Back in 2020 I finally set aside time to learn about containerisation. Docker had been on my radar for a while, but I kept putting it off, convinced it was strictly developer territory. Determined to challenge that belief, I spun up a modest Virtual Private Server (VPS) with Contabo and embarked on a learn‑by‑doing journey, self‑hosting a handful of services. The concepts felt tricky at first, but I persevered, and as I grew comfortable with the Docker CLI, the hype began to make sense.

Like many people, my daily routine involves a search engine. Whether it is a quick fact, the latest headlines, or a deep dive into a new topic, search engines have always been the gateway to discovery. For me, that changed in 2023, when I discovered Perplexity—an AI-powered research tool that’s quietly revolutionised how I find and use information. The Evolution of Everyday Search Traditional search engines have always been about links. Type a query, get a list of websites, and hope the answer is buried somewhere on page one. Perplexity takes a different approach: it delivers direct, well-sourced answers to your questions, right up front. No more hopping between tabs or skimming articles for a single fact. The experience is more like chatting with a knowledgeable assistant than using a search engine.

At the end of 2023 OpenAI introduced GPTs — custom versions of ChatGPT that anyone can create. These tailored assistants are designed to perform specific tasks, help with workflows, or simply provide a more personalised experience. In this post, I’ll explain what GPTs are, why they’re so useful, and how I created a custom GPT to help me learn Linux commands. What Are GPTs? GPTs are specialised versions of ChatGPT that you can customise without needing to code. You can:

I’ve been exploring the capabilities of the recently announced successor to GPT-4: GPT-4o (“omni”). At a glance, GPT-4o feels like a natural evolution of the GPT family. Not just bigger or faster, but more flexible in how it interacts with me across text, voice and vision. Where GPT-4 felt like a defining step into multimodality, GPT-4o feels more like that step applied across real workflows I care about. What’s Different: From GPT-4 to GPT-4o When comparing GPT-4 to GPT-4o, a few clear shifts stand out.

I recently discovered Ollama (ollama.com), a platform that allows you to run open-source Large Language Models (LLMs) directly on your local machine. As someone with an M1 Pro MacBook Pro, I was intrigued by the idea of leveraging my hardware to run advanced AI models privately. After exploring Ollama and experimenting with open-source models like Llama, Mistral, and Phi, I’ve been amazed at the speed and quality these models deliver—especially considering they’re running locally.