations that carry a full argument, and short clips at the other, cut from that longform for the feed. The middle, meaning the eight-hundred-word blog post or the three-minute explainer, is too long to consume while scrolling and too short to tell a specialist anything, so it fails both audiences at once.
How many clips should one longform piece produce?
More than most teams make. An hour-long conversation usually contains eight to fifteen genuinely distinct ideas and a substantial essay four or five, each capable of carrying its own clip.
Teams that produce one clip per piece are treating it as a trailer. A clip works better as a standalone object that happens to have somewhere to send people who want the full argument.
Does short-form content help with AI search visibility?
Very little on its own. Answer engines pull passages of text they can attribute, and posts on X are largely unreachable because the platform restricts automated access.
Longform published on your own domain is what gets retrieved and cited. Clips do the reaching among people, longform does the reaching among the systems that answer questions about your category.