Hey {{first_name|Conscious Church Fam}}
Big week, and one of the more fun ones we've had in a while. ChatGPT's creature obsession has finally been explained, an AI just changed the math on one of the deadliest cancers, and Gemini is about to take over your car. Let's get into it.
In today's recap:
🧌 How ChatGPT got obsessed with goblins — and why OpenAI had to ban raccoons too
🔬 Mayo Clinic's AI can now spot pancreatic cancer up to 3 years before doctors
🚗 Gemini is replacing Google Assistant in millions of cars
Let's dive in 👇
✍️ Josh's Musings
I was listening to Greg Isenberg this week, he had Airtable CEO Howie Liu on talking about their new thing HyperAgent.
Quick note — they’re giving the first 25 people who sign up through a link $1,000 in credits. Full transparency: you get $1,000, I get $100. I tried it, you can run Claude Opus 4.7 inside it, so it’s actually decent to test with. Link here
Anyway, I asked it something simple:
“How would this help a small–mid UK church?”
No other context.
The answer was basically:
“The bottleneck isn’t ideas. It’s admin.”
Which feels… accurate.
It pointed to things like:
weekly bulletins (same format every week)
event campaigns (Christmas, Easter, Alpha etc.)
reports + charity/admin paperwork
visitor follow-up so people don’t get missed
Nothing groundbreaking. Just the stuff that always gets delayed.
The bit I hadn’t really thought about: pastoral memory
Things like:
“You said you’d call David 10 days ago”
“It’s Sarah’s mum’s anniversary next week:”
Not replacing anything. Just holding the list most people carry in their head.
It also flagged its own boundaries:
don’t use it to write sermons
be careful with sensitive data
don’t replace actual relationships
If I was using it, I’d probably just start with one thing, bulletin, inbox, or rota, and see if it actually saves time.
That’s it really.
And for personal projects/business, you might want to play with setting up unique agents to fit areas of your business (social media manager, chief of staff, researcher etc).
Food for thought.
Have a great Bank Holiday Monday (for those in the UK).
🙌 Stay Curious, Stay Conscious, Stay Wild
— Josh
Image: OpenAI | The Conscious Church
If you noticed ChatGPT peppering recent responses with goblins, gremlins, and assorted fantasy creatures... you weren't imagining it. OpenAI just published a fascinating deep-dive tracing the whole thing back to a single reward signal in ChatGPT's "Nerdy" personality preset, and the way it spread is genuinely wild.
The Details:
After ChatGPT-5.1's November launch, "goblin" mentions in user conversations jumped 175%, with "gremlin" up 52% and other fantasy creatures spiking similarly
The culprit was the Nerdy personality preset — it was driving two-thirds of all goblin mentions from just 2.5% of total user traffic
Fine-tuning loops then recycled those creature-heavy outputs back into ChatGPT's default mode, spreading the quirk to users who never touched the Nerdy setting
OpenAI retired the Nerdy preset in March and shipped GPT-5.5 with a prompt specifically banning goblins, gremlins, ogres, trolls, raccoons, and pigeons (yes, pigeons)
Conscious Take:
This story is delightful because it's a perfect window into how weird, and fragile, these models can be under the hood.
One reward signal in a niche personality mode created a creature obsession that spread globally across millions of chats.
It's funny, but it's a useful reminder: what we experience as AI "vibes" are often the product of very specific, very traceable decisions.
Image: Mayo Clinic | The Conscious Church
This one hit different. Mayo Clinic just published new results from an AI called REDMOD that can detect pancreatic cancer on standard CT scans — up to three years before a doctor would typically catch it. Pancreatic cancer is one of the most lethal cancers we know, precisely because it almost always goes undetected until it's too late.
The Details:
REDMOD reviewed nearly 2,000 routine CT scans that specialists had originally read as completely normal — before those patients were later diagnosed with pancreatic cancer
It caught 73% of cases early, and at the two-year mark before diagnosis, spotted roughly 3x as many early cancers as experienced radiologists
The AI reads "hundreds of quantitative imaging features" — texture and structure patterns that are simply invisible to the human eye
Pancreatic cancer has a 5-year survival rate below 15%, making early detection life-or-death — and REDMOD runs on scans patients are already getting, adding zero friction to the system
Conscious Take:
This is the kind of AI story that actually matters. Our news and feeds (mine atleast) can be filled with visuals/creative and things like that surrounding AI - but this is what happens when AI gets directly applied to the hard stuff.
Pancreatic cancer kills because we find it too late. REDMOD is changing that, on scans that already exist. More of this, please.
Image: Google | The Conscious Church
Google is officially swapping out Google Assistant for Gemini in every car with Google built-in — and this isn't just a rebrand. The upgrade brings genuinely conversational AI to your dashboard, which means you can finally talk to your car the way you'd talk to a person.
The Details:
The rollout starts in the US in English, covering both new cars and existing ones via a software update
You can ask for navigation help, control your car's climate and radio, manage messages and music, and ask questions pulled directly from your vehicle's owner's manual
A beta Gemini Live mode supports longer conversations for brainstorming and learning while you drive, with Gmail, Calendar, and Home integrations coming later
General Motors is rolling it out across approximately 4 million of its vehicles from model year 2022 onwards
Conscious Take:
Car assistants have been clunky forever, you'd ask Google Assistant to play something and it'd mishear you three times and open Maps instead, and it’s not just a Google thing, Apple CarPlay does the same for me.
Gemini is genuinely conversational, so this could finally make in-car AI feel less like fighting a bad voice menu. The awesome feature though is the owner's manual integration - AI that can answer "why is this warning light on?" from an actual source of truth. Pretty cool! Although I would just be happy with an early 90s diesel Landrover Defender 90 or 110 (anyone else?).
📬 One quick ask...
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.
That's all for now
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.
