Qwen: layered image editing with AI

ALSO: How to write a LinkedIn post in minutes with AI

Author

Zain Kahn
December 22, 2025

Welcome back, Superhuman. For all their power, AI image generators still can't quite match the surgical control of traditional design software — yet. But new releases like Qwen’s layered editing model are slowly chipping away at that moat.

Today: Tweak ChatGPT’s personality, learn how to write a LinkedIn post in minutes with AI, and get the latest prompts and trending social posts.

TODAY IN AI

Click to learn about Qwen’s new layered editing AI model. Source: Qwen.

1. Alibaba’s Qwen launches new image model with layered editing: Qwen-Image-Layered lets users break images down into separate, editable RGBA layers to provide more specific, prompt-based instructions. This layering feature allows users to edit, recolor, or swap one element without impacting the rest of the image — just like most traditional design tools.

2. OpenAI releases ChatGPT tone updates, in talks for funding: OpenAI now lets users personalize ChatGPT’s conversation settings by adjusting factors like warmth, tone, enthusiasm, or emoji use — in addition to selecting personality types like quirky, professional, friendly, and more. To find this update, go to Settings —> Personalization in ChatGPT. In other news, OpenAI is also reportedly trying to raise up to $100 billion in a round that would value the AI leader at up to $830 billion.

3. AI pioneer reveals secretive new startup: Meta’s former VP and Chief AI Scientist, Yann LeCun, officially announced his new startup on LinkedIn: Advanced Machine Intelligence (AMI Labs). LeCun will serve as executive Chairman, with Alex LeBrun (co-founder and CEO of Nabla) taking over as CEO. The startup will focus on world models, and is reportedly trying to raise roughly $586 million at a $3.5 billion valuation before launching.

How do top companies like Yahoo and IKEA explain complex info to employees and customers? They use Guidde, the AI that turns workflows into powerful How-To videos instantly.

Your first 25 videos are 100% no-cost

  • Go to Taplio and sign up.

  • Click “Write a Post” and open AI Assist. Describe what you want to write about and who it’s for.

Sample Prompt: Based on my LinkedIn profile, past posts, and target audience, suggest 10 post ideas that would resonate with my audience right now. Prioritize topics that spark conversation, reflect my expertise, and fit my usual tone of voice.

  • Pick one idea and ask AI Assist to create a clear outline with strong hooks and key parts of the post body

  • Refine the outline and ask AI Assist to draft the full post.

  • Edit for clarity and personality.

  • Click “Schedule Post” to publish directly from Taplio.

P.S. We asked the team at Taplio if we could get a discount code for our readers, and they’re giving you 30 days for just $1 with the code SUPERHUMAN1X1.

FROM THE FRONTIER

One of AI's most respected voices just published his key insights from 2025 — Here’s the TL:DR.

Made with Midjourney

Andrej Karpathy's 2025 LLM Year in Review has racked up 2.2 million views — with four major takeaways you should know.

There’s a new training recipe for LLMs. Reinforcement Learning from Verifiable Rewards (RLVR) emerged as a breakthrough new way to train LLMs. The training system works by teaching LLMs how to reason, without constant human feedback. Karpathy estimates that this change — not bigger model sizes — powered most of 2025’s capability gains.

AI is a genius…and an idiot. LLMs are extremely capable where rewards are verifiable, but remain shockingly easy to trick elsewhere. This result: AI that can solve PhD-level math problems, but gets fooled by a simple trick question. Karpathy calls this "jagged intelligence” and cites it as one of the reasons why he’s losing faith in benchmarks.

Vibe coding has changed who can build software. 2025 marked the year that barriers to coding were removed almost entirely. Vibe coding empowers regular people to start writing software and programming professionals to write a lot more. Karpathy estimates that this shift will end up benefiting regular people much more than corporations or governments.

Claude and Gemini separate themselves. Karpathy called Google Gemini Nano Banana “one of the most incredible, paradigm-shifting models of 2025” and Claude Code “the first convincing demonstration of what an LLM agent looks like.” But while these models have made massive strides, he believes the industry has still only achieved 10% of its potential.

AI is a must-have for nearly every enterprise. But should you buy AI or build it in-house?

This guide helps you decide while balancing innovation, control, and speed.

  • Lessons from failures (and what to do differently)

  • How to combine the speed of buying with the precision of building

IN THE KNOW

What’s trending on socials and headlines today

Meme of the day

📝 Gemini Thread: Explore 10 wild use cases for how people are using Gemini, including live dashboards, robot simulation, and personal learning tools.

🤯 Claude Chrome Ext: Learn the internet’s new favorite productivity hack (this post has 2,200 bookmarks).

📸 Instant Editor: Gemini can instantly resize almost any photo if you use this prompt:

📈 Opus 4.5: A user on X posted “one of the most important charts in AI” that showcases just how fast AI is expanding.

😺 AI Viral Video: This car crash seems avoidable at first — until you realize who’s driving the car.

PRODUCTIVITY

5 New & Trending AI Tools

  • 💼 The AI Consultancy Project*: Turn your AI expertise into a 6-figure consultancy as the market explodes. We’ve trained 1,200+ consultants – apply now for early access.

  • 🧑‍💻 Antigravity: Google’s agentic development platform, evolving the IDE into the agent-first era.

  • 📱 Loki.Build: Design and ship studio-grade landing pages with AI.

  • 📷️ CyberCut: Create viral social clips with AI.

  • 💁 Incredible: Create AI agents that work 24/7 for you.

* indicates a promoted tool, if any

PROMPT STATION

Image-to-Insights

[upload image]Analyze the uploaded image or chart in detail.
Summarize the key patterns and trends you observe.
Identify any anomalies, outliers, or unexpected data points, and explain why they matter.
Translate these observations into 3 clear, practical recommendations that someone could act on immediately.
If assumptions are required, state them explicitly and keep the insights concise and decision-focused.”

The 9-to-5 Zoo

Source: yelyzera

Midjourney Prompt: A hyperrealistic studio photograph of a stylish dog wearing round sunglasses sitting upright at a table while typing on a laptop, minimalist indoor setting with a clean white wall and soft pastel tabletop, ultra-detailed fur texture, visible skin pores on the nose, realistic reflections in the sunglasses and laptop screen, humorous yet sophisticated mood, bright diffused professional studio lighting during daytime, shot with a Canon EOS R5, 50mm f/1.8 lens, natural neutral color palette with warm beige tones, extreme sharpness, cinematic depth of field, lifelike realism --ar 3:4 --style raw

Tip: Use the same prompt in Gemini or ChatGPT without the parameters (at the end) to get similar results

Whenever you’re ready to take the next step

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.

What did you think of today's email?

Your feedback helps me create better emails for you!

Login or Subscribe to participate in polls.

Until next time — Zain & the Superhuman AI team