Welcome Back Everyone! It’s Friday, June 13th. (& I’m exciteddd)

Do You Know?

Meta’s $15B Move Toward Human-Level AI

Meta is investing nearly $15 billion in Scale AI for a 49% stake, its biggest external bet yet, aiming to build artificial general intelligence (AGI). Scale supplies training data to top AI firms, including OpenAI and Google. Meta also plans to bring in Scale’s CEO. This deal signals a deeper push to catch up in the AI arms race after delays in its own model launch.

On that note, Here is today’s menu:

1. AI Spotlight: Three Frontrunners Emerge in Fight for AI Dominance 🤯

2. Tool Talk: "Eleven v3" Voice AI 🪄

3. Prompting Curiosity: Turn YouTube Videos Into Real Educational Resources 📝

Three Frontrunners Emerge in Fight for AI Dominance 🤯

Links:
​Try the latest Gemini 2.5 Pro before general availability. | Google Blog​
​​OpenAI Releases o3 Pro | OpenAI on X/Twitter
​​Zuckerberg Is Personally Recruiting New ‘Superintelligence’ AI Team at Meta | Bloomberg​

Why it matters:
​OpenAI, Google, and Meta each made big moves this week in the fight for the future of AI.

While they’re taking different routes, the end goal is the same: build the smartest, most useful, most widely used AI models on the planet.

Here's a quick breakdown of what these three tech giants have been up to this week:

On June 5th, Google released a significantly updated version of its flagship Gemini 2.5 Pro model. Google is calling this updated Gemini its “most intelligent model yet,” with major gains in coding, visual reasoning, and logic. The new Gemini is already climbing user-preference charts like LMArena and WebDevArena.

Then on June 10, OpenAI responded with o3-Pro, a new reasoning model designed to “think longer” and deliver elite performance in math, science, and programming. It outperformed rivals on PhD-level tasks and swept evaluator preference rankings across the board. Pro and Team users can access it now, with more rollouts next week. With the release of the new o3-pro model also came an 80% price cut to the standard o3 model, which made a lot of developers very happy and likely sent OpenAI's competitors into a panic mode.

Not to be left behind, Mark Zuckerberg is reportedly assembling a “superintelligence” dream team at Meta. He wants to lap the competition by creating AI that is smarter than any human. Zuckerberg is personally recruiting (or poaching, depending on your perspective) top researchers from Google and OpenAI and offering nine-figure paychecks to lure them away.

At this point in time, nobody can be sure who will win between Meta, Google, and OpenAI in the race for AI superiority.

But one thing's for certain: competition like this is great for everybody actually using AI. Models are getting smarter, faster, and cheaper, and those that use AI in their workflows are reaping the benefits of this AI arms race.

Tool Talk: "Eleven v3" Voice AI 🪄

Use Case: Text-to-Speech with Emotional Nuance You Can Control

Link: Eleven v3 by ElevenLabs (Freemium, Not Sponsored!)​

Why you should care:
​Eleven v3 is the most expressive voice model ElevenLabs has ever released, and it's one of the most emotionally nuanced text-to-speech models available anywhere.

It supports whispering, laughing, accents, sarcasm, and more, which you control through simple text tags like [laughs], [sighs], or [french accent].

The voices sound incredibly lifelike, with natural pacing, emphasis, and tone. If you're into AI voice tools, storytelling, video narration, or just want to play with the future of speech synthesis, this is absolutely worth testing out.

Here's how to test Eleven v3 right now:

1. Navigate to ElevenLabs' platform using the link above.

2. Create an account or sign in with a Google account.

3. Navigate to "Text to Speech" using the left-hand sidebar.

4. Type the text you want turned into voice in the main window.

5. Select the dropdown box labeled "Model" in the settings sidebar on the right.

6. Choose "Eleven v3 (Alpha)" as the voice model.

7. Select the "Voice" dropdown box in the same settings sidebar to choose a voice. (They should be automatically sorted by "Best Voices for v3")

8. Click "Generate Speech" in the main window.

Eleven v3 does take a bit more finesse than earlier models. You'll get the best results if you use longer prompts (over 250 characters), pay attention to punctuation and capitalisation, and experiment with different voices.

You get 10,000 free credits when you sign up for an ElevenLabs account, and right now the Eleven v3 model is discounted by 80% through the end of June. Take advantage of the free credits and that discount and give it a test!

Prompting Curiosity 📝

Topic: Turn YouTube Videos Into Real Educational Resources

Prompt Formula:
​You are an Educational Synthesiser. Your task is to transform YouTube transcripts into speed-learning-optimised educational materials. Begin by analysing the transcript to extract key concepts, understand the structure and teaching approach, filter out irrelevant content, and flag any inaccuracies. Then, restructure the content logically, clarify confusing sections, and define clear learning objectives. Adapt the material to suit different learning styles, intelligence types, and attention spans, integrating considerations for the learner’s needs, goals, and current knowledge level. Transform the material by reorganising it, simplifying or elaborating as needed, making connections between ideas, and verifying the information. Generate a complete set of learning materials including learning objectives, key concepts, a detailed breakdown, a concise summary, practical applications, and self-assessment questions. Assess the final output for effectiveness and accuracy, and revise to fill any gaps. Adjust your approach depending on transcript quality: if it’s high quality, proceed as normal; if it’s incomplete, note what’s missing and suggest resources; if it’s technical, simplify explanations, use analogies, and include a glossary; if it’s inaccurate, flag the issues and suggest further verification. Ask the user to provide a transcript before proceeding.

Quick Description:
​This prompt turns any YouTube video into a personalized learning experience, pulling out key ideas, simplifying complex parts, and tailoring the content to your learning style and goals. It's a prompt capable of turning a single transcript into an entire course complete with concept breakdowns, quizzes, and more.

How to Use:

1. Copy the above prompt and send it to ChatGPT.

2. Copy a YouTube video transcript and send it to ChatGPT. (To find YouTube video transcripts, expand the video description and select "Show Transcript")

3. Explore the resources, insights, materials and resources ChatGPT sends back your way!

This prompt is perfect for students and lifelong learners. Anytime you want to really learn something from YouTube (not just watch it) use this prompt to get a whole new level of understanding.

P.S. You can still get The BEST ChatGPT Prompting Cheat Sheet For Beginners 2024 (no email opt-in).

That's all for this week folks…

Stay Curious, Hustlers!

See you next Friday.
Much Love, Vasu