o3 = mind blown 🤯

You might have missed a few AI updates but I've got you...

A lot can happen in two weeks in the world of AI, so while the weeks have been short, there’s a lot of new stuff I’ve been testing in preparation for the Leads to Listings Events starting Thursday in Brisbane.

Here’s a whistlestop tour of what’s happening and what I’ve been testing. Buckle up, it’s going to be a wild ride…

1. OpenAI Launches o3 and o4-mini Models

Over the Easter period, Open AI released two new AI reasoning models that represent a significant advancement in AI capabilities:

  • o3: A new agentic reasoning model that's designed to "pause and think" before answering questions. It excels at multi-step tasks and can use tools like web search, code interpreter, and memory in a loop. Early testers report it outperforms competing models like Claude 3.7 Sonnet and Gemini 2.5 Pro in speed and problem-solving capabilities. I would agree with that based on the example I posted in the community this morning here. This model is super smart and some of the other prompts I’ve been playing with will be the subject of the Premium Subscriber Monthly Deep Dive on Friday morning - don’t forget to RSVP!

  • o4-mini: on the other hand, is a smaller version of OpenAI's next-generation o4 model, primarily focused on improving coding capabilities.

The o3 model, in particular, is described as "deep research-lite,ā€ allowing it to perform complex research tasks in minutes rather than the 10-20 minutes that deep research typically requires. Highly recommend you all research yourselves with the prompt:

tell me everything you know about [your name] from [your company]

Mine looks like this:

And the result is here. I think you’ll agree it’s almost CIA level comprehensive!

Meanwhile, I found o4-mini’s ability to fix some snarky CSS on eliteagent.com to be pretty phenomenal (read: fast, smart, accurate), only to have o3 roast me about the fact that I should be getting someone else to do those types of jobs. I don’t know… if it gives you the ability to do this stuff and then tells you not to…

2. Anthropic Launches Google Workspace Integration

Anthropic has introduced two major updates to Claude:

  • Google Workspace Integration: Claude can now connect directly to Gmail, Google Calendar, and Google Docs, turning it into a workplace assistant that can organise emails, meeting notes, and action items.

  • Research Feature: A new capability that performs multi-step searches with source citations, aiming to solve complex queries in under five minutes. This puts Claude in direct competition with Microsoft Copilot and other AI research tools, such as Grok, o3, and Perplexity.

I haven’t tested Claude’s new research feature thoroughly as it’s still not available in Australia. I did connect up the transfixed with what o3 can do right now that Claude is going to have to step up.

I did however connect up my Gmail and Google Drive to test the Workspace integration and that has proven pretty handy when it comes to finding ā€œneedle in the haystackā€ kind of emails, eg ā€œCan you find my license number for Screen Studio Proā€, but give it a basic question like ā€œWhat emails do I need to respond to?ā€ and it gave me a whole bunch of newsletters.

Side note: Open AI has released a Google Drive integration, but what I’d really like is for it to link into my Notion Databases. Please and Thankyou.

3. There are Rumours about Open AI developing a Social Network

Multiple sources have indicated OpenAI is working on its own social media platform with an X (the artist formerly known as Twitter) like feed.

If it happens, it could be a pretty smart move. It would give OpenAI a pipeline of user-generated data, which is basically gold dust when you’re training AI models. Right now, companies like Meta and X have a big advantage because they’ve already have this on their platforms.

This isn’t totally out of the blue, either. Sam Altman had already hinted at building some kind of social app after Meta started making noise about launching a standalone AI app.

So... watch this space. Of course, I have no opinions.

4. Google Releases Veo 2 for Gemini Advanced

Google has launched Veo 2, its latest text-to-video AI model, which is now available to Gemini Advanced subscribers. The model can generate 8-second, 720p cinematic-quality video clips from text prompts, with improved realism in motion and physics. This release also includes easy sharing options to platforms like TikTok and YouTube.

I gave generating some ā€œmarket updateā€ B-Roll a try, and I think it looks a lot like Broadbeach waters, which is a significant improvement over some of the past generations with Sora. But Sora is has gotten better than it was too - once someone (ie Google) raises the bar, everyone has to raise their bars too…

Please note that both videos are displayed here as GIFS and, as a result, appear at a lower resolution than they actually are. AI still has not solved the ability to embed a video in an email… (which would be nice, but we might be waiting a while…)

Cinematic drone footage of an elegant, modern luxury property located alongside a serene canal in Broadbeach Waters, Gold Coast. Capture aerial views showcasing the property's architecture, expansive outdoor spaces, sparkling infinity pool, lush landscaping, and the calm waterways. Bright, sunny afternoon lighting, smooth drone movements, and vivid, natural colors.

Sora - all looked great until I realised the pool weirdly has windows in it šŸ¤”

Google’s Veo 2 - love the palm trees blowing in the breeze… 🌓

5. Microsoft Launches Major Copilot Update

Microsoft has launched its "wave two" Copilot updates, including an agent store for research and analysis. Features include a ā€˜skill discovery’ agent that helps humans find others with specific expertise, brand-compliant image generation, Notebooks combining various work formats into one searchable space, and natural language search across Microsoft apps and third-party tools. You’re on your own in testing all this Microsoft stuff - I’m a Mac.

There’s more, but this newsletter is already a long one, so I’ll have to tell you more about Open AI's new shopping tool tomorrow!

Happy Hunting šŸš€

Sam

Co-Founder of Elite Agent

Have a question about this post? Join the Community

Need an intro to AI for Real Estate? Check out our Free Live Events in Sydney, Brisbane and Melbourne next month here.

PS: If you find this newsletter valuable, please share it with a colleague. Every share makes me do a little dance like Bill Gates in the 90s.

Reply

or to participate.