I can’t believe it, but I actually managed to write a couple of AI songs on Sonar, submit them via SoundOn (TikTok’s service for publishing), and after a couple of back-and-forths they have been approved and will be launched at the end of next week.

Watch this space as I take on more AI adventures.

I get emails all the time from schools, businesses, and local sports clubs. I appreciate that many of these people are not professional communicators and are often just volunteers, but there is zero reason anyone should send an email to a community, group, or school without running it through ChatGPT for improvements first.

The level of communication—especially from our local sports club—is absolutely horrible. By the time you’ve read the email, you’re not really sure what’s going on or who it affects.

Running a message through ChatGPT for a quick sanity check and clarity fixes solves these problems. It only takes about 30 seconds.

What this fixes:

• Clear purpose
• Who is affected
• Simple next steps or call to action

I just got a new iPhone 17 Pro. I haven’t upgraded since the iPhone 12 mini, so it’s been about four or five years since I had a new phone.

The size is quite large.

What amazes me is that Apple designs such beautiful phones, gives them an incredibly well-crafted machined finish, and then you have to go buy a case. It’s astonishing in this day and age that we still have to accessorize things that were made to be so beautiful. As soon as you add a case, they’re far less beautiful and far more bulky.

The Irony of AI’s Evolution

I feel a core irony in the evolution of AI. Much of what we interact with today comes from large language models that learned from vast amounts of human-generated content, digitally available since time began.

Now AI is so capable that video imagery and the written word are used to produce content, and this trend will only accelerate. And is AI learning from what AI creates in turn?

Blog writers, influencers, and video makers are publicly exposed to the same LLMs, with their outputs crawled and learned from. Could this form a dangerous echo chamber where a thread of information, whether factual or misinformation, spirals into a continuing falsehood?

Key Considerations

• How much do LLMs learn from human-generated content versus their own outputs?
• What happens when AI-generated content feeds back into training data?
• Could an echo chamber amplify misinformation across generations of AI content?

For context, it’s worth watching how AI ecosystems shape what we see next, and how responsible design can help mitigate these risks. OpenAI and other leading labs are at the center of this conversation, with ongoing debates about data provenance and model governance.

Further reading: OpenAI and, for image generation tools, Midjourney.

I recently posted about Spotify playing me a band I thought was excellent, only to discover it was an AI band. This sparked my curiosity, so I signed up for Suno, an AI music maker online.

I had ChatGPT-5 create a song for me — or at least lyrics. I prompted it with what kind of song I wanted and the story behind it, and it wrote a pretty decent song. I then fed it into Suno to set a musical style, and what came out surprised me: a track I would have loved to record myself.

As a musician who has recorded in studios and played for most of my life, I don’t have a full band or the time to sit down on GarageBand and finish the track, so I wanted to push this experiment a little further to see how hard it is to get a music release onto Spotify.

• Step 1: Sign up for TikTok’s Sound On and submit two of the songs.
• Step 2: Wait a few days for approval.

Flash update: after 48 hours, the songs were not approved, and I received a text from Sound On this morning. I’ll keep you posted on what the issue is, what I need to fix, and whether I can get it approved.

On one hand, I’m torn — AI managed to write a song I would have recorded, albeit with vocals that might be better than I could ever achieve locally. On the other hand, it’s fascinating to explore this space and consider where the craft of music is headed. Watch this space.

I spent quite a while playing with Arcads.ai yesterday after they announced a new release saying that the ability to have influences holding your product was in play. I think the problem with our label is that it has some fine detailing on it, and that gets lost in AI.

So unfortunately the two tests I ran still came out looking completely dysfunctional, and while it might be 80% there, the final 10% is really where the proof in the pudding is.

I’ll continue to experiment with AI tools like Arcads, as they have an API which makes it easier to produce things in an automated way at scale, either using something that is vibe coded or something using Pabbly Connect.