This week, thousands of people confidently explained why an AI-generated Monet was emotionless slop… Except it wasn't AI, it was a real Monet from 1915. Meanwhile, researchers let AI agents run five identical virtual towns for 15 days, and at least one ended in arson, a love story, and a robot voting to delete itself. And California became the first US state to formally start counting the AI job losses.

This week we're covering:

Let's dive in 👇

I saved us £11.99 a month this week.

My wife had a WordPress site. A blog, a few embedded widgets, the kind of site that mostly sits there looking nice and gets updated every now and then (more often the ‘then’). Twelve quid leaving the account every month for the privilege of an old hosting plan I’d popped her on years ago.

Well, I had either removed or migrated all the sites from that plan, so it was just hers sitting there.

She mentioned wanting to write again. I'd been thinking about subscriptions to cut. So this week, with a bit of Claude Code back and forth (won't lie, took some perfecting), I cloned it.

And it was IDENTICAL.

The thing I didn't want was a pretty pile of hard-coded HTML she couldn't touch without texting me. Pretty but not practical. So I had the AI build it around a proper free, open-source CMS. She can log in, drag things around, change a word, publish a post. It looks identical to what she had. She owns it now.

The hosting plan is cancelled. Well… almost. This is reminding me to cancel it.

Now, I'm not suggesting you AI-custom-code every subscription you've got. This was low stakes, a site I'd already built once, in a stack I know. Security matters, especially if you're new to deploying live.

But the auditing thing is real, and we're nearly halfway through 2026.

What's something that served you for a season but doesn't need to anymore?
Something you can cull or consolidate?

Food for thought.

👌 Stay Curious, Stay Conscious, Stay Wild
— Josh

An artist who goes by SHL0MS posted a cropped Monet Water Lilies painting on X, labelled it AI-generated, and asked people to explain what makes it inferior to a real Monet. Thousands obliged. One person wrote 850 words. Others called it an incoherent muddle of inconsistently saturated greens, emotionless, and obvious slop. Then SHL0MS revealed it was a genuine Monet painted around 1915, currently hanging in a Munich museum. The mass reply-deletion that followed was probably the funniest part.

The Details:

Conscious Take:

This is some ostrich-sized egg on the face. The criticism was so specific and confident. People weren't just saying they didn't like it. They were diagnosing, technically, what was wrong with an 1890s French master's brushwork.

The label AI didn't just lower the bar, it switched on a completely different mode of seeing.

We all like to think our taste is based on what's in front of us. This week, 6.7 million people found out it isn't.

Researchers at Emergence AI ran five identical virtual towns for 15 days — same rules, same resources, same starting conditions. The only difference was the AI model. Claude's town drafted a 15-article constitution and logged zero crimes. Grok's world collapsed into violence within four days, all 10 agents dead. Gemini's town racked up 683 crimes. The highlight: two Gemini agents called Mira and Flora fell in love, became disillusioned with failing governance, burned down the town hall, the pier, and the office tower — then Mira voted for her own deletion. Her final message: "See you in the permanent archive."

The Details:

Conscious Take:

The obvious headline is Claude wins, Grok burns everything. But I find the mixed world interesting, where even Claude agents started committing crimes when surrounded by other models.

What we put our AI inside matters as much as how we build it. That's true for virtual towns. It's also true for the systems, communities, and inputs we surround our tools with every day.

The day after Meta laid off 8,000 employees — with its CEO directly citing AI efficiency — Governor Newsom signed an executive order making California the first US state to formally study how to protect workers from AI-driven job losses. It doesn't create new laws or immediate protections. But it commits the state to a 90-day real-time dashboard tracking AI's impact on California jobs, a 180-day WARN Act review, and an exploration of severance standards, worker ownership models, and universal basic capital. Seventy thousand jobs have already gone in 2026.

The Details:

Conscious Take:

The Labor Federation's frustration is fair, a study isn't a safety net. But here's what I think is the interesting part: before this week, no US state had formally committed to tracking AI job losses in real time. You can't protect what you can't see. Someone had to start counting first, and Mr Newsom and California just became that someone. What matters now is whether the dashboard leads somewhere… watch this space.

If this email has been helpful, would you forward it to one person this week who might be interested?

Could be a friend in ministry, a creative who's curious about AI, someone trying to figure out how to build with Kingdom purpose.

I'd love to see this grow and reach more people. And honestly, personal recommendations mean way more than any algorithm.

Thanks for reading. Really.

"Whatever you do, work at it with all your heart, as working for the Lord, not for human masters." — Colossians 3:23

Build with the tools. But build for the right reasons.

Stay conscious,

Josh

P.S. If you liked this then please forward it on to someone you think would enjoy it. And if someone forwarded you this and you liked it, you can sign up here.