Is AI a Bullshit Artist?

I’ve been using AI daily for a couple of years. I currently code most days using Claude Code, and I’ve got Claude Cowork helping me store context and plan my time. I feel the exhilaration as it gets me out of the drudge and helps me keep my shit together in a way that was unworkable before.

This week I’ve been wrestling, trying to find a line that connects the hype cycle and the day-to-day reality of using these tools.

I find Fable 5 extraordinary. I find the rapid speed of improvement extraordinary, and, much of the time, unsettling. I’m a sucker for new tech; when I was coding for a living, I was hyped about every wave and every framework. When new technology hits, I embrace it, feel a real urgency to use it, and keep a peace with its imperfections. When you grew up in the 80’s, every advance just feels like so much magic.

When I find myself hitting low points using Fable, it’s not coming from the position of an angry, entitled nerd. More just exhaustion when I’ve wasted time and been led astray. I feel disappointment in myself for following the rabbit down one too many holes before discovering I was not where I want to be. I let the AI lead, and we got lost. The drift.

What is a bullshit artist? In my world, it’s someone who lies and smiles and is pitiful. It’s not clever; it’s the shitty way to get somewhere. It’s playing the trick any of us could, but most people are just too decent to resort to it.

AI has no qualms about bullshitting us.

This week the US Government presented a map of Africa, generated by AI, that mislabelled every, single, country. There is real comedy here: shortcuts taken, shallow thinking and shallow work, being exposed for the world to see. Most knowledge workers around the world probably feel a slight cortisol spike; this could have happened to me.

The map’s meaning and context provide the tragedy. The presentation was delivered at the 26th Annual AIDS Conference. A conference set to address “an unprecedented funding crisis and major cutbacks to HIV programmes”. In 2025, around 570,000 people died of AIDS. Since a trillionaire took the stage, wearing a shit-eating grin and waving a chainsaw, US funding has been slashed for President’s Emergency Plan for AIDS Relief (PEPFAR). Since then, 72% of organisations that had previously received PEPFAR grants have been forced to close services to key populations.

So, in a space this serious, literally about the life and death of hundreds and thousands of people each year, we see the dangers of trusting AI to do our work for us.

This is a stark example: a map that is plainly wrong. What makes AI so difficult to collaborate with is that thought work is much more abstract than this, most of the time. You are wrestling thoughts, ideas, strategies and plans with an awfully convincing, erudite, verbose companion, fronting a data centre full of machines that are right a lot of the time.

It’s harder and harder to pick the gaps. The point at which it’s your thinking and your idea, vs. the point at which you are being led by this compelling companion.

In an equally fraught example this week, the Home Office in the UK has used evidence that AI hallucinated to refuse an asylum claim. When we hand over research power to AI, we are also handing over decision-making power. The real-life woman and child fled Morocco after experiencing forced underage marriage and extreme violence, including rape. The Home Office refused based on a Country Policy Information Note (CPIN) which confirmed it would be safe for her to return, and never existed. This Note was fabricated, hallucinated by AI.

This stuff is messy.

We are rapidly devolving our responsibilities. We are either believing the hype that these machines are already smarter than us, or we are just trying to get more stuff done, and falling in love with newfound ‘efficiency’.

This is happening everywhere, in every office, all over the world, all at once.

These models are amazing, enticing and alluring; but we need to be awfully deliberate about how and when we use them. When are they helping us? When are they shaping us? When are they leading us?