Challenge 2: Dashboard Walkthrough + Nash Demo
Try out Nash!
Challenge 2: Dashboard Walkthrough + Nash Demo
Global Hack Week: Agent Week | Backboard.io
Time needed: about 10 minutes
Difficulty: Easy. You are clicking around and trying things, not writing code.
What you need: Your Backboard account from Challenge 1.
WALKTHROUGH VIDEO 1 HERE
WALKTHROUGH VIDEO 2 HERE
No time for videos, or want to go at your own pace? This written tour covers everything. Follow along in your own browser and you will not miss a thing.
Part A: The Backboard Dashboard tour
Log in at . The dashboard is mission control for everything you build this week. Here is what each area does and what to try in it.
1. The home / analytics view
The first thing you see. It shows:
- Activity charts: how many prompts and tokens you have used, viewable over 24 hours, 7 days, or 90 days. (A “token” is roughly a chunk of a word. AI models bill by tokens, so this chart is basically your fuel gauge.)
- Model distribution: a donut chart showing which AI providers you have been using.
It will look empty right now. By the end of this challenge it will not be.
Try this: Note what your charts look like now, then check again after finishing this challenge and watch the numbers appear.
2. The built-in chat
Find the chat area in the dashboard. This is the fastest way to talk to any model without writing a single line of code.
- Thread history lives in the sidebar. Every conversation is saved as a “thread” and you can jump back into any of them.
- It supports memory and built-in document search (RAG), so it behaves like the agents you will build later.
Try this (required for the challenge):
- Start a new chat.
- Make sure memory is turned on (look for a memory toggle or setting in the chat).
- Tell it something about yourself, for example: My name is Alex, I am competing in Global Hack Week, and my favorite pizza topping is pineapple.
- Now start a brand new thread and ask: What is my name and my favorite pizza topping?
- Watch it recall your facts from a totally different conversation. That is Backboard’s #1-ranked memory doing its thing. This is the superpower your agents will have.
3. The model library
Find the models section. This is the catalog of 17,000+ models you can call through Backboard: OpenAI, Anthropic, Google, Mistral, open-source models via OpenRouter, and more.
Try this: Search for a model you have heard of. Then pick one you have never heard of and chat with it. Switching models on Backboard is just picking a different name from a list. No new accounts, no new API keys.
4. The memory page
Find the memory section. Remember the facts you told the chat earlier? They live here as individual memory entries.
- Facts are saved automatically when memory is enabled.
- You have granular control: view, edit, or delete any single memory.
- Memories belong to an assistant and are shared across all of that assistant’s threads. That is why a new thread still remembered you.
Try this: Find the memory that says your name. You should also spot your pizza topping. Delete one if you want to see the control you have. Your data, your call.
5. Settings: API keys and billing
Under Settings you will find:
- API Keys: where you made your key in Challenge 1.
- Billing / credits: your balance, including the promo credits from GLOBALMLH2, plus recent usage.
Try this: Confirm your promo credits are showing.
Full dashboard docs:
Part B: The Nash demo
Nash is Backboard’s chat app, built on the same platform you just toured. Think of it as “one chat app, every model.”
Why does that matter? Most people juggle separate AI apps and paste their data into whichever one is open. Nash is one sanctioned place where you can use any model, text or image, without scattering your info across five different tools.
Take Nash for a spin
- Go to https://hellonash.ai.
- Sign in.
- Start a conversation with the default model. Ask it anything. Maybe: Explain what an AI agent is like I am 13.
- Switch models mid-conversation. Find the model picker, choose a different model, and keep chatting in the same conversation. Notice you did not lose your chat history.
- Try an image. Nash is multi-modal, meaning text and images. Ask it to generate an image, for example: Generate an image of a robot hacking at a hackathon, cartoon style.
- Compare answers. Ask two different models the same question and see how their answers differ. This is a genuinely useful trick for the rest of the week: when one model gives you a meh answer, ask another.
How to complete this challenge
Submit two screenshots:
- The Backboard dashboard chat remembering your facts in a new thread (Part A, step 2).
- Your Nash conversation, showing a model reply or a generated image (Part B).
- You’ll need to combine both screenshots into the same file.
Troubleshooting
| Problem | Fix |
|---|---|
| Chat says it does not remember me | Make sure memory was ON when you told it the facts, and that the new thread uses the same assistant. Check the memory page to confirm the facts were saved. |
| Model will not respond | Check your credit balance in Settings. If it is empty, make sure you registered through app.backboard.io/hackathon with code GLOBALMLH2 (see Challenge 1). |
| Nash sign-in trouble | Try the same account/email you used for Backboard, and disable aggressive ad-blockers if the page misbehaves. |
Next up → Challenge 3: Install R-CLI (or Backboard Studio). Time to put an AI coding agent in your terminal. This is the tool that will build your projects for the rest of the week.
Submissions are only open during the event.