AI Assistance: Who's Really in Control?

I've been using Artificial Intelligence (AI) to help with various tasks for a while now, programming included. It's become a genuinely useful tool that I reach for regularly. But there's an important distinction to make about who's doing what.

Building This Site

This entire site was built using Claude Sonnet 4.5. But let's be clear - I was the one calling the shots. I specified React, Tailwind, and Markdown from the start. I knew exactly what I wanted: a system to transform Markdown into HTML that could be statically served using Vercel.

The AI didn't dream this up. I did. It just helped me implement it faster and more efficiently than I could have done on my own.

My Writing Process

My approach is straightforward. I draft articles in rough list format, capturing the key points I want to make. Then I ask the AI to use my persona to transform them into clean, readable text. Most artwork on this site is either tidied up or generated with Google Nano Banana.

It's a practical workflow that plays to each side's strengths - my experience and judgment, the AI's ability to structure and polish.

The Reality Check

Here's the thing: AI is a tool, not the master. It has real limitations and needs careful guidance from a human who knows what they're doing. Without proper direction, it'll happily produce confident-sounding nonsense.

My Take on Large Language Models

Personally, I think Large Language Models (LLMs) are a dead end on the journey to Artificial General Intelligence (AGI). LLMs will continue to be costly yet valuable tools, but they're quite limited. Too many AI personalities are overselling the capabilities to further their own agendas.

The AI bubble is real. I reckon we'll soon see more people realise they've been sold a dream that won't be delivered with LLMs alone. The hype has outpaced the reality by a considerable margin.

Moving Forward

Don't get me wrong - LLMs are brilliant tools. I use them daily and find them genuinely helpful. But for now, AI should be assisting humans, not taking over.

The key is understanding what AI can and can't do, keeping a critical eye on its output, and staying firmly in control of the decisions that matter.

— David