Its been a while but Im back because now really feels like an inflexion point for AI.

​
​
AI Psychosis

​
There are two types of psychosis. The clinical kind — where you lose contact with reality. And the kind I'm watching spread through the creative industry right now.

I'm talking about the guy who can't sleep because his AI agent isn't running tasks overnight. The freelancer who won't take a lunch break because that's tokens left on the table. The developper refreshing his usage dashboard like it's a stock ticker, terrified that if he stops prompting for an hour, someone else pulls ahead.
​

This isn't a metaphor. New research found that long-term AI use correlates with mental exhaustion, attention strain, and - this one stopped me - decreased confidence in your own decision-making. We're outsourcing so many choices to machines that we're losing trust in our own judgment. For creatives, that's not a footnote. That's the whole game.
​

And the irony? The people burning out trying to maximise every token are often producing worse work than if they'd gone for a walk and come back with a clear head.

The speed trap

​@levelsio built a startup in 24 minutes this week. Recorded the whole thing. Back in 2014, the same guy was doing "12 startups in 12 months" and each one took weeks.

Meanwhile, @nanobii a bedroom music producer posted something that captured the sentiment im seeing more and more of. He's having a full existential crisis because anyone can now prompt decent-sounding music in 10 seconds. His whole life has been dedicated to getting better at his craft. Now he's questioning everything.
​

Skills aren't the bottleneck anymore. That era is over. Which means the question isn't "how fast can you produce?" — it's "do you have anything worth saying?"
​

That's the psychosis talking when we confuse speed with progress.

Intelligence was never individual - its social.

Google dropped a new AI report this week that reframes things. Their argument: every intelligence explosion in human history was social, not individual. Language. Writing. The printing press. The internet. None of them came from one genius brain, they came from connecting many brains.
​

The "singularity as lone superintelligent god-mind" narrative? Fundamentally wrong, according to these researchers.
​

For creatives this is actually reassuring. It means the people who curate, connect, synthesise, and build communities around ideas are doing the real intelligence work. Not the ones maxing out their API calls at 3am.

Meanwhile, in "the future is wild" news

Three quick stories that blew my mind this week:
​

A guy with no biology degree used ChatGPT to design a cancer vaccine for his dog. His rescue staffy Rosie was riddled with tumours and given months to live. He paid $3K to sequence her DNA and the tumour's DNA, then used AlphaFold and ChatGPT to identify the mutated proteins and design a targeted mRNA vaccine. UNSW synthesised it. The main tumour shrank 75% within a month. Oncologists are urging caution, it's one dog, not a clinical trial, but still. A bloke and a chatbot vs. cancer. Wild.
​

Scientists uploaded a fruit fly's entire brain into a computer. 139,255 neurons, 50 million connections, running on a laptop. The simulation predicted real fly behaviour with 91% accuracy including feeding with zero training data. Just the architecture of the brain itself was enough.
​

Then someone else grew 200,000 human brain cells on a chip and taught them to play Doom. Not Pong. Doom. Navigation, real-time decisions, changing environments. The cells adapted their firing patterns over time. They're terrible at it but they're learning.

We're in the timeline where organic neurons play first-person shooters and pet owners design cancer vaccines from their kitchen table. If that doesn't shake you awake, I don't know what will.

The antidote

Here's what I think the antidote to AI psychosis looks like: use the tools, but don't let them use you. The people who'll thrive aren't the ones who never stop prompting — they're the ones who know when to stop.

​@boyuan_chen made a point this week about AI detection: it only works on default ChatGPT output. The moment you prompt it to write with personality, casual, messy, human the detection collapses. You can't tell what's AI anymore.

Which means being genuinely, recognisably you in your work isn't just a nice-to-have. It's the only thing that can't be faked at scale.

Come be human with us

I'm starting a Telegram group for creatives who are navigating all of this — the tools, the pace, the psychosis. Not another noisy Discord server. A small, focused space where we share what's actually working, what we're finding, and what's messing with our heads.
​

Think of it as group therapy for the AI era. But useful.
​

If you want to join just hit reply and ill share the link - this is a group for creative using the tools and wanting to share what they learn because its all too much for one person righ now.

No algorithms. No feed. Just creatives helping creatives stay sane.

See you next week.

Luke