Stage 4: Connect to RAISE Playground
Keep both tabs open all week. Open in a new tab — don’t use the buttons in this page to leave the course.
a RAISE Playground project that loads and reads your AI
how a URL lets one website find work from another website
a saved RAISE project whose AI blocks can report your hand sign
Show the room slowly. Today has the most tool steps:
- Open RAISE Playground. Point out the cat sprite and the block palette.
- Click Add Extension (or the equivalent menu — show the exact path on RAISE's UI).
- Pick the AI / Teachable Machine extension.
- Paste your teacher model URL into the field. Show the new AI blocks appearing in the block palette.
- Click the prediction block by itself and show it returning "Rock" or "Paper."
- Save the project. Show campers exactly which menu saves it.
The big idea
So far we've been in Teachable Machine — the AI's home. Today we walk over to RAISE Playground — the game's home — and we hand it our AI.
Teachable Machine RAISE Playground
(your AI's home) (your game's home)
│ │
│ model URL is the bridge │
└────────────────────────────────────┘
The two websites share the AI with a URL. That is the long link you saved at the end of Stage 2 and Stage 3. RAISE Playground uses that URL to load your AI into the game. After today, the game can ask the AI, "what hand sign do you see?"
We don't build the game rules today. We do one important test: can RAISE read the model's prediction? If RAISE can answer Rock, Paper, Scissors, or Nothing, the bridge works. Tomorrow the sprite will use that answer.
Today is the handoff. Teachable Machine made the model URL. RAISE uses that URL to load the model. Then RAISE gives you an AI prediction block for the game.
- sprite
- a character in your game (like the cat)
- extension
- an extra set of blocks RAISE Playground can add — including AI blocks
- block
- a colored puzzle piece you snap together to make code
- save
- store your project so it doesn't disappear when you close the tab
You should have your model URL from Stage 3 saved. If you can't find it, open Teachable Machine, click Export Model, and grab a new one.
Build it
Step 1 — Open RAISE Playground
Click the RAISE Playground button at the top of this page. It opens in its own tab.
You should see a green flag, a red stop button, a stage with a cat sprite, and a left panel full of colorful blocks.
If RAISE asks you to sign in or create a project, do whatever your coach says for the camp. If it just opens to an empty project, that's fine.
Step 2 — Find the Add Extension button
Look in the bottom-left corner of the screen. You should see a button with a + sign or a stack of blocks icon — that's Add Extension.
Click it.
A new screen slides up showing many extensions: pen, music, video, text, and a section for AI.
Step 3 — Pick the Teachable Machine extension
Scroll through the extensions and find one named something like Teachable Machine or Image Classifier. Click on it.
The screen goes back to the main editor. You should now see new blocks in the block palette on the left — they have a special color (often pink or purple) and they mention "Image" or "Class".
Step 4 — Paste your model URL
Find the block that says something like "Use model from URL" or "Load model". Drag it into the script area.
Click on the input field inside the block and paste your Teachable Machine URL — the one you saved in Stage 3.
Make sure the URL ends with a / like this:
https://teachablemachine.withgoogle.com/models/AbC123xyz/
Click the block once to run just this block. If your URL is correct, you should see a small status indicator (a check mark, a "model loaded" message, or the sprite reacting). The model is now loaded into RAISE.
Step 5 — Prove RAISE can read your AI
Find the AI block that returns the current prediction. It may say something like prediction, class, or what do you see?
Drag that block into the script area. Click the block by itself. You do not need the green flag yet.
Make a Rock hand sign. The block should report Rock. Try Paper or Nothing too.
If RAISE can report the same words you trained in Teachable Machine, the two tools are connected.
This is the moment the bridge becomes real. Teachable Machine is not on the screen anymore, but RAISE is using the model you trained there. The AI prediction block is how the game asks the model what it sees.
Step 6 — Save your project
Right now, if you close this tab, you lose everything. Save the project:
- Click File → Save (or Save Now, depending on RAISE's version).
