- Horizon AI
- Posts
- How to Use Canva's New AI Tool to Turn Flat Images into Editable Layers ✨
How to Use Canva's New AI Tool to Turn Flat Images into Editable Layers ✨
Anthropic Says Claude Now Writes Most of Its Code and Calls for an AI "Pause Button" 🛑
Gina Acosta
June 05, 2026
Sponsored by
Welcome to another edition of Horizon AI,
One of the biggest frustrations with AI-generated images and flattened designs is that even a small edit can turn into a time-consuming process, often requiring you to start over from scratch. In today's issue, we look at a new Canva feature that aims to change that, automatically breaking any image into individual editable elements you can move, resize, replace, or remove.
Let’s jump into it!
Read Time: 4.5 min
Here's what's new today in the Horizon AI
Anthropic Reveals Claude Writes Most of Its Code and Warns of Future AI Risks
OpenAI Upgrades ChatGPT’s "Dreaming" Memory System
AI Tutorial: Turn images into editable layers with Magic Layers
AI Tools to check out
AI Findings/Resources
The Latest in AI and Tech 💡
AI News
ANTHROPIC
Anthropic Reveals Claude Writes Most of Its Code and Warns of Future AI Risks
Anthropic has released new internal data showing that Claude is playing an increasingly large role in developing the company's own AI systems, while also warning about the risks of future self-improving AI.
Details:
Anthropic says more than 80% of the code merged into its codebase was authored by Claude.
Claude's coding quality has improved significantly, with the company claiming it has reached near parity with human-written code and could surpass it within the next year.
The company says the length and complexity of tasks AI can complete autonomously also continues to increase rapidly, with Claude now capable of handling projects that can span many hours without human intervention.
Despite these advances, Anthropic argues that key limitations remain: 'research taste and judgment, including choosing which problems matter, which results to trust, and when an approach is a dead end.'
The conclusion they arrive at is that recursive self-improvement (AI that fully autonomously designs its own successor) hasn't been achieved yet, but it 'could come sooner than most institutions are prepared for.' As a result, Anthropic is advocating for mechanisms that would allow AI developers to collectively and verifiably slow down or temporarily pause frontier AI development if safety concerns become severe enough.
TOGETHER WITH MODA
Moda is the AI design agent with taste
Moda's viral launch hit 4.4 million views in two days. Tens of thousands of professionals signed up. Startups, agencies, forward-thinking brands and top firms are now using Moda to create brand-aligned slides, ad creative, reports, social carousels and more.
Most AI tools tend to create what we call "AI slop": repetitions of the same colors, layouts and fonts. And when you try to fix it, you get stuck in a loop of re-prompting.
Moda is different. Drop in your website URL, and Moda learns your brand from the ground up: your colors, your fonts, your visual language. Then it helps you generate pro-quality slides, docs, and marketing assets.
The best part? Every layer is fully editable on a real canvas, and exports to powerpoint, PDF and more.
OPENAI
OpenAI Upgrades ChatGPT’s "Dreaming" Memory System
OpenAI has updated ChatGPT’s memory system with a new architecture designed to create richer, more accurate user profiles that automatically evolve over time.
Details:
Instead of storing individual facts as separate memory entries, ChatGPT now creates a structured profile organized into categories such as work, hobbies, travel, education, and personal interests.
The profile is automatically generated from conversations and continuously updated in the background, reducing the need for users to manually tell ChatGPT what to remember.
OpenAI says the new system is designed to solve the problem of stale memories by adjusting information as a user's situation changes.
The company reports significant improvements in memory performance, including better retrieval of past information, stronger use of personal preferences, and more up-to-date contextual understanding.
Users can review, edit, add, or remove memories through a dedicated summary page that shows what ChatGPT knows about them across different categories.
OpenAI says the upgraded Dreaming architecture is more efficient than the previous version introduced in April 2025, helping enable a broader rollout — starting with Plus and Pro users in the US, with free users getting access in the coming weeks.
AI Tutorial
Turn images into editable layers with Magic Layers
Use Magic Layers From the homepage
Select Magic Layers.
Click Select Media, then choose an image.
Wait while Magic Layers processes your image.
Select a layer to edit, move or delete it.
Use Magic Layers From the editor
Open or create a design.
Click 'Uploads' to upload or add an image to your design.
Select the image.
Select Edit from the editor toolbar.
Select Magic Layers.
Wait while Magic Layers processes your image.
Select a layer to edit, move, or delete it.
Once your image is separated into layers, you can:
Move individual elements to new positions
Delete unwanted objects from your design
Resize specific elements independently
Apply effects to individual layers
Replace elements with new images or graphics
This is especially useful for AI-generated images. Want to move something or change a small detail? Now you can do it directly, no more going back, rewriting the prompt, and regenerating from scratch.
Notes:
Magic Layers supports single-page PNG and JPG files only (to use it on a PDF, convert it to a PNG or JPG image first).
The feature works best with graphic designs, illustrations, and stylized visuals.
AI Tools to check out
🗣 Miso: Build voice agents your users will love. Real-time latency, one-shot voice cloning, and on-premises deployment.
🌐 CrafterQ: AI Agents for business websites. Turn your website into a conversational AI experience with custom agents trained on your content, products, and business knowledge.
📈 Veltrix: AI finance copilot for cash flow, margins, and growth.
⏱ Exmergo: Turn messy data into reliable dashboards in minutes.
📊 ChartCool: Create professional charts with one prompt
TOGETHER WITH MODA
Moda is the AI design agent with taste
Moda is an AI design product where you prompt what you need, get a complete on-brand design, and edit every element on a full canvas.
Our viral launch hit 4.4M views in days, tens of thousands signed up, and executives at major finance and tech companies now use it.
AI Findings/Resources
🍏 Apple's annual Worldwide Developers Conference 2026: Everything to expect
🤯 Agentic traffic is growing so fast that bots have now passed human traffic online for the first time in the Internet's history
🎬 The casting for Christopher Nolan's upcoming Odyssey is stirring up discussion. This Reddit user used AI to show what their dream cast would look like
🍁 Canada released its national AI strategy, outlining how the country plans to adopt the technology over the next decade
🤖 A developer trained an AI agent in simulation and deployed it onto a real robotic air hockey table, and it can now challenge even skilled human players
The latest in AI and Tech
Top AI executives, including Sam Altman, Dario Amodei, and Demis Hassabis, have signed a letter urging Congress to strengthen protections around synthetic DNA and RNA orders. They argue that advances in AI could lower the barriers to developing biological weapons, making stronger safeguards necessary.
A new study from Stanford Law School found that law professors preferred AI-generated answers over those written by fellow professors in 75% of head-to-head evaluations. Researchers tested AI systems, including Gemini 2.5 Pro and NotebookLM, using questions commonly asked by first-year law students.
The new 550-billion-parameter open MoE model is built to 'help long-running agents complete tasks faster while lowering costs.'
The United States and Japan have announced a new science and AI partnership, with each country committing $500 million to the effort. The initiative is part of the U.S. government's Genesis Mission, which aims to accelerate scientific discovery through AI.
The International Mathematical Union has endorsed the Leiden Declaration on AI and Mathematics, which warns that growing tech industry involvement could reshape the future of mathematical research. The declaration raises concerns about unreliable AI-generated proofs, improper attribution of human work, and increasing dependence on proprietary AI systems, calling for stronger safeguards, greater transparency, and policies that keep human judgment at the center of mathematics.
That’s a wrap!
We'd love to hear your thoughts on today's email!Your feedback helps us improve our content |
Not subscribed yet? Sign up here and send it to a colleague or friend!
See you in our next edition!
Gina 👩🏻💻