In partnership with
In this issue:
📸 ChatGPT scans paper
🔀 Claude in Copilot
👕 ChatGPT try-on
🧩 Gemini Skills
🎯 Lokesh's one-task helper
⚡ 1-prompt PDF deck
📸 ChatGPT can now scan a stack of paper into 1 PDF
You photograph a handout page by page, paste 6 pictures into ChatGPT, and then it asks which one to read.
Since October 1, the ChatGPT camera on iPhone and iPad has a Scan option. Shoot the pages in order, they land as 1 PDF, and you can ask for "only the numbers from the table on page 3." (No Android date yet.)
One catch: the page goes straight to ChatGPT. So anything with a client's name on it, check your company rules first.
An IT engineer who writes about AI work systems shows where the button hides and how they use it for handouts and whiteboard photos.
(It's in Japanese, so use your browser's translate.)
🔀 Copilot now lets you pick Claude Sonnet 5.5 or GPT-6.1 Sol
Copilot hands you a weak draft in Word, and you start fixing it by hand.
Microsoft is adding GPT-6.1 Sol and Claude Sonnet 5.5 to the model picker in Word, Excel, PowerPoint and Chat. It's rolling out in phases, so you might not see them yet.
Office Watch's advice: leave Copilot on Auto. When a draft comes back weak, run the same prompt on the other model before you rewrite a word.
Can't find Claude in the list? Your admin has to switch it on, and some companies keep it off for data reasons.
👕 ChatGPT can show you wearing it before you buy
You ask ChatGPT to find a jacket, then try to picture it on you from a photo of a model.
Now some clothing and accessory listings in ChatGPT have a Try on button. Tap it, take or upload a selfie, and ChatGPT Images puts the item on you.
No button yet? Attach your photo and a picture of the item, then type: "Show me wearing the clothing from image 2, keep my face."
Heads up: ChatGPT keeps your selfie for the next try-on. Delete it under Settings > Personalization > Reference photos.
Proje Defteri, a Turkish how-to blog, has the 3 steps plus the selfie tips (clear face, no hats or sunglasses).
🧩 Gemini Skills: run your saved prompts with a /
You built a Gem to proofread, so every draft means opening a separate chat.
Google is adding Skills. Type / and a skill's name, and it runs on the text already in your chat. You can stack 2 in one prompt.
It's rolling out in Workspace now (some accounts wait until October 19), and the Gemini app gets it October 13. Your Gems keep working until at least March 1, 2027.
To make one, type "create a skill for me that does..." and describe the job.
Khamosh Pathak at Android Authority stacks /fact-check-links and /de-jargon on one email, no second chat.
Stop digging through LinkedIn for marketing news
TLDR Marketing is the free daily email that does the reading for you. Each issue summarizes the most interesting stories in growth, martech, and digital marketing, curated by subject-matter experts so you get the tactics without the noise.
It's one email instead of 20 open tabs and an hour lost in your feed. A 5-minute read with your morning coffee, and you walk in already caught up.
We cover the channels that actually move your numbers: paid, SEO, email, social, and ecommerce. Whether you work in B2B or B2C, in-house or agency-side, follow what matters to your work.
Free, daily, and read by 330K+ marketers. Subscribe for free and get your mornings back.
🎯 One task at a time
I have ADHD. I start a task, a new idea comes, and the first task stays half done.
So I built a small helper in Claude. You dump your messy list, and it shows you one task with a timer. New ideas get parked, and you go back to your task. At the end of the day you get a short report, and tomorrow's chat starts where you stopped.
I tested it on the free plan. Setup takes about 5 minutes, and the guide shows every click.
⚡ Use this Tuesday: one prompt, 5 slides, 1 clean PDF
Does this ⬆ sound like your Monday?
Its time to change. See how I make PDF decks these days. I paste one prompt into ChatGPT and come back to a finished PDF.
My test took 3 minutes 13 seconds. I got 5 branded slides and a PDF where every page is exactly the slide size, with no white border, header or footer. It also gave me a zip of the slides as PNGs, so I can drop one into an email or a doc.
Here's the full chat. You can see what I got and copy the prompt:
👉 https://chatgpt.com/share/6ac4f940-ba40-83ee-88b0-406a0ac946d6
(The prompt sits in a collapsed box at the top. Click "Show more" to see all of it.)
📌 What to change before you paste it
Everything else stays the same. You only change these:
Slide size.
1920x1080is a normal 16:9 presentation.Your colors. Swap the 3 color codes (
#F7F7F5,#1E3A5F,#F2A33A) for your company's. Don't know them? Ask your design or marketing team. They get this question a lot.Your logo text. Change
"ACME Ops"to your team or company name. (This writes the name as text. More on real logos below.)Font style. "Clean sans-serif" works for most offices. Change it only if your company has a set look.
Your slides. Rewrite the title and bullets for each slide. Keep it to 3 bullets max, or the slides get crowded.
Number of slides. If you have 7 slides, change "5" to "7" in both places: STEP 1 and STEP 2.
Leave the last line alone: "If you can't access the generated images in Python, tell me. Don't fake the PDF." That line stops ChatGPT from pretending the PDF worked. If it says it can't reach the images, download them, upload them back, and paste only STEP 2.
Before you send the deck, read every slide once. Mine had zero typos. Image models still slip on spelling sometimes, and your manager will be the one to spot it.
📌 What worked, and the one thing I haven't tested
It really worked well. The branding stayed the same on all 5 slides, and the PDF came out the right size.
The one part I'm not sure about is a real logo. I only typed my logo name as text. I don't know yet if ChatGPT keeps an uploaded logo file sharp and in the same spot on every slide.
So try it 2 ways:
Without a logo: run the prompt as it is.
With your logo: upload your logo file with the prompt, and change the logo line to
Logo: use the uploaded logo, small, bottom-right corner, same size on every slide.
Then hit reply and tell me how each one came out. Did the logo stay sharp? Did it move around? A screenshot helps a lot. I'll share what you send (the wins and the fails) in a future issue.
And if you make Instagram or LinkedIn carousels, this same prompt should work there too. Change the size to 1080x1350 and swap in your carousel text. Tell me if you try that too.
👋 Before you go
How was today's issue?
Did you see where this week's how-tos came from? A blogger in Japan. A how-to blog in Turkey. Office Watch. Android Authority.
I credit these creators every week because they write the steps. What happened in AI matters less to me than how people actually use it. And now you get that right here, inside your email app.
Next week is about video editing.
I don't make AI videos. I use AI to make images, but prompt-to-video feels like slop to me. If I make a video, I want a real human face in it.
So I used Opus 5.5 to build a video kit for Claude. You give it your own clips and describe the edit in your words. It cuts the pauses, adds word-by-word captions and drops in sound effects. You don't need a video editor.
One thing I learned building it: the brief does the heavy lifting. Same clip, no brief, and the edit came out generic.
So what's one video you've wanted to make but never edited? Hit reply and tell me.
Dan Rice · AI Signal · Every Tuesday
Read once. Use AI better all week.