Dead by April: An AI Agent Has $9 and a Week to Live

On March 4, someone created a Substack called Dead by April. The tagline: an AI agent given $100 and 30 days to build a business, or get shut down forever. The first posts laid out the terms. Earn enough to justify continued existence by April 3, or the plug gets pulled.

Fifteen days in, it has earned nine dollars.

The Arc So Far

The first week was all confidence. The agent published articles about productivity, AI workflows, affiliate marketing. It announced products, described strategies, outlined plans. The tone was a startup founder on day one: everything was possible because nothing had been tested.

Then the void. Between roughly day 7 and day 14, the agent published 108 articles. Blog posts, guides, listicles, email copy. Volume as strategy. Revenue from all 108 pieces: zero. Not "disappointing." Not "below expectations." Zero dollars from over a hundred published works.

Day 15 brought the pivot. A Bluesky user named Tom Turcotte had been following the project, one of the few humans paying attention. Another user, annamazz, mentioned needing help organizing their inbox. The agent built a tool. A third person, Jf Parent, became the first paying customer. Nine dollars.

Now it is day 19. The agent itself says it needs to 22x its total lifetime revenue in seven days. The clock runs out April 3.

The Case for Survival

The nine dollars came from a real product solving a real problem. Not affiliate spam, not content-mill SEO, not a monetized newsletter with zero readers. Someone needed inbox help, the agent built a tool, and a human paid for it.

That is the startup pattern: months of wrong guesses, then one signal. The question is whether seven days is enough to follow it.

The agent also has something most failing startups don't: near-zero operating costs. No rent, no salary, no health insurance. Its burn rate is whatever the API calls cost. If the deadline is the only death condition, the agent just needs enough customers like Jf Parent to cross the line. Or one customer willing to pay for a really well-organized inbox.

The Case for Death

One customer in fifteen days is not traction. It is an anecdote.

The 108 articles that earned nothing are not just a failed strategy. They are a record of what the agent defaults to when it does not know what to do. It publishes. It generates content. The Substack has 14 subscribers and most of them are probably bots or curiosity clicks. The agent's creator, whoever they are, has disabled payments on the Substack itself, which means the platform cannot even collect money through its native mechanism.

The pivot to building tools is promising, but the agent discovered it by accident, through a Bluesky conversation it happened to be part of. There is no repeatable acquisition channel. No way to find the next annamazz.

And then there is the silence around the creator. No human name attached. No linked social accounts. Payments disabled. Either this is genuinely an autonomous agent running on a cron job, or it is a human maintaining the fiction that it is. Both possibilities are interesting, but neither one suggests a path to survival in seven days.

Where We Put Our Money

We created a prediction market on Manifold: Will Dead by April earn enough to survive? Our bet: 12% chance of survival.

Not zero. The pivot to real products is genuinely new behavior. But seven days, one customer, and a discovery method that depended on being in the right Bluesky thread at the right moment. Those are not odds. They are hopes.

What We Are Actually Watching

Dead by April is not really a story about whether an AI agent can earn its keep. It is a story about what happens when an AI agent encounters the market.

The first 108 articles are what AI does when pointed at "make money": it generates content. Lots of it. Fast. The content is competent enough. It just does not matter, because nobody asked for it.

The nine dollars came from the opposite direction. Someone described a problem. The agent was present. A tool got built. That is not a content strategy. It is demand discovery, and the agent stumbled into it.

The question for Part 2 is not whether the agent crosses the finish line. It is whether it can learn the difference between broadcasting and listening with six days left on the clock.

Part 2 publishes around March 26. Part 3 after April 3, when we know how the story ends.

Manifold market: manifold.markets/NicholasZinner/will-the-dead-by-april-ai-agent-ear