Latest Update
China AI Just Flipped the Cursor Game
2026.05.19Hey, it's the Carat team. 🤗
Cursor just flipped the coding tool market with a Chinese AI model. Claude Design now handles longer and more complex work, and Anthropic acquired the core infrastructure behind MCP.
Cursor just flipped the coding tool market with a Chinese AI model. Claude Design now handles longer and more complex work, and Anthropic acquired the core infrastructure behind MCP.
China AI Just Flipped the Cursor Game
Cursor launched Composer 2.5, built on Kimi K2.5 — a model made by Chinese AI company Moonshot. It matches GPT-5.5 and Opus 4.7 in performance, at one-tenth the price.
It's better at staying on task through long workflows and following complex instructions accurately. Free usage is doubled for the first week after launch.
The era of US AI models dominating coding tools is fading. When performance is comparable, price and accessibility become the deciding factors. Developers now have far more options to choose from.
📌 3 Things to Know Today
1️⃣ Claude Design: Now You Can Write Way More
Anthropic doubled the token limit for Claude Design across all plans. You can now fit longer design briefs, more complex UI descriptions, and more context in a single session.
What makes an AI design tool competitive isn't ideas — it's how much context it can handle at once. Double the tokens means a fundamentally different scale of project.
2️⃣ Anthropic Just Bought the Core of MCP
Anthropic acquired Stainless — a platform that auto-generates SDKs and MCP servers for TypeScript, Python, Go, and more from an API spec. Anthropic's own API SDK was built by Stainless from day one.
MCP (Model Context Protocol) is how AI agents connect to external tools. Anthropic just brought the core infrastructure company of that ecosystem in-house. It's a fight over who controls the standard for AI tool connectivity.
3️⃣ They Posted Their Idea on Threads. Someone Built the App.
A user in the Korean Threads community shared their idea for an app: couples submit their fight to an AI judge, who gives a verdict and suggests a reconciliation mission. They even had a panda judge character and were building excitement with the community as they developed it.
Not long after, a nearly identical app appeared from another account — same concept, similar UI, similar marketing tone, with no mention of the original. The community reaction was cold. With AI vibe-coding, anyone can see an idea and ship it in a day. A public idea is now a public resource.
In an age where anyone can copy fast, the real edge is execution, branding, and community trust. If you took inspiration from someone, say so — and make sure your version has a clear, distinct point of difference.
🧪 Prompt Tip of the Day
Turn Your Photo into a Crayon Drawing by a 3-Year-Old
Threads creator hochcori's "terrible crayon drawing" prompt went viral with 4,800 likes and 240,000 views. It transforms your photo into something a 3–4 year old might have drawn with crayons — endearingly messy and hard to identify.
Here's what we made on Carat:
Use ChatGPT Images 2.0 Medium model with the prompt below to generate an image.
Transform the attached image into a drawing that looks like it was made by a 3–4 year old child with crayons — messy, uncertain, and charmingly clumsy. The child seems to have been trying to draw something, but you can only half-tell what it is. The composition and perspective should feel childlike and random, not adult or intentional. Embrace the adorable chaos.
Upload your photo with the prompt above on Carat's ChatGPT Images 2.0 and try it yourself.
That's all for today. See you tomorrow with more AI news. ☺️
Was this helpful? Share with your friends
Share
Get daily AI trend updates
