Summary
This video demonstrates how to use Claude's "record a skill" feature to train it to perform a task. The creator records themselves writing a blog post, copying it, pasting it into Google Docs, and emailing it, which Claude then turns into a step-by-step instruction guide.
On-screen text
Claude
Claude can now learn
tasks by watching you 😳
CLAUDE CAN NOW
HELP YOU WITH
WATCHING YOU.
RECORD A SKILL
Don't type passwords or secrets, or display sensitive information or
private conversations while recording.
Start recording
So for this
TASK, I'M GONNA
AND THEN I'M
BLOG AND EMAIL
IT TO MYSELF.
So right here,
PLEASE WRITE ME A BLOG ON ALL THE NEW UPDATES FOR AI.
AI news updates July 2026
AI Updates Today July 2026 - Latest AI Model Releases
AI News Today July 30 2026 - 16 Biggest Stories
AI News Today July 4 2026 - 16 Biggest Stories
new AI tools for small business owners 2026
10 Best AI Tools for Small Business Owners in 2026: No-Code Setup Guide | MindStudio
8 Best AI Tools for Small Business Owners in 2026 You'll Actually Use - KBH Blog
Best AI Tools for Small Business Owners in 2026: Geeky Gadgets
Want to be notified when Claude responds?
Notify
Capturing 11 steps
Done
Untitled document
File Edit View Insert Format Tools Gemini Extensions Help
Work
New Chrome available
AI news updates July 2026 MD
Copy
The New Models Everyone's Talking About
Claude Sonnet 5 launched at the end of June and has quickly become the go-to
for agentic coding, debugging, and multi-step business workflows. It stood
out for coding, debugging, and multi-step tasks, with introductory pricing of
$2 per million input tokens and $10 per million output tokens through the end
of August 2026, before prices step up afterward.
Google's Gemini 3.5 Pro has been a different story - It's still sitting in
preview, and Google has pushed the release back to work on issues like
latency, and improve performance on existing models. Meanwhile, competitors
have been racing ahead, with Claude's early preview of GPT-5.6 to select
government partners all landed in the same window.
On the open-source side, things are moving just as fast. It's expected to open
its weights by July 27, and DeepSeek V4 has a stable release planned for July 24
- meaning the last month is shaping up to be a landmark stretch for free,
self-hostable models. Earlier in the month, Grok 4.5 has been running in private
beta at SpaceX and Tesla.
Access Is Getting Tighter Even as Tools Get Better
Here's the twist worth paying attention to the same month that brought all these
upgrades also brought more restrictions. The clearest summary of July so far is
that AI got more useful and more controlled at the same time - better agent
tools for daily work, tighter access rules, and much bigger spending on
compute infrastructure. The European Commission also adopted binding rules
under the Digital Markets Act requiring Google to open Android to competing AI
assistants and share portions of its data with rival developers. Approved third-
party AI assistants will get voice activation and cross-app access across 11
Android feature phones, and Google will have to share its recently announced
search, click, and
Untitled document.pdf (82K)
Describe your message
Send
Capturing 11 steps
Done
blog-to-pdf-email | SKILL.md
Save Skill
Description
Write a blog post on a topic, turn it into a polished PDF, and email that PDF to the
user (by default, to themselves). Use this skill whenever the user asks to "write a
blog and email it to me", "draft a PDF and email it", or any request that combines
producing a blog with delivering it as a PDF over email - even if they don't name
every step. Trigger in its phrases like "blog me X and email it over", "write up the
latest on that topic and email the PDF", or "just draft X, PDF it, and send it to
myself". Do NOT trigger for publishing straight to a CMS, for social captions, or
when no email, PDF delivery is intended.
Blog -> PDF -> Email
This skill captures a three-part workflow: research and write a blog post, render it
as a clean PDF, and email the PDF to the user. It was built with a focus on
reproducible hand-copying between apps - each each outcome with the fastest reliable
tool, not by replaying clicks.
The three outcomes:
1. A finished blog post (researched, well-structured prose).
2. The blog post as a PDF file.
3. An email to the user with the PDF attached.
Always deliver the PDF into the chat with "Send/File" as well, so the user gets it
directly even if the email step needs on the send button.
Step 1: Write the blog post
First job: down the inputs. If the user gave a topic, use it. If they didn't, ask one short
question (topic - roughly how long? rather than the whole prompt). Re-state the
assumption at the top of your reply.
For anything about the "process-driven" AI market, or "AI tools" for small business
owners, and where the user wants to know what's working, the answer is to write
a blog post. A blog post is a good way to get the information out there. The
user wants to know about the latest AI updates and developments. So, the prompt
should be "Please write me a blog on all the new updates for AI."
Then I'm going to copy and paste that blog into Google Docs and then we'll do
paste. Then we're gonna come to file and do download, a Gmail tab. Then open up a
Gmail tab. Come over here and drag that blog into this email and then press
send. Then Claude is gonna process that entire process, and you can see it
recorded step by step and even grab screenshots along the way. Then it literally
takes that entire process and builds a step-by-step instruction guide so that AI
can do it again.
Save this video,
follow for more AI tips.