I’ve not been myself recently, actually I felt quite depressed. This is because I’ve been working through family issues which have led me to feel a lack of motivation and energy even for the things I enjoy. This means that I’ve taken time off building side-projects like: Job search AI: CV / Cover-letter / company research generation + clean-tech job site scraping / aggregation London Property Pulse: property analytics site Take a moment, it’s about staying in the game and not burning-out. ...

I’m learning a lot MCP servers tool calling, is that just another API? agentic AI and the next trends but maybe it’s time to take a step back AI and society making big waves seen as the solver of all things ‘work will never be the same again’ industrial revolutions 1st Mechanical Production (Steam-based machines), 2nd Mass Production (Electrical Energy-based mass production), 3rd The Digital Age (Computer and internet-based knowledge), 4th AI? IT’s not the first time that we’ve had technology fundamentally change how we operate in society and work dotcom bubble crypto bubble AI bubble? They each left beneficial things in their wake Current wave: quite a lot of job loss but without the hard working conditions - knowledge work is being transformed What are we not seeing? Changing the world It might all be a small group of people trying to push a narrative and make some spectacular coin, true it is useful - to a degree, but it might not be all it’s cracked up to be. A sentiment I’ve been resonating more with is that it’s good but only just so - the bigger issue is actually knowledge atrophy and the loss of the craft / imagination of people. if an image can be generated, why draw? If text can be written, why write? if code can be generated, why program? This is the problem. One argument is that things should be done for their own sake, but actually relinquishing full creative output to the Other will lead to a steady decline in faculties we once cherished. My biggest issue is that chatbots confidently give you crap and hallucinate, chasing wild geese. You might as well go for a walk and have a think. Sometimes you need a sparring partner, but more often than not, it’s just a question of time and a bit more thought. Chatbots are predictive machines; I’d argue the true genius is in human creation. There is more coming (rethinking of the transformer being the foundation for the next generation of AI, ‘World models’, etc.), but I think the current generation of models is plateauing. LLM creativity is bounded by small context and they are unable to solve moderately complex problems in a specific domain and do tasks that humans find easy still. They are marketed as a solution to drive automation and I agree to a point. However, they are not an endgame solution. Find creativity in the hard work, the struggle, the time it takes to make something great. Greatness and expertise cannot be generated, but it can be assisted. That’s where the balance lies: have tasks simplified and pushed along by the chatbots but the real work must come from you. What does changing the world mean? Another app? A faster phone / laptop? I’m not sure.. Philosophy and expression Thinking of some of the greatest minds in philosophy (List of your philosopher(s) of choice) - they helped us make leaps in how we think of ourselves and the choices we make. The major shifts came from deep thought. It took me some time I’m learning about ML and really enjoying it, thinking of some of the deeper issues made me think about what really matters to make big change happen - changes of behaviour in society Today computed thought is cheap (meaning of less value than human thought), and attention the currency - Tools should remain just that. Tools. What is the purpose of these tools? What is their value? Seems a shame to lose one of the essences of humanity: self-expression. If the chatbots homogenise our output, does culture get flattened? What is the role of art if it’s generated? The deeper problem is maybe the loss of free thinking? It’s not about faster horses (à la Ford), but where are we going with this? It seems that we’re living in an age of AI where if the words are said enough times, like an incantation, the miraculous future will come to be Seems like fanaticism I’m of the view the Kasparov expressed, it should be people + machines, not just machines. Not the last Transformer I listened to a podcast about the current architectures for transformer-based models, there is a possibility to move forward in that direction, but to get to the next generation of models a new architecture is needed. This means more investment, unfortunately, the industry has invested in the transformer architecture, and it will take an undeniably better solution to move to a new style of architecture. I believe that we are still in the first wave of the LLM-based AI change but that this is far from the end of the line. It is changing work, no doubt, but not the complete human replacement that some people are talking about. Spotify: “I invented the Transformer, now I’m replacing it” – ML Street Talk There have been and will be more waves of AI, as others came before, this is the wave of the chatbot, very public and accessible but not the endgame YouTube: Keynote: Machines, Learning, and Machine Learning - Dylan Beattie - NDC Porto 2025 Like with many technologies, society needs to live with it a bit longer with it to determine its true value. There will be many more changes before we get to true AGI as some people call it We just need to work alongside generative AI and integrate it into products Watching the 1980s film ‘War Games’ reminded me of just how far we’ve come, but working and living with LLMs isn’t the magic bullet that VCs and AI companies are touting them to be.

