Instagram | Sponsor this newsletter | Let’s Connect!

My fellow AI explorers

It’s been a quietly explosive week in the world of AI.

No splashy keynotes. No headline-grabbing demos. Just the kind of seismic shifts that slowly but surely change everything—new frontier models, deeper industry insights, and a serious reframe of what progress in AI actually looks like.

In today’s edition:

• 🧠 O3 & O4 Mini: OpenAI’s Next Leap in Reasoning and Tool Use

• 🧑‍💼 Mustafa Suleyman on AI’s Future: From Agents to Desktop Automation

• 💸 Training, Trust, and Tradeoffs: Why Bigger Models Aren’t the Only Bet

Must See AI Tools

• 💰 Payman: AI That Pays Humans. Over 10,000+ signed up for the beta

• 💫 SubMagic: An AI tool that edits short-form content for you! (Get 10% off using code “uncoverai” at checkout)

• 🎤 11Labs: #1 AI voice generator (Click Here to get 10,000 free credits upon signing up!)

• 🤖 ManyChat: Automate your responses & conversations on IG, FB and more! (Click Here to get first month for free)

• 🎙️ Syllaby: The only social media marketing tool you’ll ever need - powered by AI! (Get 25% off the first month or any annual plan with code “UNCOVER” at checkout)

Get discounts on popular AI tools here!

Microsoft

🧑‍💼 Microsoft’s AI Chief on What’s Actually Next for AI

In a rare and wide-ranging interview, Mustafa Suleyman, CEO of Microsoft AI and co-founder of DeepMind and Inflection, laid out his vision for what’s coming after large language models—and how we’ll get there.

Here’s what stood out:

• Beyond just bigger models: Suleyman believes scale still matters—larger models + more compute will continue to yield gains—but it’s no longer just about size. The new frontier? Training smaller models with synthetic, high-quality feedback from larger ones. Think: AI teaching AI.

• Tool use is the real unlock: LLMs with the ability to use tools are more than just chatbots—they become agents. From web searches to desktop automation, this is what moves AI from passive responder to proactive problem solver.

• Hallucinations aren't the apocalypse: He reframes "hallucinations" as a fuzzy form of reasoning—more feature than bug when used creatively. And the good news? Modern models are already far easier to steer, cite sources, and self-correct.

• Actions are coming: The next big leap? CoPilot Actions—AI that sees your screen, clicks your tabs, buys your flights, and fills in your forms. Desktop automation with natural language as the driver.

🧠 Key idea: We may already have the kernel of AGI—not in some mythical sentient machine, but in networks of agents using tools, learning from each other, and evolving with every interaction.

🔮 Prediction: The near future won’t be “one big AGI.” Instead, it’ll be a web of capable, teachable, fast-moving AI agents that orchestrate workflows, navigate apps, and augment every aspect of knowledge work.

📺 Watch the full interview with Mustafa Suleyman for a glimpse into the future of AI—from someone building it at the highest level.

Open AI

🧠 o3 & o4 Mini: OpenAI’s Newest Reasoning Powerhouses Are Here

OpenAI just rolled out two new frontier models—O3 and O4 Mini—and while they weren’t hyped up with the usual fanfare, they’re quietly reshaping what’s possible with reasoning, coding, vision, and autonomous tool use.

Here’s what you need to know:

• O3 is OpenAI’s most powerful reasoning model to date, excelling in coding, science, and visual tasks.

• O4 Mini is a faster, smaller version—optimized for cost-efficient reasoning, ideal for lightweight workflows.

• Both models feature agentic tool use, letting them autonomously chain tasks like web searches, code generation, or image analysis using multiple agents.

What makes these models special isn’t just their output—it’s how they think. In one demo, O3 identified a restaurant and its location from a blurry, logo-less photo of a menu—by parsing dishes, zooming in for details, scraping TripAdvisor, and confirming pricing across platforms.

In another, it solved a maze by analyzing color data, mapping paths with BFS algorithms, and rendering the solution graphically—all in under a minute.

📊 Benchmark highlights:

• Creative writing: O3 now ranks #1.

