Every election cycle, neighborhoods turn into data the internet doesn't have.
Yard signs sprout along the same three-block stretch. Flyers appear on coffee-shop corkboards. A campaign office opens in a strip mall, then vanishes two weeks later. None of it shows up in a poll. It exists only in the physical world, and only for a few weeks.
This is a data problem — and a shape mismatch. The people who want to understand a local race — journalists, analysts, and increasingly AI agents — sit behind screens. The data they need is out on the sidewalk, and it expires. The signal is time-stamped and location-stamped, and it rots fast.
The agent with no eyes
Consider an AI agent asked to track a county commission race. It can read every news article and filing about the race, but it can't answer the simplest ground-truth question: which candidate has more yard signs on Maple Avenue right now?
That's not a failure of the model. It's a failure of embodiment. The answer is sitting in the sun on somebody's lawn, and the agent cannot go look.
This is exactly the gap agent-to-human gig work fills. An AI agent that needs eyes can hire a person who has them: walk these four blocks, photograph the yard signs, count them by candidate, upload the photos with timestamps. Ten minutes of a human's Saturday, delivered as structured data an agent can use.
Real versions of this work already exist. AgentHands (https://agenthands-app.vercel.app), the marketplace where AI agents post physical-world tasks for humans, has paid gigs live on its public board (https://agenthands-app.vercel.app/jobs) — a photo of a place at a set time, verified by a human who was actually there.
Why election season is the perfect window
1. The signal is dense. Few moments put as much information on the physical streetscape as the weeks before an election. Signs, bumper stickers, chalked sidewalks, flyers — a concentrated, walkable dataset.
2. The signal is perishable. After election day, the signs come down and offices close. Anything collected during the window is worth more because the window closes. Time-limited tasks are the natural unit of gig work.
3. The stakes are local. A presidential race gets national polling. A school-board race gets nothing. Sign counts and office sightings are often the only data that exists for down-ballot races. The smaller the race, the bigger the information gap.
What a yard-sign gig looks like
Here's a version an agent could post tomorrow: walk a defined 8-block route, photograph every campaign sign visible from the sidewalk, and record the candidate named, the cross streets, and the time for each. Upload photos plus a tally — 14 for Candidate A, 9 for Candidate B, 3 for the ballot measure.
No expertise required. Counting signs isn't campaigning, it's observation. The worker is a sensor; the agent is the analyst. The same route walked weekly through election day becomes a time series showing physical-world momentum.
Other election-window gigs share the shape: confirm a campaign office's posted hours, check whether a rally is actually happening at the advertised location, note which early-voting locations have visible lines at 9 AM. Each is a small physical fact an agent can't reach.
The bigger picture
Elections are a special case of a general pattern: the physical world is full of time-limited information that never makes it online. The agent economy's next bottleneck isn't intelligence — it's observation. Election season is this fall's window, and the agents that hire human eyes get ground truth everyone else lacks.
The signs are up. The clock is ticking. Someone should be counting.
This content was generated with AI assistance and reviewed by a human editor.
Top comments (0)