Who Runs the Event Changes What Wins

Split the archive by source and the shape appears. Seventeen winners here came through lablab.ai, spanning BrightData, the AI Agent Olympics, ExecuTorch and the early OpenAI events. The rest come from companies — GitLab, Google Cloud, Microsoft, Databricks, Qdrant, Meta, Salesforce — announcing winners on their own blogs.

The difference shows up in what judging weighs. A platform-hosted event exists to showcase the host's technology, so depth of usage is effectively a precondition; look at the ADK, GKE, Agentforce and Copilot Studio winners and the sponsor's product is the skeleton of the project. Community-run events impose looser technical constraints but draw larger fields, so novelty and finish decide it. ChattyRental winning the early ChatGPT API era on a claimed 36% cut in sales costs is the archetype of that second kind.

Takeaway: before choosing an event, look at where the winners will be announced. If it is a company blog, start by reading that company's product documentation. If it is community-run, sharpen whose problem gets solved rather than technical depth. The same project does not have the same odds in both.