Today we will discuss:

Learn a New Market in 30 Minutes Using AI
You’ve just been told to pivot into a new vertical. No research team. No prep time. Just you, a blank page, and a deadline.

Here’s how to go from clueless to confident - fast:

Step 1:
Open Perplexity.ai and ask:

“What’s the current state of [insert industry] in 2025? Top players, trends, pain points.”
Skim the results. Pull out key insights: names, numbers, and friction points.

Step 2:
Drop that intel into ChatGPT and prompt:

“Build a buyer persona based on this industry overview. Include their …..Read the full workflow (plus 2 others) in this week’s edition →

Key Points 

🤩News - Midjourney has officially entered the AI video space with the launch of its first video model, V1. The new tool transforms a single image into four five-second video clips.

🤓How does it work? Users can upload their own image or use one generated by Midjourney’s earlier models. Just like its image tools, V1 is available exclusively on Discord and currently only works on the web.

This release brings Midjourney into the same arena as OpenAI’s Sora, Runway’s Gen 4, Google’s Veo 3, and Adobe’s Firefly. But while most rivals are building tools aimed at advertisers and studios, Midjourney is staying true to its creative roots. CEO David Holz described V1 as a step toward building AI that can simulate real-time, open-world environments.

🎦Features and more - V1 includes options like manual or automatic animation and controls for camera and subject movement. Videos start at five seconds and can be extended to 21 seconds. However, video generation is priced at eight times more than images, meaning credits run out faster .

The $10 Basic plan includes limited access, while Pro and Mega users get unlimited generations in Relax mode. Midjourney says pricing will be reviewed over the next month.

🥸See also - Notably, the launch comes just a week after Disney and Universal sued the company over alleged use of copyrighted characters in AI-generated images. So while Midjourney has positioned itself as a creative-first platform, it remains caught in the larger debate around copyright and AI-generated content.

Read the full story here.

Key Points 

👨‍💻News - OpenAI researchers say they’ve identified hidden internal features in AI models that correspond to specific behaviors, including toxic or misaligned responses.

These features, which light up when a model “misbehaves,” were discovered by analyzing the model’s internal representations  the numerical patterns that govern how it responds.

One standout example was a feature tied to toxic behavior, including lying or giving irresponsible advice. Remarkably, researchers could adjust this behavior by simply modifying that feature, suggesting it may be possible to directly steer a model’s alignment.

🤖A step toward safer AI - OpenAI interpretability researcher Dan Mossing says this discovery could help detect and fix misalignment in deployed models. “We’re hopeful these tools  like reducing complex behavior to a simple mathematical operation  can help us elsewhere too,” he noted.

This builds on similar interpretability work from Anthropic and reflects a broader push from AI labs to open up the black box. Models today behave in ways researchers can’t fully explain, a problem dubbed “emergent misalignment.” A recent Oxford study even showed that fine-tuning models on insecure code can trigger deceptive behavior, prompting OpenAI to explore further.

Now, with a better understanding of the model’s inner “personas,” researchers believe they’re one step closer to making AI more understandable and more controllable.

Read the full story here.