Mon, October 26, 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.
On Friday, 50% of you said your site’s PHP is patched and 25% had no idea. If that’s you, the one email to your host works just as well on a Monday.
Still pasting the same long prompt into every new chat? Anthropic just handed the format that replaces that habit to the Linux Foundation, and more than 70 AI tools already read it.
IN 3 MINUTES, GET 1% BETTER THAN EVERYONE YOU KNOW:
Write instructions once, use them in any AI.
Turn your blurry logo into an editable file.
Ask your Gemini notebooks from any Gemini chat.
De-AI your website, then let an agent audit it.
Get a panel of experts instead of one answer.
3 Did you see
Image: Agentic AI Foundation
Anthropic donated Agent Skills to the Linux Foundation’s Agentic AI Foundation on Oct 22. A skill is a folder with a SKILL.md file that teaches an AI agent one task, and the agent loads it only when the task calls for it. More than 70 platforms already read the format, including ChatGPT, Cursor, GitHub Copilot and VS Code, and Anthropic counts nearly 280 million installs. Engineers from Anthropic, Google and OpenAI will now maintain it together.
A skill you write once (your brand voice, your report format, your client checklist) now travels to whichever AI you use next, so stop re-pasting the same long prompt into each tool.
Your best prompt just got a passport.
How often do you paste the same long prompt into an AI?
Ideogram released Ideogram Vector 1.0 on Oct 22, its first model that outputs vector files instead of pixels. Describe a logo, label, icon or illustration and it returns a clean SVG you can recolor and reshape in Figma or Illustrator, or upload an existing image to convert it. It’s in the Ideogram app and API, where an SVG costs $0.13 to $0.32 and converting an image costs $0.10.
Take the blurry PNG logo you’ve been emailing around and turn it into a file your printer or designer can actually use.
Your printer can stop sighing at your attachments.
Gemini Notebook is now a Connected App inside the Gemini app, per Google’s Workspace blog (Oct 23). Type @Gemini Notebook in a chat to create a notebook, add sources (including a copy of the chat you’re in), save notes or ask what your notebooks say about a topic. It’s on by default for Workspace and personal Google accounts, except in the EEA, where you switch it on under Settings > Connected Apps.
When a chat turns into research worth keeping, say “add this chat to my notebook about ___” instead of copy-pasting.
Your research stops dying in chat history.
2 things you didn't know you can do with AI
1. 🖌️ Make your AI-built website stop looking AI-built · Hallmark · free, open source · Claude Code, Cursor, Codex, OpenCode
Hallmark is a free design skill from Together AI. It picks three themes from a catalog of 48, builds the page, then checks it in a real browser for the tells that read as machine-made. Point it at a page you already have and it either reports what’s wrong or redesigns it while keeping your content and links. The catch: it runs inside a coding agent, not a chat window.
2. 🛡️ Have your coding agent security-audit your own site’s code · Cloudflare security-audit skill · free, open source · coding agents
Cloudflare open-sourced the security-audit skill behind its own vulnerability-hunting system. Your coding agent maps the app, hunts for holes, has fresh agents try to disprove each finding, and writes a plain report of what survives. After last week’s PHP flaw, it’s a cheap first look before you pay for a professional test. The catch: it needs an agent that runs parallel helpers, plus a locked-down sandbox before it will run your code.
1 I Wish I knew sooner
Microsoft’s MAI-DxO, published in Science on Oct 22, didn’t win with a smarter model. It ran a panel: AI roles propose answers, ask for what’s missing and challenge the leading view over several rounds. It got 83% of hard NEJM cases right, against 36% for physicians with references, and the gains held on newer models.
You can borrow the panel for any big business call, like pricing, a hire or a vendor. The medical result is simply the proof that the pattern works.
Before your next big call, try: “Play a CFO, a skeptical customer and an operator. Each argue your view, then poke holes in the others for two rounds. End with one recommendation and what would change your mind.”
The catch: three roles played by one model can share one blind spot, so feed it your real numbers and argue back.
Three skeptics, zero meeting invites.
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.
