Kimi K2 Outsmarts GPT-5

Hello there! Ready to dive into another upgrading, Mind-boggling, and Value-filled Shift?

Today we have:

🧠 Kimi K2 Thinking Redefines Open AI Reasoning

👀How to Make Scroll-Stopping UGC Ads Using Google Opal

🤖 The State of AI 2025: Pilots, Agents, and Impact

🔨Tools and Shifts you cannot miss

🧠 Kimi K2 Thinking Redefines Open AI Reasoning

Moonshot AI unveiled Kimi K2 Thinking, an open-source reasoning model rivaling GPT-5 and Claude 4.5. Built for stepwise problem-solving, it handles 200–300 tool calls autonomously and achieves state-of-the-art scores in reasoning, coding, and search benchmarks, all at under $5M training cost.

The Shift:

1. Reasoning Performance - K2 achieved 44.9% on Humanity’s Last Exam, surpassing GPT-5’s 41.7% and doubling human web search accuracy on BrowseComp (60.2% vs 29.2%). It reasons coherently across hundreds of steps, solving PhD-level math and multi-domain analytical challenges through structured tool use.

2. Agentic Coding and Search - The model reached 71.3% on SWE-Bench Verified and 61.1% on SWE-Multilingual, demonstrating strong multilingual coding precision. It autonomously cycles through thinking, browsing, and coding phases, capable of building full web components.

3. Efficiency and Creativity - Powered by INT4 Quantization-Aware Training, it delivers twice the inference speed with minimal accuracy loss. In creative writing, it produced 1,595 thinking tokens per sentence, showcasing narrative depth and contextual fluidity far beyond typical models.

4. Economics and Global Significance - At under $5M training cost, K2 Thinking challenges frontier AI economics. Accessible via API or OpenRouter for pennies, it signals China’s open-source labs reaching near-frontier parity.

Kimi K2 Thinking isn’t just another open model, it represents a shift toward affordable, scalable, and self-reasoning AI. With elite benchmarks, low cost, and human-like depth, it redefines what open-source AI can achieve in global competition.

^{TOGETHER WITH MORNING BREW}

Your career will thank you.

Over 4 million professionals start their day with Morning Brew—because business news doesn’t have to be boring.

Each daily email breaks down the biggest stories in business, tech, and finance with clarity, wit, and relevance—so you're not just informed, you're actually interested.

Whether you’re leading meetings or just trying to keep up, Morning Brew helps you talk the talk without digging through social media or jargon-packed articles. And odds are, it’s already sitting in your coworker’s inbox—so you’ll have plenty to chat about.

It’s 100% free and takes less than 15 seconds to sign up, so try it today and see how Morning Brew is transforming business media for the better.

Check it out

👀 How to Make Scroll-Stopping UGC Ads Using Google Opal

Creating authentic UGC-style ads used to mean juggling 5 tools and praying it syncs. Google Opal changes that. Here’s how to go from static assets to UGC-style video in one smooth flow:

🧠 Step 1: Open Google Opal and hit Create New. This platform lets you build AI workflows without code.

✍️ Step 2: Write Your Prompt - Example prompt:

“Build an AI app that takes a product photo and a user photo, generates a realistic image of the user holding the product, scripts a short review, animates it into a 5s UGC video with voiceover and music, and outputs a download link.”

🖼 Step 3: Upload Your Assets

Provide:

• A clear product image

• A photo of someone chest-up, with visible hands (this person becomes your ‘reviewer’)

🛠 Step 4: Let the AI Stack Run

Behind the scenes:

• Nano Banana handles image compositing

• Gemini 2.5 writes the review script

• Lyria composes background music

• Veo animates the face, lips, and gestures into a UGC-style video

📥 Step 5: Download & Launch

In ~1 minute, you’ll have a polished, real-looking UGC ad ready to test. No video editing, no creator back-and-forth, no extra tools.

Why It Works: UGC converts because it feels real. With Opal, you don’t fake it, you just automate the best parts. Perfect for founders and marketers who want content velocity without compromising authenticity.

🤖 The State of AI 2025: Pilots, Agents, and Impact

McKinsey’s global AI survey of nearly 2,000 organizations shows that while 88% now use AI, most remain in pilot mode. Only a third have scaled adoption enterprise-wide, revealing a clear divide between experimentation and transformation as AI agents begin entering workflows.

The Shift:

1. Adoption and Scaling - Nearly two-thirds of organizations are still experimenting, and only 33% have scaled AI across business functions. Agentic AI is gaining attention, with 62% exploring and 23% scaling agents, primarily in IT and knowledge operations.

2. Measurable Impact - 39% of respondents report EBIT impact from AI, though just 6% saw over 5% gains. Top performers redesign workflows to enable cross-functional innovation, proving that AI’s real value comes from structural transformation, not tool adoption.

3. Innovation and Workforce - 64% credit AI for enabling innovation, while 32% expect job reductions and 13% anticipate growth. The strongest performers repurpose labor toward higher-value creative and strategic tasks instead of pure automation.

AI’s next phase hinges on agents, workflow redesign, and outcome-led scaling. As 2025 marks the shift from pilots to performance, success will depend on embedding AI into core operations to drive lasting innovation and efficiency. 

^{TOGETHER WITH RIPPLING}

Don't get SaaD. Get Rippling.

Disconnected tools don’t speed you up – they slow you down. That’s SaaD (Software as a Disservice).

Rippling unifies HR, IT, and Finance so you can save time, cut costs, and grow faster.

Find out if you and your team are at risk of SaaD today.

How SaaD is your org?

🔨AI Tools for the Shift

🗂️ WinFeedback – Transforms messy client notes into clear, structured, and quantified tasks with priorities, time estimates, and one-click sync to your project boards.

🎨 Kolors AI – Your AI design assistant that transforms ideas or existing images into fresh, creative visuals instantly.

🌐 VibeTrans – Offers fast, accurate, and context-aware translations across multiple languages and dialects for global communication.

💬 Linkeddit – Finds Reddit users actively discussing your niche, delivering warm leads who are already seeking solutions like yours.

🧩 UndetectedGPT – The go-to platform for generating text that consistently passes AI detection tools with human-like fluency

🚀Quick Shifts

⚖️ Seven families are suing OpenAI in California, alleging ChatGPT caused emotional harm that led to four suicides. OpenAI recently added parental controls and AI distress detection updates to mitigate risks. 

🏗️  OpenAI asked the Trump administration to extend the 35% Chips Act tax credit to AI data centers, servers, and grid components, arguing it would reduce capital costs and accelerate U.S. AI infrastructure growth.

🧠 Google Cloud added major upgrades to Vertex AI Agent Builder, enabling developers to build, scale, and govern AI agents with new context control, observability dashboards, evaluation layers, and native agent identities for security and compliance.

💔 After his $1T Tesla pay approval, Elon Musk used xAI’s Grok Imagine to generate a video of an AI woman saying, “I will always love you,” prompting mixed reactions and online ridicule. 

🎬 ‘Breaking Bad’ creator Vince Gilligan declares his new Apple TV show ‘Pluribus’ was “made by humans.” Gilligan emphasized his anti-AI stance, calling AI “the world’s most expensive plagiarism machine,” and added a clear disclaimer in the show’s credits to highlight its fully human creation.

📂 Google launches File Search Tool, a fully managed RAG system that grounds Gemini in users’ data, offering an integrated, scalable way to deliver more accurate, context-aware AI responses across applications.

🤳AI Nugget of the Day

That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition let us know how much:

Forward it to your pal to give them a daily dose of the shift so they can 👇