Post-hackathon discussion

[chenlica]

We use this post to discuss any ideas and thoughts after the hackathon event. Here are some comments received by attendees:

1. Several participants wanted to form teams in future hackathon events;
2. One participant mentioned that using two different agents (e.g., Claude Code for coding, Codex for code review) would be useful
3. One participant suggested that it would be great if at the beginning of the hackathon, there can be a more thorough demonstration of a Texera use case
4. One participant highlighted the importance of powerful laptops (e.g. larger amount of memory), as it can allow developers to launch multiple claude code sessions at the same time
5. A few Texera members, shared that sometimes Claude code took many detours when developing a small feature or bug fix.

[xuang7]

Adding some personal thoughts:

For group vs. individual participation, I think it may depend on the expected deliverables of the hackathon. If we encourage group work, the tasks may need to be slightly larger in scope so that team collaboration makes sense.

For the tutorial process, it may also be helpful to provide some context or preparation materials a day ahead, so participants can get familiar with Texera before the event.

For the agent-related comments, I agree that agents still cannot perfectly replace human effort, especially because they can make mistakes or struggle with highly context-specific tasks. At the same time, agents are still powerful and may evolve quickly. I think there are already some efforts in the repo to provide better agent-facing documentation or project context, so the agent can better understand the codebase and reduce unnecessary detours. This could be a useful direction for future improvement.

[nathant27]

I think it would be nice to have a unified form of communication like Slack for any announcements and for all the resources

[zyratlo]

To address 1., previous hackathons I have participated in all allowed teams (up to 3 people), but the scope of the projects were different. For this specific hackathon case, in my opinion individual work with Claude is better than many people working on one codebase, with each person using Claude since it can get messy quickly.

For 3 and 4, although I am already familiar with Texera I can see the benefit of having a longer structured tutorial phase besides just teaching how to setup the development environment on your laptop. If you don't know what the system can do it will take longer to come up with ideas to improve it, right. And we can send a disclaimer during sign up that laptop resources like RAM may be a limiting factor for productivity.

For my personal thoughts, agents are very useful for creating prototypes of ideas that work. However, whether the code is high quality or not is a different problem. In my opinion, with the introduction of agents like Claude the burden and time overhead shifts from implementation to discussion, planning, and review.

[SarahAsad23]

I think allowing teams in future hackathons could make the experience more engaging and collaborative. It would also give participants a chance to get to know each other better.