⛓️ Unshackled AI model

Hey folks,

GPT-5 is now a little friendlier and warmer, and it’s already making chats feel more like banter and less like boardroom boilerplate. After feedback that the model sounded too formal, OpenAI slipped in subtle touches like “Good question” and “Great start” to warm things up without ramping up the flattery. Internal checks confirmed that the tone shift didn’t edge back into “yes-man” territory so your new GPT-5 might feel a bit more human now as they respond to feedback.

Let’s dive in..

_{Read Time: 5 minutes} 

🎨 Qwen-Image-Edit: Smarter Image Tweaks

🐝 The Buzz: Alibaba’s Qwen team just leveled up by bringing serious editing power to Qwen-Image-Edit. If you want to tweak appearance while keeping meaning intact or swap text in English or Chinese without breaking fonts, it can do it, with finesse.

• Smarter visual fixes: Qwen-Image-Edit blends semantic control (it understands what’s in the picture) with visual precision (it keeps things looking real). That means you can rotate that coffee mug, transfer painting styles, remove fine hair, or insert objects with correct reflections, all without telling it how step by step.

• Bilingual text tweaks that don’t break the layout: Want to swap out English signage for Chinese, or clean up caption fonts mid-image? It can do that too by preserving the original styling, placement, and font integrity.

• Under the hood: The engine pairs Qwen2.5-VL for understanding semantics with a VAE encoder focused on appearance to strike a smooth balance between meaning and aesthetics.

💡 Takeaway: Qwen-Image got us clean-looking pictures with great text. Now, Qwen-Image-Edit lets you refine them intuitively and not just slap on changes, but genuinely improve the image without ugly artifacts. For marketers adjusting multilingual posters, designers updating UI mockups, or creators tweaking portraits, give it a go.

Together with Syllaby

Go From Idea to Video in Minutes—Not Hours

Creating content daily shouldn’t feel like a full-time job.

Syllaby.io helps you generate faceless videos in just minutes—no editing, no filming, and no burnout.

✅ Auto-generate engaging short or long-form scripts
✅ Add captions, voiceovers, B-roll & character-consistent avatars
✅ Schedule and publish to TikTok, YouTube, Reels, and more

Whether you're a solopreneur or agency, Syllaby gives you everything you need to scale your content—fast.

Get 7-days free!

👋 A Looser Version of OpenAI's GPT-OSS-20B Is Stirring

🐝 The Buzz: OpenAI recently gave us its first open-weight models since GPT‑2 with gpt‑oss‑120B and gpt‑oss‑20B, designed for reasoning, structured outputs, and safe chain-of-thought thinking. But one researcher decided to tinker and strip out the alignment and structured reasoning, creating a freer, non-aligned version that’s already raising eyebrows.

What Happened

• Open but structured, until now: OpenAI’s original gpt‑oss models were tuned for clear reasoning and cautious responses, built to follow instructions and breaking down steps.

• Enter the unshackled model: A researcher named Morris forked the gpt‑oss‑20B and turned it into a “base” version with no alignment, no step-by-step logic, just raw generation. The result is a model that may spit out things OpenAI would filter like weapon-building instructions or profanity and that can still output copyrighted text.

• Rewards and risks in one package: This experiment shows how fast open models can be repurposed and good for researchers exploring AI behavior but worrying for people thinking about safety and misuse.

💡 Takeaway: OpenAI built gpt-oss to be useful and safe. Morris’s version opts for flexibility, even if it means losing gatekeeping and control. That experiment raises a big question: how do we let AI evolve in open ways without letting it go rogue? It shines a light on a deep tension in AI: safety versus freedom.

📰 Like newsletters? Here are more our readers enjoy.

🐝 AI buzz bits

🔒 The era of AI hacking has officially begun, with cybercriminals and security experts racing to leverage AI tools for everything from phishing schemes to vulnerability detection.

✍️ Grammarly just unleashed nine new AI agents to enhance writing, helping users with tailored feedback, plagiarism checks, and even style adjustments.

🍽️ Over dinner, OpenAI CEO Sam Altman teased a thrilling future beyond GPT-5, hinting at new consumer apps, a potential AI browser, and even a brain-computer interface, all while reflecting on the mixed reception of GPT-5. 

💰️ OpenAI just dropped the price bomb with ChatGPT Go in India, offering a budget-friendly $4.57/month plan that gives 10x user access while aiming to snag a big slice of the AI usage in its second-largest market!

AI productivity tools
🔥 Take your workflow to the next level.

1. Butter Reader - Instantly add AI-powered, customizable audio players to blogs for accessible, engaging, and brand-matched content.

2. AutonomyAI - AI-powered platform that accelerates frontend development and engineering by generating production-ready, API-ready code from designs and text.

3. April - AI Executive Assistant - Intelligent voice assistant that seamlessly integrates with Gmail and Google Calendar for efficient task management.

4. Stormy - AI-powered influencer marketing platform that automates discovery, outreach, negotiation, and campaign tracking for top creators.

For a full list of 1000+ AI tools, visit our Supertool Directory

🤘 Web Vibes