A Week of AI & Creativity Updates (and what to expect from 2025)
New Models, New Features, and The 12 Days of OpenAI
Over the last week, a lot has happened in the AI space. Not to be outdone by the 12 Days of OpenAI, Google revealed Gemini 2.0 with the headline “Our new AI model for the agentic era.” We will hear much more about the various flavors of AI Agents throughout 2025.
I’m super excited about the potential of AI agents, especially in the creative world. I can already think of dozens of areas where a capable agent would greatly help, research for one.
This week may be a slightly shorter newsletter than usual. Again, it’s been an extremely busy week. The day job devoured nearly forty hours just in the first three days. That, together with a sick wife, doing what I can for my cancer-stricken best friend, and trying to keep the household running, hasn’t given me much time to write. Such is life.
I should give a shout-out to my daughter, who is getting married on Saturday, December 14 (at the time of writing, 00:47 AM). As I’ve said before, life has a habit of getting in the way of creativity.
Gemini? Never Used It!
I’ve been so busy playing with Claude, ChatGPT, Perplexity, and, latterly, NotebookLM for the last few months until I realized I’d never even tried Gemini. That must change. I can’t afford to drop the ball, especially when the AI space is changing so rapidly.
The first hurdle I ran across within seconds was that the free tier only supports uploading images. A bit silly, in my opinion. There must be a way to do this. So I asked Gemini, of course. When working with GenAI, forget about Google search; just ask the tool itself:
Me: Good morning. Can you tell me how I can upload files other than images to Gemini?
Ah, it’s only available in the paid Advanced tier, even sillier. Never mind. I’ll start the trial for the Advanced version, and I can always cancel it. Once that was done, which was a bit of a faff, I’m in. Next up was to present my requirements (creative writing) and ask what it can do to assist:
Me: Can you give me some ideas on where to start and how Gemini can help me in the creative writing space?
The expected answer is full of self-promotional garbage. Still, I’ll persevere, quickly moving on to the meat of the session:
Me: Thanks. Now, I want to test your creativity. Can you rewrite the section "Gemini? Never Used It!"? The section must contain 250 words, and try to write in my voice. Consider the section in the context of the rest of the blog post.
Here is where it goes downhill. First, the answer was far too long and full of typical GenAI speak. I’ll only use an extract here, as it’s almost too painful to read:
Gemini: I’ve been so busy lately exploring the creative potential of different GenAI tools. I've spent countless hours experimenting with Claude, ChatGPT, Perplexity, and even delved into the fascinating world of NotebookLM.
Not to put too fine a point on it, this is plain awful. Particularly painful is the use of the word “delved.” Aargh…
Claude to the Rescue?
To be fair to Gemini, I’ll try the same test with Claude and compare the results. First up, the word count was a lot better, 238 words compared to Gemini’s 300-plus words. But that is a minor point in this context. Below, I’ve pasted the first two paragraphs only to give a flavor of the creative quality, and also, I hate padding out an article with AI-generated text:
With all the excitement around AI tools lately, I've realized something rather embarrassing - I've never actually tried Gemini. I've been so caught up experimenting with Claude, ChatGPT, Perplexity, and lately NotebookLM, that Google's offering completely slipped under my radar. How did that happen?
It's a bit ironic, really. Here I am, writing about AI developments, yet I've managed to completely miss one of the major players. I suppose it's a testament to just how crowded and fast-moving the AI space has become. You blink, and suddenly, there's a new model or feature to explore.
I don’t know what you think, but I think this is much better and a lot closer to my voice than Gemini managed. Let me know what you think. Leave a comment or message me. I particularly like the second paragraph, which sums up my thoughts very well. But it does make me wonder where to draw the line. I want to make progress with my book and not spend hours playing with GenAI. After all, that’s the entire point of the exercise, isn’t it?
To be Completely fair, ChatGPT 4o (with Canvas)
AS ChatGPT could become my new writing assistant, darling, although it’s still in its early days. I had to give it a go with the same uploaded file and the same prompt. Again, the word count was in line with the guidelines given, and the authenticity of my voice was better. Not great, but better.
To be fair, both Claude and ChatGPT had an unfair advantage (especially Claude) in that I uploaded some good examples of my previous posts. But this only explains the lack of my writer’s voice. It doesn’t, however, explain the terrible AI-ish prose that it used.
Back to ChartGPT. What I like about it is that I can highlight a paragraph or sentence that I’m not all that happy with, right-click, and in the mini prompt, ask it to extend, change the style or tone, and anything else we can think of. Claude is still far better out of the gate with its initial offering. I wonder when Anthropic will add the ‘Canvas’ feature to Claude; I don’t think it’s far away.
A Sora Point
It’s great that Sore has been bundled with my ChatGPT Plus subscription. But what is rather irritating is that it’s not yet available in the Netherlands, my adopted country.
So, let me get this straight. Even though I pay the same as the rest of the world for my Plus subscription, I still don’t get access to it. Something doesn’t seem right about this state of affairs. From the OpenAI page , I see Sora - Supported Countries; apparently, no European Union country is listed. What gives?
It seems that the incredibly strict EU regulations are the culprit. These regulations delayed the EU release of ChatGPT a few years ago. I guess I’ll have to be patient and keep my fingers crossed.
This was supposed to be the highlight (for me) of all of the expected releases from OpenAI. I was thinking about how even a very short AI-generated video would be incredibly useful for book trailers, advertising (book releases, of course,) and any other niche we can think of to publicize our latest and greatest masterpiece.
Final Thoughts
I play Wordle every day, as I’m sure many of you still do. Earlier this week, I asked ChatGPT 4o to have a go at solving that day’s game after I’d solved it myself. After a false start, where it simply searched the Internet for the answer, I asked it to work it out and show each step along with the logic.
It did okay, but it could never really grasp the rules. For example, it knew which letters were in the word but consistently forgot the correct position rule. Good, but far from perfect. This simple experiment illustrates nicely that we can’t trust GenAI tools completely, at least not yet. But those days are not far off.
The new ChatGPT 4o with canvas is very close to Claude 3 for creative tasks. I’ll use both over the coming weeks and compare the results. Of course, you will be the first to know when I find anything interesting and potentially impactful on your writing process. For example, I’m busy playing with Claude’s custom styles (writing styles or voice to you and me); I’ll keep you posted.
Until next week, have a great and creative week.