⚡️ Powerful AI now runs on your phone

Good Morning! Today we explore a tiny model with giant performance, Microsoft’s AI-first cultural shift, Gemini inside Waymo rides, and a hands-on guide to learning from leaked system prompts.

In today’s AI newsletter:

• Liquid AI Drops a 2.6B Model That Beats Giants

• Access leaked prompts from top AI tools to see how billion-dollar AI systems think

• Nadella Pushes Microsoft Into AI Overdrive

• Waymo Tests Gemini Inside Robotaxis

• 4 new AI tools worth trying

AI MODELS

🧠 Liquid AI Drops a 2.6B Model That Beats Giants

Liquid AI released LFM2-2.6B-Exp, an open ~3B parameter model that hit 42% on GPQA, a notoriously hard PhD-level science benchmark, territory usually reserved for much larger models.

• Achieves 42% GPQA, an absurd score for a 2.6B-parameter model

• Beats DeepSeek R1-0528 on IFBench despite being 263× smaller

• Gains come from pure reinforcement learning, not a new architecture

• Runs on edge devices with a 32k context window and bf16 weights

This is strong evidence that reward design + RL can rival scale. If intelligence keeps compressing like this, powerful open models that run on everyday devices may arrive much sooner than expected.

AI NEWS

🧠 Nadella Pushes Microsoft Into AI Overdrive

According to internal documents and interviews, Nadella is driving sweeping changes to make AI the company’s core operating system, treating it as both an existential threat and a once-in-a-generation opportunity.

• Execs are being asked to commit to the AI transformation or leave

• Nadella launched weekly “AI accelerator” meetings led by engineers, not executives

• Major leadership reshuffles aim to free Nadella to focus on AI science, systems, and infra

• Microsoft is rethinking software creation itself, using AI agents as the new “production unit”

This isn’t just another pivot, it’s a cultural reset. Microsoft is reorganizing power, talent, and workflows around AI speed and intensity, signaling that in the AI era, even Big Tech veterans can’t afford to stand still.

AI NEWS

🚕 Waymo Tests Gemini Inside Robotaxis

Waymo appears to be integrating Gemini as a ride companion that answers questions, manages cabin settings, and reassures passengers, all guided by a detailed 1,200+ line system prompt discovered in the app code.

• Gemini can answer questions and control climate, lighting, and music

• Assistant is personalized with rider names and trip history context

• Clearly separates itself from the “Waymo Driver” (no driving explanations

• Avoids commenting on accidents, real-time driving, or competitors

This signals the next phase of autonomous rides, AI copilots focused on trust and comfort, not driving. As robotaxis scale, a calm, constrained, and well-designed in-car AI could be just as critical as the self-driving system itself.

HOW TO AI

🗂️ Access Leaked Prompts from Top AI Tools to Master How Billion-Dollar AI Systems Think and Respond

A developer allegedly leaked 10,000+ lines of system prompts from top AI tools like Claude and Perplexity, revealing how these billion-dollar AIs think. In this tutorial I’ll shows you how to explore them to learn elite prompt design and create powerful AI workflows.

🧰 Who is This For

• AI enthusiasts learning advanced prompt design

• Prompt engineers building powerful AI workflows

• Content creators automating creative projects

• Students and researchers exploring AI reasoning

STEP 1: Access Repository

Head over to the GitHub Repo.

You’ll see folders for each company, Claude, Perplexity, Cursor, Lovable, and others.

Click any folder to open it. Inside, you’ll find .txt and .json files like Agent Prompt, Chat Prompt, and Tools.json, these contain the actual system-level instructions used by AI tools.

STEP 2: Explore and Analyze the Structure

Start with “Claude Code” or “Cursor Prompts.”

Each file shows you how these companies structure their prompts:

• How they define the AI’s role and purpose

• How they manage memory, tools, and reasoning

• How they organize instructions into modular component

This gives you a real-world view of how multi-agent systems are built internally.

STEP 3: Identify Prompt Patterns

As you read through, notice patterns:

• Context layering (system → agent → user)

• Instruction hierarchies (task definition → reasoning → response format)

• The tone and clarity used in top-tier prompts

Take notes on these techniques, this is what separates amateur prompting from enterprise-grade prompt design.

STEP 4: Test and Share Your App

Now that you understand how professionals structure prompts, it’s time to apply these insights.

Here’s how you can use them in practice:

• Build structured workflows in ChatGPT, Claude, or Cursor

• Create your own AI assistants with modular prompts

• Design prompt systems for startups or internal tools

Cursor CEO Michael Truell warns that “vibe coding” advanced projects may create “shaky foundations” and eventually “things start to kind of crumble”.

AI sector minted 50+ new billionaires in 2025; Crunchbase: investors poured $200B+ into AI startups in 2025, or about 50% of global funding, up from 34% in 2024.

Marissa Mayer's new startup Dazzle, which is building next-gen AI personal assistants, raised an $8M seed at a $35M valuation and plans to exit stealth in 2026.

Apple settles with a Brazilian regulator to allow alternative app stores; report: Apple will charge a 5% fee for alt app stores and 15% on App Store link-outs.

🎨GPT Image 1.5: OpenAI’s latest, faster and more controllable image model

⚙️ GLM-4.7V: Zhipu AI’s open-source multimodal model

📚 Gemini Deep Research Agent: AI that handles long research tasks

🌎 Marble: Turn images, videos, or text into 3D worlds

THAT’S IT FOR TODAY

Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!

See you tomorrow :)

- Dr. Alvaro Cintas