Manus - a new LLM model has been impressing @Bruce Lloyd None of us had heard of it, but if you want to have a look and let us know your thoughts, that would be great. Deep Seek doesn't like talking about China, but there is a workaround - ask it the same question three times and it gives in! Southsea Bubble - ChatGPT was used to write a film scrip for this famous financial crash from the 1970s. Claude - there seems to me a mass migration to this platform - what are your thoughts? Join us next month when we get together to grab our coffee and talk about all things AI! Put it in your diary so that you don't forget. Here's the link

I’m an experienced AI developer who focuses on building practical, production-ready solutions, not just demos that look good on paper. I’ve worked on things like: -Custom AI agents & assistants -LLM integrations (OpenAI, Claude, open-source models) -Automation pipelines & AI-powered tools -Backend + API integrations to make AI actually usable in real products I care a lot about clear communication, fast iteration, and delivering something that genuinely helps your business — not overengineering or buzzwords. I’m happy to suggest better approaches if I see one, and I keep things transparent so there are no surprises. If you already have a clear plan, I can execute quickly. Happy to jump on a quick call or chat to see if we’re a good fit. Looking forward to working together!

Yesterday I mentioned in class that I’d posted my first reel on Instagram that includes AI video. I’ll pop the link below. It’s not perfect, but it was a good exercise for me as I’ve been intrigued how it could be used and if it would be effective. I used Runway for the first time, that although I liked and was easy to pick up straight away and use, I did get frustrated by their token system. Definitely a benefit to be more proficient now with prompting as I’ve been on the course - otherwise they’d be used up really quickly! Not sure how often I’ll do similar in the future as it was quite time consuming, but I do think it could be really handy for B-roll when I don’t have the right props at hand. I did have 3 AI video clips to start with, but changed to be a blend of just 2 with a real video clip of me (which I think is better as I’m quite fussy that I’m authentic for my brand). Let me know what you think - but be gentle 😉 https://www.instagram.com/reel/DTdV3jgCIhw/?igsh=dW1kdmExaTdhd2w0

The same prompt can produce different answers. If you are using AI for marketing copy, that may not matter. If AI is influencing strategic thinking, financial decisions or governance discussions, it absolutely does. Here is why. • AI does not read words the way we do. It processes tokens, small mathematical segments of text. • Even minor changes such as punctuation, spacing or capitalisation can alter token boundaries internally. • That shifts probability weighting. • Once the first few tokens differ, the rest of the response can drift. This is not a flaw, it's how probabilistic models work. The issue is that unmanaged variability inside a business decision environment is not harmless. It creates: • Inconsistent outputs • Hidden risk exposure• Overconfidence in machine-generated reasoning • Blurred accountability The real issue is not whether AI is being used, it is whether it is structured and bounded. Well-designed SME AI tools: • Fix the output structure • Standardise the input format • Reduce randomness for advisory use • Define clear boundaries on what AI can and cannot do • Keep human judgement firmly in control AI should sharpen thinking, it should not introduce invisible drift into commercial decisions. The risk is not artificial intelligence, it is artificial confidence. If AI is already inside your business, and it almost certainly is, the question becomes: • Do you know where it is operating, and have you defined the boundaries? That is a governance conversation worth having.

Interesting article about the impact of AI usage on the brains of writers: https://medium.com/techx-official/shocking-mits-verdict-on-ai-assisted-writing-04e2fbad8a17. Spoiler, it's not good if done the wrong way, it is good if done the right way.