Ask for rules before opinions
Devpost says participants should check eligibility rules, build requirements, required tools, URLs, and demo limits before building. The first mentor question should be boring: does this plan satisfy the rules and the track requirement?
Karen Bajza-Terlouw told Devpost she checks whether a submission fulfills requirements first. Richard Moot said standout projects clearly considered the judging criteria. A mentor cannot save a project that ignores the scoring sheet.
Bring a packet, not a monologue
Devpost's online teamwork guide says virtual teams should introduce skills and share basic project concepts before joining. Use the same discipline with mentors: one user sentence, current demo path, blocker, and the exact decision you need.
Jackie Luc's CalgaryHacks team clarified the theme with organizers, defined roles, and started with the pitch. That is the model. A mentor session should tighten the plan, not become a second ideation meeting.
Use sponsor context carefully
Google Cloud Rapid Agent Hackathon required a functional AI agent powered by Gemini and Google Cloud Agent Builder that integrated one partner MCP server. In that kind of track, ask sponsor-side helpers what qualifies as real integration before you build the wrong thing.
AgentHacks 2025 listed office hours alongside cash, Claude credits, Vercel credits, Mistral AI credits, and gift cards. Office hours can be valuable, but they are not prize money. Treat them as access to sharper constraints, not validation that your idea is good.
Let mentors kill scope
Devpost winner Ansh recommends 2 to 3 must-have features for a prototype, and AngelHack says winners get a working skeleton by the first 25 percent of the event. Ask mentors which feature is required for judging and which one is just team anxiety.
Chris Ho's API approval examples show why late dependencies are dangerous: Twilio A2P could take 2 to 3 days, and another project had to switch when Instacart approval did not land. A good mentor may save you by saying no early.
Separate mentor feedback from judge feedback
Maria Yarotska from NEAR told Devpost she looks for problem understanding and startup mindset, not only technology. A technical mentor may help the system work, but the judge still needs to see the user, market, or impact logic.
Richard Moot told Devpost he loves connecting with standout teams after a competition. That is different from live build feedback. During the event, ask for decisions. After the event, ask for follow-up, intros, or critique.
Follow up with proof
OpenAI's internal AI hackathon playbook says follow-up should classify outputs as learning examples, needing more testing, limited pilots, reusable examples, or not moving forward. That is a clean way to report back after mentor help.
Send the mentor the project link, demo video, what changed because of their advice, and the next ask if there is one. The best follow-up is specific. It turns a rushed event conversation into a real builder relationship.
< read by a human · updated as things change >
browse hackathons