Hey {{first_name|Conscious Church Fam}} 👋
Morning! Anthropic has given Claude a new app that keeps working while you sleep, an AI helped catch a security hole in most of the internet, and Microsoft's virtual doctors have been outscoring real ones.
Here's what we're covering:
Claude gets a sidekick that never sleeps
AI finds a hole in 70% of the web
Microsoft's AI out-diagnoses doctors
Let's dive in 👇
✍️ Josh's Musings
This weekend I saw a tweet about someone using Claude, the AI assistant, to build simple motion graphics. They didn’t use an animation app. They just described what they wanted. I couldn’t not give it a go, so I tried it with Scripture.
Here’s what came out: a 28-second painted short of Matthew 11:28–30.
Below is a short gif - click it to see the full video.
(For a first try it’s semi decent, but I wouldn’t use it as is.)
How it works, in plain English:
You give it a verse and a look. “Matthew 11:28–30, oil paint, like the faith shorts on Instagram.” That’s the whole brief.
It plans the shots. Thirteen pictures, a new one every two or three seconds, each matched to the words: a tired man on a dusk road with a sack on his back for “heavy laden”, a heart-shaped pond for “lowly in heart”, and a dad holding his little boy’s hand for “my yoke is easy.”
It draws every scene in code. Skies, hills and people are simple flat shapes, like paper cut-outs. There’s no AI image generator involved, so nothing is borrowed from other artists’ work.
Then it paints them. Each drawing is repainted in brushstrokes twelve times a second, so the paint gently shimmers like hand-drawn animation.
It gives it a voice. A warm British AI narrator reads the verse. The voice runs on your own computer, with no account and no subscription.
It writes the music. A soft score is composed fresh by code for each video, with no stock tracks and no samples. The music dips under the narrator so every word stays clear, and the whole thing is levelled to the volume TikTok and Instagram expect. It adds captions, a few words at a time, timed to the voice.
Before the final render, it shows you one frame from every shot so you can check them. It caught a few problems itself. My favourite: the labourer’s sack was so big it looked like he was carrying a boulder.
The tool is a Claude Code add-on called Shorts Animator (from here). The catch: it needs a paid Claude plan and someone happy to type into a terminal or in claude code.
Could be a fun way to present bible verses or a quote…
Stay Wild, Stay Conscious.
Josh
Image: Anthropic | The Conscious Church
Anthropic has launched Claude Duet, a separate app that works for you in the background. It sends a morning brief, keeps an eye on your inbox and calendar, and gets on with tasks on its own cloud computer, so your laptop can be shut and your phone in your pocket.
The Details:
It runs on Opus 5.5 with its own browser on a virtual machine, and pings you when a job is done or needs a look.
It plugs into 6,000+ connectors including Gmail, Slack, Canva and Shopify, and brings across your memory and skills from the main Claude app.
It can buy things using a one-off virtual card from Stripe's Link wallet, so your real card number is never shared.
Anything that sends, pays, signs or deletes needs your OK first. It's a research preview for US Pro and Max subscribers on a waitlist, at no extra cost.
Conscious Take:
Claude and ChatGPT and th other frontier models are getting better each week. They are doing more and more and this release shows more of the agentic work to ‘support’ you in your regular workflows and tasks. Thankfully payments require confirmation.
Image: OpenAI | The Conscious Church
PHP is the code behind roughly 70% of the websites you visit. This week its maintainers rushed out fixes for two critical flaws that were found with help from OpenAI's GPT-5.6 Sol, and some governments took their sites offline while they patched.
The Details:
Security firm Calif found the first bug in August using GPT-5.6 Sol, and Wiz separately caught both a few weeks later.
The bugs were around a decade old and let attackers take over servers without logging in.
OpenAI spent two months working with PHP's maintainers on fixes before going public, and says more big finds are on the way.
Finland, Germany and the Netherlands pulled government sites offline, and the NSA told anyone who can't patch straight away to take their systems offline.
Conscious Take:
Good news this time, the good guys got there first. BUT the same tools work for the bad guys too, so update your phone and laptop this weekend and turn on two-factor login if you haven't.
Image: Microsoft AI | The Conscious Church
Microsoft has published results for MAI-DxO, an AI system that works like a panel of virtual doctors. They suggest diagnoses, order tests and argue with each other as results come in, and on paper they're beating real physicians while spending less on tests.
The Details:
On 304 hard cases from the New England Journal of Medicine, it hit 83% accuracy against 36% for doctors using reference books and the internet.
On everyday A&E cases it scored 80%, compared with 47% for physicians.
One version reached 81% on the hard cases at less than half the cost of similar systems, mostly by ordering fewer tests.
The research is published in Science, and trials with real patients start at Beth Israel Deaconess in Boston.
Conscious Take:
Anyone who's sat in A&E for six hours will be up for this. It's still a test though, so I'd want to see how it does with real patients in Boston before getting too excited.
📬 One quick ask...
If this email has been helpful, would you forward it to one person this week who might be interested? Thanks :)
"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.
