👋 Hello hello,
Anthropic dropped Claude Opus 4.8 today. Same price as before, sharper judgment, better self-awareness, and the ability to run longer autonomous tasks. Which, honestly, feels less like a model update and more like someone quietly handed your AI intern a promotion.
While that was happening, Satya Nadella posted about a redesigned Copilot — faster, simpler, and supposedly built to keep you in flow. And Claude Code quietly shipped something that's going to make a lot of developers very, very productive (more on that below).
Let's get into it.
🔥🔥🔥 Three Curated AI Updates
Anthropic just released Claude Opus 4.8, and the headline is this: it's better at working on its own for longer. It builds on Opus 4.7 with sharper judgment, more honesty about its own progress and limitations, and improved ability to run extended autonomous tasks without needing you to babysit it.
The price hasn't changed, which makes this a no-brainer upgrade if you're already using Claude for complex workflows. If you've been testing Opus 4.7 in agentic setups — document processing, multi-step research, code review — 4.8 is worth another look.
P.S. Kushank is testing this out and will share first impressions on YouTube Shorts. Stay tuned.
Satya Nadella posted the redesign himself, which tells you Microsoft wants people to actually notice this one. Copilot has been rebuilt to be simpler, faster, and more intuitive — with less friction between you and the task at hand.
The goal is to keep you in flow, not interrupt it. Whether the redesign delivers on that is worth testing for yourself.
What do you think of the Copilot redesign?
P.S. We've got a whole channel dedicated to Copilot tips, walkthroughs, and updates. Consider following if you want to stay ahead of everything Microsoft AI.
This one's for the builders. Claude Code just shipped dynamic workflows in research preview, and it's a genuinely big deal. You describe a complex task, Claude writes an orchestration script on the fly, then spins up a coordinated fleet of subagents to tackle it in parallel.
Practically speaking: tasks that used to take hours of back-and-forth can now be broken down and executed simultaneously by multiple agents working together. To trigger it, just use the word "workflow" in your prompt.
This is early access, so rough edges are expected — but the direction is clear.
🔥🔥 Two Pro AI Tools
Firecrawl already lets you extract web content cleanly for AI use. Now they've added monitoring: enter any URL, describe what you want to track, and Firecrawl notifies your AI agent via webhook the moment that page changes. The smart part? It uses up to 90% fewer LLM tokens by only sending what actually changed, not the whole page every time. Perfect for anyone building agents that need to react to live web data.
2. 📱 Anything
Grab a screen from Mobbin (a design reference library with thousands of real app UI screenshots), drop it into Anything, and it builds a working app from it. If you've ever spent hours trying to describe a UI you already had in your head, this cuts straight to the result. Great for indie builders, founders prototyping fast, and devs who want a head start on frontend.
🔥 Learn How To Build an AI Agent
Kaggle and Google are running a free 5-day intensive on building AI agents with vibe coding, running June 15–19, 2026. It's structured, hands-on, and built for people who want to go from understanding agents to actually shipping them.
Here's how to get started:
Head to the Kaggle competition page and register for the course.
Follow along each day — the course is structured as a daily sprint, so one session at a time.
Build progressively: each day adds a new layer to your agent, from setup to full orchestration.
Use the community notebooks and Google's resources to debug and iterate fast.
By day five, you'll have a working AI agent you built yourself, not just watched someone else build.
Did you find today's post useful?
Until next time,
Team @PracticalyAI