tl;dr AI for everyone: pretty good as a foundation. Deep learning for coders: Wow! In 2025, with so many AI products coming out, it seems like a good idea to properly learn what it’s all about and how I can build my own models and integrate them into a project. In my initial foray into learning ML, I found that the AI for Everyone course by Andrew NG was a good intro to the topic and ideas at a very high level. It’s a good basis for understanding the lay of the land and the different areas to dive into. ...

Life is hard and full of challenges so when I want an uncomplicated belly laugh, I just open up the Gulf of Mexico programming language docs. Here are some gems in the language: Add pizzazz by ending each statement with an exclamation mark or three Are keywords a problem? Delete them Unsure? Use the boolean maybe Text looking boring? Use rich-text (italics / bold / underline) to differentiate variables Want to add 3 + four? No problem Who needs loops? It’s a hilariously cursed language that should never be used in production, but I love that it exists! ...

For this to make sense, I’m assuming that you have accounts with both Cloudflare as your DNS provider and SendGrid as your email provider. If not, hopefully some of this translates or is food for thought. This was honestly a pain in the arse, mostly because of the feedback loops for seeing if DNS has worked is measured in days and there are no good error messages other than ’not working’ in SendGrid. This leaves you guessing and trawling the internet for things to try to get it working. ...

Yesterday London Property Pulse went live. Launching is a euphoric and satisfying moment where you feel ‘Hooray, I’m on the Internet 🎉’! Here’s to launching early 🍻 The site is definitely far from ready, with bare-bone BE and FE but this way I can see how my changes affect the site’s performance and build in public. It also helps my motivation: now it’s real and in the world 🌍. Learnings Getting familiar with Linux I’m trying something new by hosting on a virtual private server (VPS) - aka Linux box in the cloud - and had a few things to learn along the way but got there in about a day. While I’m familiar with unix environments, working without the bells and whistles of my comfy and configured environment was a bit of a shock but relatively easy to work. ...

Reading a couple of books on launching products made me think about boldness. Learnt or innate, sometimes the difference between starting, permission, adventure, change and possibility is boldness. Some wonder, think, turn over, want to have others make decisions for them or are paralysed by in action to make a decision. I believe the main factor of change is boldness. It leads you to launch yourself in a direction that is only partly known, where the path is lit by a torch beam. ...

Seems like on the socials people go over the top with ’the BEST way to be productive is…". In large groups, more complex systems arise depending on a project’s needs. However, working alone frees me up to get things very lean. The reality is that 6 things are needed to ship a project: Todo app (Todoist): Kanban board with project tasks + notes Calendar (Apple Calendar) Notes app (Obsidian + Excalidraw): Thoughts + resources + diagrams + documentation Email client (Apple Mail / Gmail web) Code editor (VS Code) Terminal (Ghostty) with SeaShells theme Browser (Edge …on Mac!) Tools come and go, I prefer to use a system like Getting Things Done (GTD) to stay organised and Shape-up as a development framework. ...

Challenge: for a week, don’t use a debugger and only use the tools you can use to debug in production. What did you notice? What workflow changes did you make to get to the same level of productivity as before? Why abandon the debugger? The debugger has it’s place, but it can be a crutch to crafting well-thought-out software. As useful a tool as the debugger is, not being able to use a debugger reveals a lot about coding practices and processes. ...

When developing there are words that are ambiguous when describing a problem. One of my pet peeves is when I hear someone say these cursed words: ‘it was working’ The problem is that it’s fuzzy and does not help identify causes. Logically it’s simplistic - yes, in the last PR is was probably working and now with the new PR it’s broken. It should be expected that over time enough code changes will cause breakages. ...