AI wearable lets paralyzed patients control robots by thought

Good morning, AI enthusiasts. UCLA engineers have developed a noninvasive brain-computer interface (BCI) that lets paralyzed users control robotic arms with their thoughts, without any surgery.

PLUS: How to turn your messy prompts into a highly structure ones

HOW TO AI

How to turn your messy prompts into a highly structure ones

OpenAI just launched its official Prompt Optimizer for ChatGPT, and it’s completely free. This tool takes your rough, messy prompts and transforms them into polished, structured prompts designed to get the best results from models like GPT-5.

🧰 Who is this for

• Researchers who want consistent, accurate outputs

• Developers running GPT-5 in production

• Founders and builders optimizing prompts for real apps

• Anyone looking to learn prompt engineering without trial and error

STEP 1: Access the Prompt Optimizer

Go to the OpenAI Playground. You’ll see a new option next to the prompt field called Optimize.

In the main text field, type your raw or messy prompt exactly as you would normally write it.

Here's the example: "Build me a train game where I can control the speed and distance between two stations and see the train running."

STEP 2: Optimize Your Prompt

Once you type your normal prompt, click on the "Optimize" button.

The tool instantly rewrites your prompt into a more structured, clear, and detailed version.

Instead of a single vague line, you’ll now see a refined structure that includes a clear role and objective, step-by-step instructions, and added context about what you want. Each section is neatly separated into paragraphs.

This ensures GPT-5 understands exactly what you’re asking for.

STEP 3: Compare Results

Copy the optimized prompt back into ChatGPT and run it. Compare the results against your original prompt.

You’ll immediately notice the difference in clarity, structure, and usefulness of the responses

STEP 4: Use it wisely

One important note: optimized prompts are often longer, which means they consume more tokens.

If you’re running GPT-5 in production, this can increase API costs. That’s why you shouldn’t overuse it for trivial requests. Save it for when quality matters most.

💡 Pro tip: Use the optimizer as a learning companion. Try writing your own improved version of a prompt first, then compare it with what the optimizer suggests. Over time, you’ll naturally start writing better prompts on your own.

TODAY IN AI

1. AI wearable lets paralyzed patients control robots by thought

UCLA researchers built a wearable BCI system that pairs EEG brain signals with AI-powered computer vision to help infer user intent and guide robotic arms or cursors.

→ Participants completed tasks significantly faster with AI assistance

→ A paralyzed participant moved blocks with a robotic arm in 6.5 minutes with AI

→ The AI-BCI uses noninvasive EEG, avoiding the risks of brain surgery

Why is important: This advancement shows real promise for giving people with paralysis or neurological conditions more independence. By combining AI “co-pilots” with BCIs, researchers may unlock safer, more practical assistive technologies, moving the field closer to everyday use outside the lab.

2. China enforces new rules on labeling AI-generated content

China’s new regulations on AI-generated content (AIGC) officially took effect today, requiring all platforms to label AI-produced text, images, audio, video, and virtual scenes.

→ Platforms must review and label AIGC before release

→ Implicit/explicit labels required depending on detection method

→ Risk notices must be added to suspected AI content

Why is important: China is moving faster than most countries in regulating AI, setting a global precedent for mandatory labeling. These rules could influence how other governments handle AI content and reshape how platforms fight misinformation.

ESSENTIAL BITES

• China’s President Xi Jinping urged deeper AI cooperation and rejected a “Cold War mentality” at the Shanghai Cooperation Organization summit, where he also proposed a new Global Governance Initiative.

• Microsoft launched a 10B parameter version of its VibeVoice text-to-speech model, capable of generating multi-speaker podcasts up to 45 minutes in minutes, with 32K context support under an MIT license.

• Tencent unveiled HunyuanWorld-Voyager, a long-range world model with 3D reconstruction and memory, alongside Hunyuan-MT-7B and Hunyuan-MT-Chimera-7B, a joint translation system outperforming rivals in 33 languages.

⚙️ Hot AI Tools:

• 🎥 Google Vids: Create and edit videos with AI-powered tools

• ⚙️ Grok Code Fast: xAI’s new SOTA coding model

• 🧑‍🏫 Anyimg: Transform your text into stunning visual artworks.

• 📽️ DrawToVideo: Turn sketches into living animations with AI.

That's it! See you tomorrow

- Dr. Alvaro Cintas