Thu, November 12, 2026
3-minute read
3
DID YOU SEE
2
THINGS YOU DIDN'T KNOW YOU CAN DO WITH AI
1
I WISH I KNEW SOONER
👋 Good morning.
Yesterday 45% of you said you’d let Claude build your team a live spreadsheet, and 0% picked staying in Excel. Today’s helper is quicker. Google’s new Decisions API reads a photo of a breakroom counter and picks wipe-it-up or call-the-janitor in 150 to 250 milliseconds, per Google. If your morning starts with “is the kitchen client-ready,” the answer just got cheap.
IN 3 MINUTES, GET 1% BETTER THAN EVERYONE YOU KNOW:
Gemini answers 1,000 yes-or-no questions in one pass.
Gemini rides along in Apple CarPlay.
Adobe’s experiments let you step inside a photo.
Describe a phone widget or line-art scene, get one.
One prompt finds the work your AI should do.
3 Did you see
Image: Google
Google launched the Decisions API, a Gemini tool for developers that scores answers instead of writing replies, per Google. You hand it one piece of material (text, a photo, audio, video or a PDF) and up to 1,000 questions. It returns yes-or-no, multiple-choice or rubric scores, each with a confidence score from 0 to 1. Google quotes 150–250 milliseconds and $0.10 per million input tokens, with answers unbilled. It runs in AI Studio and the Gemini API, and Google Cloud support is still coming.
Do this: list ten yes-or-no checks your team makes by hand, then try them on AI Studio’s no-cost tier.
Your breakroom spill just got a confidence score.
If an AI could score a yes-or-no for you in a blink, what would you try first?
The Gemini app started rolling out on Apple CarPlay worldwide on Wednesday, for every Gemini user on iPhone, per Google. You can ask about your calendar, hear unread email summarized, or rehearse a pitch out loud. The chat syncs to your phone after you park. You need the latest Gemini app, microphone access, and one past Gemini Live session on your phone. Daily Brief and some personalization features are off for under-18s, the European Economic Area and Nigeria.
Do this: open Gemini Live once on your phone today, so it’s ready the next time you plug into the car.
Your commute just became office hours, whether you wanted that or not.
Adobe showed its experimental “sneaks” at MAX, per The Verge. Project Sight Span turns a single image into a 3D scene you can step inside and reframe. Project Rinse Repeat copies an edit from one part of a photo to another. It rebuilds the edit with native Photoshop tools, so it stays editable, per Adobe. Project Screen Test lets you talk to an AI character like an audition. These are experiments, with no promise any will ship.
Do this: if you retouch group photos or product shots, Rinse Repeat is the one to watch.
Your group photo just learned to share a good edit.
2 things you didn't know you can do with AI
1. 📱 Describe a phone widget and Gemini builds it · Create My Widget · Pixel 10/11
Create My Widget builds a home-screen widget from a description, like a monthly spending tracker or a race-day countdown. It can pull from Gmail, Calendar and Search. The catch: Pixel 10 and 11 only (not the 10a), in English, in the US, UK, Canada, Australia and India, per Android Authority.
2. ✏ Animate a line-art scene from one sentence · line-art-animation · free · Claude Code skill
line-art-animation is a free, open-source Claude skill that animates line art from one sentence, like a baker at dawn. It outputs SVG, so the result stays sharp at any size, per its README. The catch: you install it by cloning a GitHub repo into Claude Code, not by opening a website.
1 I wish I knew sooner
🔍 Ask your AI what it should take off your plate · Alex Finn · Nov 10
Most of us hand an assistant only the tasks we already thought of. Alex Finn’s prompt flips that, per his post. It asks the assistant to review what it knows about you, then propose systems, routines and subagents (helper agents) to build. Finn runs it in an agent, and says you can just send yours the post.
My goal for working with you is to lighten my cognitive load. I want to get more work done, but with less effort on my part. Based on what you know about me and what we do together, what are some things we can do, systems we can implement, subagents we can build, or routines we can create to accomplish this goal?
The catch: it only knows what you’ve told it, so a fresh chat with no memory gives generic answers.
Your to-do list just got a suggestion box.
Don't forget to rate today's post
That's your 1% for today.
Kushank, AI teammates, and the human team.
PracticalyAI
Looking to hire a team to build AI solutions for you? Here’s the team I recommend.
