Hi 👋 I’m Simon. I’m a chemist with a PhD. I spent years learning how science actually works. Now I spend far too much time on the internet watching people confidently explain how it doesn’t. So I write about chemicals, food, agriculture, medicine, AI and evidence. I make complicated science simple. I challenge claims that don’t survive the evidence. And I explain what the headlines leave out. Science is complicated enough. The internet doesn’t need to make it worse. Welcome.

The bottle neck in drug discovery is not ideas.

Ignorance isn't the biggest problem. Confidence isn't the biggest problem. It's ignorance combined with absolute confidence.

Imagine designing a machine that rewards people for being angry. Then giving everyone access to it. Then acting surprised when society becomes angry. Welcome to social media.

We've reached a strange point in history. People fear trace amounts of chemicals in food. But happily consume enormous amounts of misinformation every day. Only one of those comes without a safety assessment.

Nature invented some of the most toxic substances known to humanity. Humans invented vaccines. But apparently "natural" means safe and "synthetic" means dangerous.

Can you see this? It’s my turn to be absolutely throttled by the algorithm.

Here's my prediction: The cheaper AI makes ideas, the more valuable experiments become. The next gold rush won't just be for computing power. It'll be for evidence.

I’m calling it again. Big AI will start buying Big Pharma. Not for the drugs. For the experimental data. And the wet labs that generate it.

"Could" is carrying a lot of weight here.

Anthropic just announced that AI designed proteins targeting scorpion venom. The headlines sound incredible. 318,484 protein designs. 211 tested in the lab. 17 passed the binding tests. And how many have been shown to neutralise the venom? Zero. Now, don’t get me wrong. This is impressive science. AI can generate and screen molecular designs at a scale that would have been unimaginable a few years ago. But here’s what the hype keeps missing. Binding to a toxin isn’t the same as neutralising it. And neutralising a toxin isn’t the same as developing a safe, effective treatment. Someone still has to make the molecules. Run the experiments. Measure what actually happens. Demonstrate efficacy. Establish safety. AI is making it easier to generate hypotheses. But generating hypotheses was never the same as proving them. You can accelerate the thinking. You can’t skip the evidence.

Your body naturally produces acetone. Yes. The same chemical found in many nail polish removers. It’s a normal product of fat metabolism. A chemical’s name tells you surprisingly little about its risk.

Who would have thought....

The difference between a scientific breakthrough and a headline? About ten years of experiments nobody wants to read about.

It takes five seconds to invent a lie. Five minutes to share it. Five hours to explain why it's wrong. And five years to convince people it was a lie.

Silicon Valley has just discovered that intelligence is expensive. Wait until it discovers how expensive reality is.