- Give it a name. Use something like:
AI Collector — First Last
Saving locks in the loaded model URL too. Tomorrow when you reopen the project, your AI is still connected.
Pacing Lab
This lab is required before Stage 5. You need to know where the AI blocks are before you start building with them.
Part A — AI block scavenger hunt (15 minutes)
Find and point to:
- The block that loads or connects your model.
- The block that reports the AI prediction.
- Any block that shows confidence or class names, if your RAISE version has one.
- The regular Scratch-style say block.
- The regular Scratch-style forever block.
Do not move on until you can find the AI prediction block twice in a row without help.
Part B — Connection proof (15 minutes)
Run the project and make each sign.
Write:
Rock sign showed: __________________
Paper sign showed: _________________
Scissors sign showed: ______________
Nothing showed: ____________________
If one answer is wrong, fix the model URL or class spelling before Stage 5.
Understand it
A URL is just an address. The model URL points to the place where your trained AI lives. RAISE Playground reads the URL and loads the AI. Then the AI blocks can ask it questions.
This works like sending a video link to a friend. The link tells your friend's browser where to find the video. The model URL tells RAISE where to find your AI.
The extension is the part of RAISE that knows what to do with a Teachable Machine URL. Without the extension, the URL is just letters. With the extension, RAISE can load your AI.
The thing we didn't do today is the most important one to notice. We did not write the game loop yet. Today was the handoff: model URL in, prediction block out. The game starts in Stage 5.
Try this
Try this
Three short experiments. Predict before you run, then test your guess.
What do you think will happen if you paste the wrong URL — like a YouTube link, or a Google search? Predict, then try it (be careful, you'll need to put the real URL back).
Look at the AI blocks in your block palette. Find the one that says something like "what class is the AI seeing" or "prediction". Now look at the regular Scratch blocks (the colored ones, not the AI ones). What does that AI block return — a number, a word, or something else? Hint: your green bars from Stage 3 showed it as a word.
Stage 5 makes the cat sprite say what the AI sees. There's a Scratch block called say. There's an AI block called something like prediction. Look at both blocks. How do you think we'll combine them?
Test your stage
- RAISE Playground is open in its tab.
- The Teachable Machine / AI extension is added (you can see special pink/purple AI blocks in the palette).
- You pasted your model URL into the Use model from URL block.
- When you clicked the block, the model loaded without an error.
- Clicking the AI prediction block returns one of your class names.
- You saved the project with your name.
- Design check. Switch to the Teachable Machine tab and back to RAISE. Can you find both tabs in under five seconds? You'll be switching a lot from now on.
If it breaks
- I don't see an AI extension in the Add Extension menu. RAISE Playground sometimes updates the names. Look for Image Classification, Teachable Machine, or AI. If none exist, raise your hand — the coach may need to enable it for the camp.
- The model URL paste didn't do anything. Make sure the URL ends with a
/(forward slash). If it doesn't, add one. Click the block to run it again. - Error: "Could not load model." Open the URL in a brand-new browser tab. If it doesn't load there either, the URL is broken — go back to Teachable Machine and export a new one.
- My project disappeared after I closed the tab. You didn't save. Reopen RAISE, recreate the steps, and save before closing the tab next time.
- The new AI blocks look totally different from a video I watched. Different versions of RAISE name blocks slightly differently. The behaviors are the same — load a model from a URL, ask the model what it sees. Find blocks that do those two things.
This is the hardest stage of the week because the tool has several steps. RAISE Playground changes sometimes, so the exact button names may not match this page. Have a backup plan: record a 60-second video of the current Add Extension flow and play it for the room.
The most common failure: the URL is missing the last /. Walk the room and confirm every camper's URL ends in /. This bug can take 20 minutes to find in real time.
If a camper finishes early, push them into the medium stretch (explore every AI block) — it primes them for Stage 5 and means tomorrow goes faster.
If RAISE Playground itself is down (it happens), the backup is paper-based: have campers draw the AI Collector Game by hand. Stage 4 is the only stage that requires RAISE to be online.