👋 Hello hello,
❗Before diving in, this is for the business owners and entrepreneurs: we're building AI automation workflow guides: step-by-step, click-by-click, tool-by-tool so the repetitive work runs itself. Tell us what you'd automate.
This week a handful of companies decided to kill that trip entirely. Anthropic put Claude inside Slack as something you tag like a colleague. A scrappy email app started drafting your replies before you even open them. And a Mac tool now finishes your sentences in whatever app you happen to be typing in.
Let’s dig in.
🔥🔥🔥 Three Exclusive AI Updates
Anthropic launched Claude Tag, a version of Claude that lives in Slack as a shared teammate. You add it to a channel, give it access to the tools you choose, and anyone on the team can tag @Claude with a request. It breaks the task into stages, does the work, and posts the result back in the thread.
The interesting part is that it's one shared Claude per channel, so a teammate can pick up exactly where you left off, and it builds context as it follows the conversation. Turn on "ambient" mode and it starts nudging quiet threads on its own.

It runs on Opus 4.8 and is in beta for Enterprise and Team customers, replacing the old Claude in Slack app. The proof point Anthropic is leaning on: 65% of its product team's code now comes from its internal version.
Most AI video tools cap out around 15 seconds, so longer clips mean stitching pieces together and hoping the seams don't show. ByteDance previewed Seedance 2.5, which claims one continuous 30-second clip from a single prompt, plus support for up to 50 reference inputs to keep a character, product, or style consistent across the whole shot.
It's in enterprise beta now with a public launch targeted for early July, so treat this as a teaser rather than something you can open today. If the claims hold up under real testing, full ad spots and longer scenes in one pass get a lot closer.
Quick context: OCR is the tech that turns a scanned PDF or photo of a page into text a computer can actually use. Mistral OCR 4 goes further than plain text. It tags where each block sits on the page, labels what it is (title, table, equation, signature), and scores how confident it is, across 170 languages.
That structure is what makes documents usable for AI search, automation, and source-cited answers instead of a wall of jumbled text. It runs in a single container, so sensitive files can stay in-house, and pricing starts at $4 per 1,000 pages. The benchmark wins are Mistral's own numbers for now, so worth a real test before you commit.
🔥🔥 Two Tools Worth Trying
1. ✍️ Cotypist

You know how your phone finishes your sentences as you text? Cotypist does that for your Mac, everywhere you type. As you write an email, a Slack message, or a Notion doc, it suggests your next few words and you hit Tab to accept them. It learns your phrasing over time, so the suggestions actually sound like you instead of generic filler. The whole thing runs on your Mac, so nothing you write gets sent to the cloud. Best for anyone who churns out a lot of repetitive writing and wants the speed without handing their drafts to a server somewhere.
2. 📬 Upstream

An AI email app for teams that drafts your replies before you open them, using the thread, the recipient, and your past messages with that person. It also resurfaces important conversations at the right moment with a draft already waiting. Best for small teams living inside a shared inbox.
🔥 One AI Resource To Help You Learn
Two Anthropic engineers recorded a 24-minute walkthrough of the features that usually get missed. It teaches you almost everything you need to know about how to use Claude.
📋 Help shape Practically AI
We’re constantly refining the daily insights to make sure they actually help you in your day-to-day work. Your feedback would make this possible - it takes less than 2 minutes.
👉 Share your feedback here
Don't forget to rate today's post
Until next time,
Team @PracticalyAI
