In partnership with

Welcome back!

There’s a new brain in town, and it’s not from OpenAI. Moonshot AI’s open-weight “Kimi K2 Thinking” just outperformed GPT-5 and Claude 4.5 on reasoning benchmarks. Let’s unpack…

Today’s Summary:

• 🔥 Kimi K2 Thinking beats GPT-5 in reasoning benchmarks

• 🎨 Google launches Pomelli marketing AI

• 🛰️ Google plans solar AI datacenters in space

• 🍎 Apple nears Gemini deal for Siri

• 📣 Altman addresses $1.4T rumors

• 🏆 LMArena launches Expert leaderboard

• 🛠️ 2 new tools

TOP STORY

Open-source Kimi K2 Thinking beats GPT-5 in reasoning benchmarks

The Summary: Moonshot AI has released Kimi K2 Thinking, a trillion-parameter reasoning model that outperforms GPT-5 and Claude 4.5 Sonnet on several key benchmarks. Built as a “thinking agent,” it can chain up to 300 sequential tool calls and sustain complex reasoning over a 256K context window. The model is open-weight under a modified MIT license requiring attribution only for massive scale commercial use.

Key details:

• Backed by Alibaba and led by ex-DeepMind researchers

• 1 trillion parameters, 32B active, native INT4 inference

• Estimated training cost ≈ $4.6 M

• Benchmarks: BrowseComp 60.2 % (↑5.3 vs GPT-5), HLE 44.9 % (↑3.2)

• Early testers report it feels “smarter than the benchmarks”

• Handles very long reasoning chains without collapse, solves PhD-level math via 20+ tool calls

• Modified MIT license requiring attribution only for massive scale commercial use

Why it matters: Open-weight models have stopped chasing the frontier and are now starting to set it. Moonshot’s K2 Thinking shows that what once needed billion-dollar compute budgets can now be built for under $5 million and still beat GPT-5 on reasoning tasks.

Model card

FROM OUR PARTNERS

Build Your Own AI Agents in Minutes

The Simplest Way to Create and Launch AI Agents and Apps

You know that AI can help you automate your work, but you just don't know how to get started.

With Lindy, you can build AI agents and apps in minutes simply by describing what you want in plain English.

→ "Create a booking platform for my business."
→ "Automate my sales outreach."
→ "Create a weekly summary about each employee's performance and send it as an email."

From inbound lead qualification to AI-powered customer support and full-blown apps, Lindy has hundreds of agents that are ready to work for you 24/7/365.

Stop doing repetitive tasks manually. Let Lindy automate workflows, save time, and grow your business

Get $20 Worth of Free Credits

GOOGLE

Google launches Pomelli for automated brand marketing

The Summary: Google Labs and DeepMind have launched Pomelli, an AI tool that helps small businesses to generate on-brand marketing campaigns. The system extracts a “Business DNA” from the existing brand, including tone, visuals, and color palette.

Key details:

• Analyzes a business website to extract brand identity elements

• Generates campaign ideas and full creative assets in three steps

• Designed for small marketing teams with no design staff

• Public beta limited to English users in the US, CA, AU, and NZ

Why it matters: Pomelli is a simple tool for producing consistent branded content without the need to juggle between many different tools. It turns brand guidelines into simple editable campaigns, matching the company tone and visuals. While still an early experiment, it gives small teams a way to produce marketing assets that stay on-brand.

Try it in Google Labs

FROM OUR PARTNERS

Work Smarter with This Free AI Tool

From Boring to Brilliant: Training Videos Made Simple

Say goodbye to dense, static documents. And say hello to captivating how-to videos for your team using Guidde.

1️⃣ Create in Minutes: Simplify complex tasks into step-by-step guides using AI.
2️⃣ Real-Time Updates: Keep training content fresh and accurate with instant revisions.
3️⃣ Global Accessibility: Share guides in any language effortlessly.

Make training more impactful and inclusive today.

The best part? The browser extension is 100% free.

Check out Guidde

RESEARCH

Google plans AI datacenters powered by the Sun in orbit

The Summary: Google has revealed Project Suncatcher, an experimental effort to scale AI computation beyond Earth. The company plans to test two prototype satellites by 2027, each carrying custom TPU chips powered directly by sunlight. The research envisions orbital constellations that act like space-based datacenters, where near-continuous solar energy replaces terrestrial grid power.

Key details:

• Satellites to orbit in sun-synchronous formation at 650 km altitude, receiving up to 8x more solar energy than Earth-based solar panels

• Trillium AI chips withstood radiation equal to 15 years in orbit

• Demo achieved 1.6 Tbps optical link between satellites

• Projected launch costs could fall to $200/kg by the mid-2030s, making orbital AI compute economically competitive with terrestrial centers

Why it matters: Space-based AI compute could ease Earth’s growing energy load. The Sun offers effectively infinite, uninterrupted high-density power, and harnessing it directly in orbit could sidestep the usual constraints of cooling, water, and grid.

Research paper

QUICK NEWS

Quick news

• Apple near deal with Google Gemini to power next-gen Siri

• Altman responds to rumors about $1.4T infrastructure push

• LMArena launches ‘Expert’ leaderboard to test AIs on pro prompts

TOOLS

🥇 New tools

• Floqer - AI automation + 75 sources to make your CRM up-to-date

• Sora 2 for Android - OpenAI’s generative-video app lands on Android