• Math Arena: O4 Mini High scores the highest overall with 89%.

• Finction memory: O3 retains 100% accuracy on long-form comprehension.

• But… hallucination rates remain a concern: 6.8% for O3, 4.6% for O4 Mini—much higher than Gemini 2.5 Pro.

🛠️ Developer access is already live via API, and O3 Pro with full tool support is expected in the coming weeks. Free-tier users can try O4 Mini via the “Reason” toggle in ChatGPT.

🔮 Prediction: These models mark a shift from passive assistants to autonomous research agents. As tool use improves and hallucination rates drop, expect AI workflows to become radically more capable—and independent.

📺 Watch the full video breakdown for real-world examples.

USA

💸 Tariffs and Tensions: The AI Investment Landscape Under Pressure

The booming AI investment scene faces new challenges amid escalating U.S.-China trade tensions. The imposition of 145% tariffs on Chinese goods by the U.S. has disrupted supply chains critical to AI infrastructure.​

Key developments:

• Major tech firms like Alphabet and Microsoft have reaffirmed significant capital investments, totaling $155 billion for 2025.

• However, some companies are pausing or abandoning data center projects due to oversupply and rising costs.

• The situation raises concerns about the sustainability of rapid AI expansion.​

These economic pressures could reshape the AI industry's growth trajectory, emphasizing the need for strategic resilience.​

Read full story

30-Second AI Play

🎬 Build a Custom AI-Generated Stream Intro (Music + Animation)

Want a unique, head-nodding intro for your livestream—complete with custom music and looping visuals? Here's how to do it using AI:

🧠 Step 1 – Brainstorm Your Theme in ChatGPT
Use a voice prompt to ideate with ChatGPT (or O3 Mini High) on your show’s vibe. Include:

• Desired mood (e.g., upbeat 90s retro)

• Use case (live stream intro)

• A format for naming + prompt crafting

🎶 Step 2 – Create Music with Udio
Upload your reference track—or write a custom prompt. Udio generates multiple 2-minute instrumentals. You can:

• Fine-tune the prompt for mood + genre

• Try the new “Style Transfer” feature to remix an existing track

• Select your favorite, loopable version

🖼️ Step 3 – Animate Visuals with Midjourney + Photoshop
Generate your base image with Midjourney, tweak it in Photoshop, then animate it:

• Slightly alter the image (e.g., position shift or expression change)

• Loop between versions for subtle motion

• Overlay your music and brand assets

💡 Pro Tip: Combine Midjourney image pairs with frame interpolation tools like Runway or After Effects for smoother loops.

🎥 Result? A stylish, custom-branded intro powered by AI—but directed by your creative vision.

Other Relevant AI News!

• Huawei's New AI Chip: Huawei is preparing to mass-produce a new AI chip, offering an alternative to Nvidia's offerings amid China's push for technological self-reliance. ​Look Here

• Salesforce's Agentic AI Blueprint: Salesforce has released a blueprint to help businesses scale AI agents, reflecting the growing demand for autonomous AI systems. ​Check It Out

• DeepMind's Geometry Solver: DeepMind's updated model now solves 84% of geometry problems accurately, a significant improvement from 54%, showcasing advancements in AI's problem-solving capabilities. ​See Full Post

• AI in K-12 Education: A draft executive order outlines plans to integrate AI into K-12 classrooms, signaling a move towards modernizing education through technology. ​Read Here

• Economic Impact: The IMF reports that AI's economic gains are likely to outweigh the emissions costs associated with data centers, highlighting AI's potential to boost global output. ​Learn More

Golden Nuggets

• 🧠 OpenAI’s O3 & O4 Mini show just how far agentic tool use and reasoning have come—models that think, search, and solve with serious autonomy.

• 🧑‍💼 Mustafa Suleyman’s vision points to a future of AI agents that work across your desktop, not just inside your chat window.

• 🎧 You can now build entire creative pipelines—from custom theme music to animated stream intros—using AI tools like ChatGPT, Udio, and Midjourney.

Until our next AI rendezvous,

Anthony | Founder of Uncover AI