Claude Just Changed: The 10 Big Updates You Need to Know
A 13-minute walkthrough of every Claude launch from one month, with a verdict on who should try each one and who can skip it.
Posted
yesterday
Duration
Format
Review
educational
Views
3.7K
112 likes
57 · 43
Big Idea
The argument in one line.
Anthropic shipped ten launches in one month, and the useful move is to sort them by what you actually do: pick one model per job type, then try only the tools that match your daily work.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use Claude every week but have no time to read each Anthropic announcement and want a quick sort of what matters.
You are non-technical and want to know which of the new Claude tools (docs, decks, dashboards, video) you can try today without code.
You use Claude Code and want to know what Projects, the wrap-up allowance and mods change in practice.
SKIP IF…
You want hands-on setup walkthroughs. This is a fast overview of each launch, not a tutorial.
You only use Claude through the API and do not touch the app or Claude Code.
TL;DR
The full version, fast.
Anthropic released ten things for Claude in one month. Three new 5.5 models split into Opus for long hard jobs, Sonnet for everyday work and Haiku for cheap repeated tasks. Cowork and chat merged into one Claude that decides how big a job is, and Docs, Slides, Design, Dashboards and Motion all build from a normal chat. Claude also works inside Google Docs, Sheets and Slides and on CarPlay. Claude Code gained goal-based Projects that run parallel threads, a wrap-up allowance that lets a response finish when you hit your limit, and mods that reshape the interface. If you try one thing, ask Claude to build a deck from your notes.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Duncan previews the highlights, states the format and introduces himself.
00:49 – 02:23
02 · New models: Opus, Sonnet, Haiku 5.5
Three sizes of the same brain, with a rule of thumb for which to use when.
02:23 – 03:30
03 · Cowork and chat are one Claude
One place for quick questions and big background jobs, with scheduling and permission controls.
03:30 – 04:53
04 · Docs, Slides and Design
Three tools that build documents, decks and designs from a normal chat.
04:53 – 05:55
05 · Claude Dashboards
Describe the numbers you want and Claude builds live charts with traceable sources.
05:55 – 07:02
06 · Claude Motion
Animated video from a prompt, built from code rather than an AI video generator.
07:02 – 08:15
07 · Claude in Google Docs, Sheets and Slides
A sidebar in Google files that edits with permission and respects version history.
08:15 – 08:57
08 · Claude on CarPlay
Voice-only Claude for your commute, with two stated limits.
08:57 – 10:22
09 · Projects
Goal-based Claude Code projects that split work into parallel threads.
10:22 – 10:51
10 · The wrap-up allowance
A small extra usage buffer so a response can reach a safe stopping point at the limit.
10:51 – 12:45
11 · Claude Code mods
Add-ons that reshape how Claude Code looks and behaves, from games to safety logs.
12:45 – 13:25
12 · Recap
Quick summary of all ten launches and one thing to try this week.
Atomic Insights
Lines worth screenshotting.
If you are unsure which Claude model to pick, pick Sonnet 5.5: it handles documents, slides, spreadsheets and coding at lower cost than Opus.
Haiku 5.5 is the cheapest and fastest model Anthropic has made, best for small tasks you run over and over like captions or sorting ideas.
Cowork and chat merged, so you no longer decide where a task belongs: Claude judges whether it needs a quick answer or a bigger background job.
Claude's bigger jobs can keep running after you close your laptop, and you can check on them from your phone.
Docs, Slides and Design are out of beta and work on every plan, including free, all from a normal chat.
Claude Dashboards let you describe the numbers you want in plain words, and clicking any number shows exactly where it came from.
Claude Motion does not use an AI video generator. It writes code that animates your own words, charts and images, so everything stays editable.
Claude in Google Docs, Sheets and Slides asks permission before each edit, and every change lands in normal version history so you can undo it.
Claude on CarPlay is voice only, cannot control your car or phone, and has no wake word, so you open the app each time.
Claude Code Projects are built around a goal, not a folder of chats: a lead Claude splits the goal into jobs that each run in their own thread in parallel.
The wrap-up allowance gives a response in progress a small extra amount of usage to reach a safe stopping point when you hit the five-hour limit, and it counts against your weekly limit.
A Claude Code mod is like a browser extension for the tool: it can change what you typed, block or approve actions, or add a panel to the screen.
Mods have the same access to your computer as Claude Code itself, so only install ones you trust.
