Ask any AI model what happened this morning and watch it make something up. Every LLM is frozen at its training cutoff, and the moment the world moves, it starts guessing. Ground Truth is a one-day sprint to build agents that don't guess.

Ground Truth is hosted by The Builder Series, You.com and localhost:nyc in New York City. Every project is built on the You.com API, which gives agents real-time, cited access to the web. You have one day to build an agent that searches, reads, reasons and shows its receipts.

We're looking for agents that do real work on live information: watching, researching, verifying, deciding. We're not looking for a chatbot wrapper with a search box bolted on, or a slide deck about an agent you'll build later. If your project would work just as well with a model trained last year, it's not the right project.

Pick one of the three build tracks and build an agent that is useful precisely because it knows what's happening right now.

• Live Wire (real-time agents): an agent that watches a market, a competitor, a regulator or a job board and acts the minute something changes.
• The Analyst (deep research): pick a profession (VCs, lawyers, journalists, recruiters, doctors, policy staffers) and turn its 40-hour research memo into a 4-minute one, every claim cited.
• Receipts (trust and verification): an agent that fact-checks claims, catches hallucinations and verifies sources before anyone relies on them.
• Best Content (open to every team): post about what you built, tag You.com, and compete for a dedicated prize picked by the hosts.

Each criterion is scored out of 5.

• Design (out of 5): Is it something a real user would enjoy using? We look at clarity of the interface, how fast you get to value, and whether the agent's sources and reasoning are easy to see.
• Uniqueness of Idea (out of 5): Is this a problem only a live-web agent can solve? We reward ideas that are specific, surprising, and impossible with a model frozen in the past.
• Technical Execution (out of 5): Does it work live, on today's data? We look at how deeply the You.com API is integrated, the quality of the agent loop, and how it handles bad or conflicting sources.
• Business Execution (out of 5): Is this a company, not a project? 
• Use of You.com (out of 5): Is the You.com API the engine of your agent, not an add-on? We look at which endpoints you use, how much of the product depends on them, and whether every answer shows its sources.

• 1st Place Overall: $500 cash
• Best Content: $150 cash

• Design (out of 5)
  Would a real user enjoy it? Can they see the sources?
• Uniqueness of Idea (out of 5)
  Only possible with a live-web agent.
• Product and Technical Execution (out of 5)
  Works live, on today's data.
• Business Feasibility (out of 5)
  A real customer, a market and a way to make money.