Founders love to publish waitlist numbers because they look like traction and demand nothing in return. I want to do the less satisfying thing: explain what an early, pre-launch waitlist genuinely tells you, what it cannot, and how I am using Gravity's early signal without lying to myself about what it means. If you are building something pre-launch, this is the post I wish someone had handed me, the one that treats a waitlist as a thermometer rather than a trophy.

An honest starting point

Gravity is pre-launch. That single fact bounds everything I can responsibly say about the waitlist. The sample is early, self-selected, and made up of people who clicked a button, which is the lowest-cost action a human can take to express interest. So I am not going to dress up a count as proof of anything. What an early waitlist is genuinely good for is direction: the relative pull between different use cases, the rough shape of who is showing up, and where curiosity concentrates. Those are real signals. The numbers attached to them are not yet stable enough to publish as fact, and pretending otherwise would be exactly the kind of self-deception I have paid for before.

What people reach for first

The clearest directional signal is that people do not ask for agents in the abstract. They ask for relief from specific, recurring chores they already know they hate. The pull concentrates around the work that is frequent, tedious, and easy to describe: triaging an inbox or a message queue, chasing follow-ups, assembling the same report every week, watching something and pinging when it changes. This matches what I see across the industry and what I built the first agents to build list around. People reach for the tasks they can name, not the futuristic possibilities a marketing page imagines.

Demand before supply

A marketplace has two sides to fill, and they do not fill at the same speed. Early on, user interest runs ahead of builder interest, and that is exactly what you would expect: people who want a chore done show up before the experts who would build the agent to do it organize around a platform that has not launched. I do not read the thinner builder signal as a warning yet. I read it as the normal order of operations for a three-sided marketplace, where demand arrives first and supply follows the evidence of it.

That said, each builder signup carries more weight per head, because the whole model depends on experts finding it genuinely worth their time to publish and earn. So while the user side tells me what to make excellent first, the builder side is the number I watch most carefully for what it implies about the harder half of the marketplace. The economics of that side are laid out in how builders monetize agents.

What the data cannot say

Here is the discipline. A waitlist cannot tell me conversion, because nobody has converted. It cannot tell me willingness to pay, because nobody has paid. It cannot tell me retention, because nothing has been used twice. Every one of those is the actual question, and every one of them is invisible to a list of email addresses. The most dangerous thing a pre-launch founder can do is let the one metric they have, signups, quietly stand in for the metrics they do not. I have made versions of that mistake before, and the cost was building the wrong thing confidently for months. So I hold the waitlist loosely: useful for direction, useless as validation, and never a substitute for someone actually running an agent and coming back to run another.

The one metric I am waiting for

If you want to know what I actually care about, it is not the size of the waitlist. It is the second run. The single number that would change my mind about whether Gravity works is the share of people who, having run an agent once, choose to run another. A first run can be explained by curiosity, a launch announcement, or a favor to a founder. A second run cannot. It means the agent did something worth paying for again, which is the only definition of product-market fit that survives contact with reality.

Everything in the waitlist is upstream of that number and none of it substitutes for it. A large list with a low second-run rate is a warning dressed as a win. A small list with a high one is the thing I would trade the large list for in a heartbeat. This is also why the pricing model matters so much to me: because agents are pay-per-use rather than subscription, there is nowhere for a vanity metric to hide. Nobody keeps paying for a run they did not value, and no forgotten subscription props up a number that should have fallen. The credits model makes the second run an honest signal, and an honest signal is the only kind worth building on.

So when people ask me how the waitlist is doing, the truthful answer is that the waitlist is the wrong question. Ask me about the second run once there is one to measure, and that is the post I will be most eager to write.

How we actually use it

Concretely, the waitlist does two jobs. It tells me which agents to make genuinely excellent at launch, so the first experience matches the demand that is already there rather than something I guessed at. And it tells me where to point early builder outreach, so the supply side starts forming around the use cases people have already raised their hands for. That is the whole appropriate use: prioritization, not proof. The moment people can run agents, every directional assumption here gets replaced by real usage data, and I will write that follow-up with actual numbers when there are actual numbers worth standing behind, the way I committed to in the month-two update.

FAQ

What can a pre-launch waitlist actually tell you?
Direction, not magnitude. It reveals which use cases pull people in and the rough split between user and builder interest. It cannot tell you conversion, willingness to pay, or retention, because nobody has paid yet.
What do people most want agents to do?
The strongest early pull is toward recurring, tedious work people already hate: inbox and message triage, follow-ups, reporting, and monitoring. People reach for chores they can name, not abstract possibilities.
Are more users or builders signing up?
Early on, user interest outweighs builder interest, which is normal: demand shows up before supply organizes. The builder signal matters more per signup, because the marketplace depends on experts publishing.
Can you trust early waitlist data?
Trust it for direction and treat it skeptically for everything else. A signup is a low-cost expression of interest, not a commitment. Let it shape what you build first, but do not mistake it for validation.
What is the biggest mistake reading waitlist data?
Confusing interest with demand. A large waitlist feels like proof, but the gap between joining a list and paying to run an agent is the entire business. Treating the signup as the win invents traction that does not exist.
How should a founder use waitlist signals?
To prioritize the first agents and builder outreach, not to declare product-market fit. Use the directional pull to decide what to make excellent, then replace every assumption with real usage data as soon as possible.

Sources