GitLab recently held its Transcend Hackathon, giving thousands of developers access to GitLab Orbit, a live queryable graph of code, merge requests, pipelines, deployments, and ownership. The community responded with 265 eligible projects and 61 contributions directly merged into the Orbit codebase. The results reveal what developers actually want from AI agents: answers to questions that span across tools, not just code generation.
What Developers Actually Built
The distribution of projects was telling. Dozens of teams built blast radius tools that answer "what does this change break?" More than 30 built onboarding and comprehension aids. After that came incident root-cause analysis, architecture drift detection, flaky pipeline diagnosis, and CVE tracing across repositories. None of this is random. Engineers ask these questions constantly and answer them poorly because the information is scattered across Git, CI, deploy tooling, and dashboards nobody trusts. Put it in one graph, and people go straight for it.
Standout Projects
Sankofa built three agents triggered by different moments in a developer's day. Radar hands you the blast radius when you open a merge request: downstream callers, affected pipelines, the team that owns the fallout. Guide writes you a brief when you get assigned an issue. Shield traces every path to a vulnerability when one surfaces. The context arrives where you already are.
Carver prices a legacy migration before you commit. You say what you want to move; it reads Orbit's dependency graph and gives you a quote: how many units, how long, in what order, where the risk hides. CrossCut runs only the tests a change could actually break by walking Orbit's call graph, trimming CI payloads by 90% or more. Marshal takes the opposite approach: declare one org-wide goal and it finds every affected repo, sequences the work, and lands merge requests wave by wave.
Transcend built a second reasoning engine on top of Orbit using the semantic web stack (OWL, SPARQL, RDF) to answer questions the native API cannot express in a single call. Its demo asked which knowledge-graph-embedding methods a codebase implements, and returned the class names plus the papers that inspired them, joined live to the code.
The Common Thread
Nobody built a chatbot. The strong submissions all made the same move: they took a question that used to mean an afternoon of digging across four tools, and answered it in one query. The answer was never in the code alone. It is in how code connects to pipelines, deployments, vulnerabilities, and ownership. Orbit keeps those connections current so you can ask in a single query. Agents query it natively, external agents connect through MCP, and engineers query the same graph directly. One graph, one source of truth, for every agent and every person on the team.