IdeaNet

Inspiration

There are several common problems when generating ideas. When working as a team, it isn't easy to ensure everyone is fully heard, and also there are too many problems surrounding idea generation. Brainstorming is a very common situation, everyone has done it at some point. We wanted to make the process more efficient, and make the ideas well constructed. It is happened too many times where teams think of an idea, and go ahead with it in full excitement only to realise that the idea was not as good (which can be a nightmare to handle when you try to execute it).

What it does

Through the use of multiple different AI personalities, users can prompt and speak to the agents when they have difficulties in the brainstorm. There are 4 key agents here:

spark: sparks the conversation by adding on to existing ideas, and also promotes ideas to ensure people feel confident and heard when presenting a new idea.
probe: the cynical agent, it will challenge an existing idea, picking out potential flaws and issues. This can be used when everyone is getting excited with an idea, and we need to check for any potential drawbacks
anchor: this bot is called automatically after every few messages, to remind the group what the actual task at hand is, and if they have veered too far away from the initial goal, counters 'scope creep'
facilitator: This is the main bot that oversees the entire session. Users can send their ideas to this bot, which asks the users questions about their idea to ensure that the idea is strong and feasible, then the bot will suggest the idea to the main chat. This ensures that everyone is heard, including the shy people of the team.

The bots are incorporated into the chat, and can be called at any time. There is also a running mindmap which users can edit as they go along, to keep a track of ideas discussed. At the end of the session, a summary is generated including the key insights discussed, the main ideas presented, the actions that need to be done after this session. The session summary is saved for users to come back to later, and users can also save the mindmaps individually.

How we built it

initially built in python using the slack API, openAI's API (as claude was paid) as a bot for slack. we then realised that we can add more functionality to it if we entirely made it a webapp. So then we coded with json and react. We used ai-assistance to make the app UI smooth, and also to create a login system. We then created the agents, and integrated it into the app.

Challenges we ran into

We initially had the idea of creating a slack bot, so until 5pm yesterday we developed that, but realised it is not as impressive as it could be, hence we had the idea of creating a web app really late. This meant that we had to rush the development, and use AI to help structure and create an initial app which we expanded on. Also, everyone lives far away from the actual venue, so the travel was definitely something we had to put up with (especially getting back home extremely late), it took up valuable time we could have spent hacking away.

Accomplishments that we're proud of

We worked well as a team, not knowing each other initially, we worked together to create an idea we firmly believe in.

What we learned

We learnt several key things:

ensure the idea is good, impressive and not too simple
focus on a track, then an idea (rather than fit an idea to a track)
delegate, rather than everyone working on a single task at once

What's next for IdeaNet

We could push a similar system but instead using actual voices rather than text, so expand the system to real-time meetings (Integrating this to notion, teams and google meet). We could also make the agents interact with each other to produce stronger, well-reasoned ideas. Adding another agent, a data agent, that scrapes data and uses facts and figures to back ideas or to state why an idea isn't the best. Finally, we could make the platform more accessible (e.g. light/dark mode, text-to-speech, font-size etc).