I want to show you two prompts that I found super useful with OpenAI's O3 model.

The first one is a brainstorming prompt that surfaces unexpected ideas:

You have consumed more information than anyone in the history of the world and you've demonstrated an extraordinary ability to make connections among them. What are the most important non-consensus or even not yet hypothesized things?

If you run this prompt, you'll likely want to add a specific topic underneath as an extra building block. For example, "generative AI for product marketing". The results can be surprisingly insightful, like "Most teams still treat Genai as a content engine. The real unlock is feedback, letting models both write and instantly metacritique variations."

Another powerful prompt comes from Amin, a trend analyst on X. He uses O3 to spot trends before they hit:

Analyze Reddit, Twitter, and Product Hunt to suggest a YouTube long-form or Instagram Reel idea covering an emerging trend that could solve a pain point for marketers.

This prompt has O3 scour various platforms, find new search terms, and then synthesize ideas for content that could address underlying emotions and pain points. For example, it might suggest "Your AI content is killing your brand voice" as a potential video topic.

These two prompts showcase the incredible ideation and trend-spotting capabilities of OpenAI's O3 model. I highly recommend trying them out for your own brainstorming and research needs.

Moving forward, I can use my voice to dictate anything in near-perfect quality as AI delivers it versus the built-in tool. Here's how to set it up:

1. Get the shortcut from the 11 Labs team that enables this functionality.
2. In the shortcut editor, add your 11 Labs API key and your OpenAI API key.
3. Assign the shortcut to your iPhone's action button or double-tap gesture.

Now, when you activate the shortcut, it will use the powerful speech recognition models from 11 Labs and OpenAI to transcribe your speech with superior accuracy compared to the default iPhone dictation feature.

To demonstrate, I'll try out the new shortcut and compare it to the built-in dictation:

Shortcut transcription: "Okay, there's some permissions I needed to give it. Hopefully this is just the first time while using it. Here I am testing my brand new shortcut. The quality of this should be way higher in theory. If it works, I'll use it everywhere."

Built-in dictation: "Okay here I am testing my brand new shortcut The quality of this should be much higher in theory If it works I'll use it everywhere."

As you can see, the shortcut provides better punctuation and sentence separation, delivering a more accurate transcription. This AI-powered upgrade to iPhone dictation is a game-changer for anyone who relies on voice input in their daily workflow.

The latest release of Midjourney, version 7, has brought some notable updates to the platform. The most significant change is the introduction of a new web-based user interface that provides users with more control and editing options.

The new interface integrates the previous paint tools, but now also includes layers. This allows users to bring in various images, combine them in a way they prefer, and continue working on top of the composition. This added flexibility can be particularly useful for users with a specific creative vision, as it gives them more control over the final output compared to the traditional Midjourney workflow.

While Photoshop has had layer-based editing and generative AI features for over a year, the new Midjourney interface can be a great addition for users who prefer a more streamlined and accessible approach to image editing. The web-based platform is now available to all Midjourney subscribers, not just the yearly ones, making it accessible to a wider user base.

Overall, the Midjourney 7 release represents a step forward in the platform's capabilities, providing users with a more robust and customizable image editing experience within the Midjourney ecosystem.

GenSpark, a US-based company, has released a new feature that focuses on creating slides, or PowerPoint presentations. While it doesn't work as well as the widely-regarded Gamma app, this new feature takes a more agent-based, chatbot-style approach.

I tested the feature by providing some bullet points from the application's notes. After answering a few follow-up questions, the tool started generating slides. After about 6 minutes, it produced 4 different slides.

Although this is not a traditional PowerPoint presentation, the output is more akin to an interactive infographic or landing page. It includes elements like audio players, though they didn't work in my test.

The tool seems to be better suited for enhancing presentations rather than replacing skilled human editors. Users can customize each element, as the output is HTML code with some JavaScript.

While not the most practical solution, this new feature from GenSpark represents a novel, agent-based approach to slide creation that is worth noting as part of the evolving landscape of AI-powered productivity tools.

Descript, a suite of tools for people who don't want to use complex editing software, has introduced a new feature that takes video editing to an agented level. This new release removes the AI features and instead lets an agent control the underlying tools and prompts to handle the editing process.

The key aspects of this new agented video editing assistant are:

• It is designed for content that is educational or podcast-like, where the editing needs are relatively simple.
• Users can simply talk to the agent and give instructions, such as "Can you edit this down?", and the agent will use the various tools and prompts to carry out the editing tasks.
• This approach aims to make the video editing process more conversational, similar to working with a freelancer, rather than directly manipulating a timeline or using complex editing software.
• While it is still a long way from replacing skilled human editors, this is seen as a significant step towards more agented video editing workflows.

Descript is currently accepting applications to test this new agented video editing feature, providing an early glimpse into the future of simplified video editing powered by AI-driven agents.

The key takeaways from this week's AI news are:

1. OpenAI's Image Generation API: The availability of OpenAI's image generation capabilities through an API opens up new possibilities for integrating AI-powered image creation into various applications. The ability to generate high-quality images programmatically is a significant development.

2. Powerful Prompts for GPT-3 and Whisper: The presenter showcased two effective prompts for GPT-3, one for brainstorming unexpected ideas and another for identifying emerging trends. Additionally, a tutorial was provided for using Whisper, OpenAI's speech recognition model, as a superior alternative to the built-in dictation on iPhones.

3. Midjourney 7 Update: The latest update to Midjourney introduces a new web interface with more editing capabilities, including the ability to combine multiple images and work with layers, providing users with greater control over the image generation process.

4. GenSpark's Slide Creation Tool: While not as polished as the industry-leading Gamma, GenSpark's new slide creation feature demonstrates an interesting agentic approach to generating presentation content, blending text-based prompts with AI-generated visuals.

5. Dcript's Agentic Video Editing: The presenter highlighted Dcript's upcoming agentic video editing feature, which aims to simplify the editing process by allowing users to interact with an AI agent that handles the underlying tasks, rather than directly manipulating a timeline.

Overall, this week's AI news showcases the continued advancements and diversification of generative AI tools, with a focus on expanding the capabilities of existing models and exploring new agentic workflows for tasks like image creation, slide generation, and video editing.