Takeaway
Match each launch to the work you do
WHAT TO LEARN
Most of this month's Claude launches only matter if they match your daily work, so sort them by job instead of trying everything.
02New models: Opus, Sonnet, Haiku 5.5
Default to Sonnet for everyday documents, slides and spreadsheets, switch to Opus for long hard jobs, and keep Haiku for small repeated tasks.
03Cowork and chat are one Claude
Because Cowork and chat are merged, describe the outcome you want and let Claude choose between a quick answer and a background job.
04Docs, Slides and Design
Docs, Slides and Design build from a normal chat and save to the artifacts tab, and all three work on every plan including free.
05Dashboards and Motion
Dashboards and Motion are in beta and plan-limited, so check your plan and that your data tool is supported before planning around them.
06Claude in Google Docs, Sheets and Slides
In Google Workspace, Claude edits only after asking permission and every change sits in version history, so edits are reversible.
07Claude on CarPlay
CarPlay is voice only with no wake word, which makes it good for thinking out loud on a drive and nothing more.
08Projects and the wrap-up allowance
Claude Code Projects pay off on jobs that touch many places at once, because a lead Claude runs the pieces in parallel threads.
The wrap-up allowance reduces half-finished edits at the limit, but it draws on your weekly usage and may not be enough to finish.
09Claude Code mods
Mods let you reshape Claude Code without writing code by asking for one, but they run with your computer's access, so install only trusted ones.
Glossary
Terms worth knowing.
Cowork
Claude's separate workspace for bigger multi-step jobs. It has now merged into the main chat so Claude decides which mode a task needs.
Wrap-up allowance
A small extra amount of usage that lets Claude finish a response already in progress after you hit the five-hour limit, instead of stopping half done.
Claude Code mod
A small add-on that listens for events like tool use or permission requests and can change input, block actions or add things to the screen.
Claude Code Project
A goal-based workspace where a lead Claude splits work into parallel threads, each on its own copy of your code, with shared memory.
Artifacts tab
The side panel in the Claude app where documents, decks and designs that Claude builds are saved so you can edit and share them.
“If you only try one thing this week, ask Claude to build you a deck from your notes.”
single concrete action→ IG reel cold open↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
analogy
Anthropic shipped 10 major new things for quad this month, and it's easy to miss half of them. So I went through every release and boiled it all down. So you don't have to read it yourself.
Here's a couple of the things I am most excited about. You can now make an animated video from a prompt with no video software, you can build a full deck doc or design right inside a normal chat. And you can talk to quad from your car.
Plus, there's one quad code update, I can't wait to show you and I'm saving it for last. For each launch, I'll tell you what it is why it's useful and whether it's worth your time.
If you use cloud and you don't have time to read every anthropic announcement, this video is for you. I'm Duncan and I spent 15 years as an art director and motion designer for brands like Apple, PlayStation and Nissan. And now I teach people how to build cool things with cloud even if you can't code.
So let's dive in starting with all the new models. The first big change is a brand new family of models and it can make your work faster and cheaper. A model is the brain behind cloud and this month there are three of them, Opus 5 .5, Sonnet 5 .5, and Haiku 5 .5.
Think of them as three sizes of the same brain. Opus 5 .5 is the biggest and smartest model. It's built for hard jobs that take a long time, like deep research that you don't have to babysit.
It's also what I've been using for creative thinking and for editing videos. Anthropic says it costs about 40 % less than the last Opus. Sonnet 5 .5 is the everyday model.
It's faster and cheaper than Opus, and it's very good at what most people do, which is making documents, slides, and spreadsheets, plus coding. If you're not sure which model to pick, pick Sonnet. Haiku 5 .5 is the small, quick, cheap model.
It's the cheapest and fastest model Anthropic has ever made, so it's great for simple jobs you run over and over or as a helper that handles small tasks while a bigger model does the thinking. So here's how I'd use them. Start with Sonnet for your everyday work, like writing a script, building a deck, or putting together a spreadsheet.
Reach for Opus when a job is long and hard, like creating an Instagram carousel or motion graphics. And then save Haiku for the small stuff you run. all day like writing quick captions or sorting a list of ideas either way you get better results for the same money or the same results for much less if you want to master cloud code in 45 days i co -founded the cloud code club you can get access to all my resources already done for you and learn with 8 000 other builders the links in the description the second thing anthropic announce is going to completely change how you use the app co -work and chat are now one Claude.
So here's what that means. Before, Claude had two places, chat for quick questions and co -work, which is a big separate space for bigger multi -step work. You had to decide which one your task belonged in and your work didn't carry over between them.
Now it's all in one place and you just ask. Claude figures out whether you want a quick answer or a bigger job and it just does the right thing. For bigger jobs, Claude can keep working in the background even after you close your laptop and you can check on it from your phone.
You can also set a task to repeat on a schedule like a report every Monday morning. By default, Cloud asks before it takes an action, so you stay in control the whole time. If you trust it more, you can switch it to only check in when it needs you.
And if you used Cowork before, your chats, projects, connectors, and skills all came with you, so there's nothing new to set up. If you ever stop to wonder whether a task belonged in chat or in Cowork, you can stop wondering. You just say what you want done.
If you never touch Cowork, this change matters less to you. After that, Anthropic shipped Docs, Slides, and Design, three tools that are about to change how you make documents, decks and graphics.
All three work the same way. You just ask Claude in a normal chat and it builds the document, deck or design. Then Claude saves it in the artifacts tab on the side of the app so you can find it, edit it and share it.
So first docs. Claude writes a document with you and your teammates can work in it at the same time. Quad drafts sections, asks you questions, and leaves comments.
It's private until you share it, and you can export it to Word, PDF, or Google Docs. Use it for something like a guide for new hires or a plan for a campaign instead of copying chat answers into another app. Second is Slides.
So give Claude your notes, a report, or a file, and it builds a deck for you. You can edit any slide yourself or ask Claude to change it. You can even present from inside Claude or download the deck as PowerPoint or PDF.
Slides is the tool to reach for when you have a meeting Monday and a pile of notes. third is design it makes visual things like mock -ups one -pagers pitch decks and landing pages using your own brand style design now works in any chat so there's no separate app to open all three are out of beta and they work on every plan including the free one so you can try them today and your documents decks and designs now come from the same chat where you already think The next feature Anthropic shipped is dashboards and it changes everything for people who've ever had to wait on a data person.
Dashboards are still in beta and a dashboard is a page that shows your numbers in charts so you can see how things are going at a glance. Normally someone has to build one and that takes a data person and a lot of time. With Claude you describe what you want in plain words like show me this quarter's revenue by region or how many people signed up this week compared to last month.
Claude writes the questions that pull the numbers from your data, builds the charts, and keeps them up to date. It works with tools like Snowflake, BigQuery, and Salesforce. You can trust what you see too.
Click any number and Claude shows you exactly where it came from, and every chart shows when it last refreshed. If you want to go deeper, you can send the dashboard to other tools like Mixpanel, Postog, or Grafana. Dashboards are on the paid plans.
If you've ever waited on an analyst for one simple number, you'll want this one. And if your numbers don't live in a tool like the ones I just named, just check that Cloud connects to it before you get all excited. After that, Anthropic announced Motion and it makes video as easy as typing a prompt.
Motion is also in beta right now, and it makes animated videos from a single prompt. So say you have a quarterly report and you want a 30 second video to explain it to your team. All you have to do is describe it and Cloud builds it.
But here's the part that surprised me. Motion doesn't use an AI video generator, so there's no fake footage and no fake people.
Instead, Claude writes code that animates your own words, numbers, charts, shapes, and images. Because it's code and not a flat video, everything stays editable. If you want to change a number, change it.
If you want the timing faster, just ask Claude. Nothing is baked in or permanent. When you're happy, you can export it as a video file or open it in tools like Descript or Runway.
To start, just type slash motion in a chat. Good uses are an explainer for an all hands meeting, adding movement to a chart and a board deck or a quick walkthrough for new customers. Right now, motion is only available on team and enterprise plans.
So if you've ever put off an explainer video because you didn't want to learn video software, definitely keep an eye on this one. The next big change puts Claude inside of Google Docs, Sheets, and Slides, and it's going to make your work faster because you stop copying things back and forth. So here's how it works.
You install Claude from the Google Workspace Marketplace. Then in any Google file, you click Extensions, then Claude, and a Claude sidebar opens right next to your document. So you don't have to copy things back and forth anymore.
You just talk to Claude while you're looking at the file. In Docs, you can say, turn my notes on. next steps into a table with an owner and a due date.
And it does. In Sheets, you can say, find out why this formula shows an error and fix it. In Slides, you can say things like, make all my titles the same style and under eight words.
Cloud asks your permission before each edit, so it won't change things behind your back. And every change shows up in your normal version history, so you can undo anything. There's also a second piece.
Cloud can create and edit Google files from a normal Cloud. chat all this is beta right now on pro max team and enterprise plans so if you spend your day in google files this is a big one because you get cause help without ever leaving them if you don't use google much you can skip it next anthropic also put claude on carplay and it's a fun one your drive time now becomes your thinking time if you don't know carplay is the screen in your car that shows your iphone apps and claude now works on it so open quad from it and you can just talk it's voice only so your eyes stay on the road and you can start a new chat or pick up a recent one where you left off.
Use it to think out loud on your commute, brainstorm a video, or ask questions and get spoken answers. There are two limits. Quad can't control your car or your phone and there's no wake word so you open the app each time.
Anthropic didn't announce this one so everything I'm telling you comes from news coverage so just make sure your quad app is up to date and then try it on your next drive. The last three updates are all for Cloud Code, starting with a big redesign of projects that's going to change how you handle big jobs. The redesign projects is in beta.
Before, a project was basically a folder of chats. Now, a project is built around a goal. You tell Cloud the goal and point it at your code.
Then a main cloud, like a team lead, splits the goal into smaller jobs. Each job runs as its own thread on its own copy of your code working at the same time as the others. The team lead keeps of progress, reviews what comes back, and puts all the results together.
The project also has shared memory so Quad remembers decisions like why you dropped something or what date changed. This is Anthropic's own example if you want to remove an old version of your product from your website, your app, and your backend. Instead of doing three jobs one after the other, Quad runs one thread for each and tells you which changes need to go in first.
You can check on it and steer it from your phone and it keeps running after you walk away. The work can run in the cloud or on your own computer and every pro and max user can get it right now projects are built for big jobs that touch a lot of places like that product example because the work gets done in parallel and you manage it all in one place right now it's tied to cloud code and anthropic says it's coming to everyone soon so if you're not using cloud code yet this one can wait the next update is the wrap -up allowance and it's small but it's going to save you a lot of headaches So, Claude has a usage limit that resets every five hours.
Before, if you hit it in the middle of a task, Claude just stopped and your work was left half done. Now, if Claude is already in the middle of a response, it gets a small extra amount of usage to reach a safe stopping point. You'll see a message that says, wrapping up.
That extra usage counts towards your weekly limit and it won't always be enough to finish the job. It's a soft landing, not a full refill. It's a small change, but it means fewer broken hands.
half -finished edits when your limit runs out. And the last update is Cloud Code Mods. It's the one I've been wanting to tell you about because it lets you reshape how Cloud Code looks and works.
So a mod is a small add -on that changes how Cloud Code looks or behaves. Just think of it like a browser extension, but for Cloud Code. Under the hood, a mod listens for things that happen, like Claude using a tool, asking for permission, or drawing the screen.
Then your mod can step in, it can change what you typed before Claude sees it. It can block or approve actions, or it can add something new to the screen, like a panel that shows you your build status. So Boris Cherny, who created Claude Code, put it like this.
Each person works differently, so there's no reason everyone should have the same Claude experience. Mods let you make Claude your own, and they can be really fun. One person made a mod that drops you into a real game of Doom while Claude is working.
When Claude finishes, it puts you right back. Another one lets you play Tetris right above your prompt. You type slash arcade, and the game pauses itself when Claude is done.
There are super cute ones too, like one shows a calm breathing animation while Claude works. Another puts the Claude mascot in a detective hat while it thinks. And there are of course useful ones too.
One shows a live panel of which helpers Claude is running right now so you can see the jobs as they happen. Teams can use mods for safety too, like asking for confirmation before anyone changes a live setting or keeping a log of everything Claude does. You don't need to write any code, just ask Claude Code to make you a mod, like make a mod that shows what Claude is working on.
Claude writes it, you approve it, and it turns on for that session. To find more mods, just type slash plugin or browse the cloud directory.
Just remember that mods have the same access to your computer as cloud code, so only install the ones that you trust. If you've ever wished cloud code worked a little differently, this is how you change it. So that was everything cloud launched this month.
Here's the quick recap. You got three new models, Opus for hard jobs, Sonnet for everyday work, and Haiku for fast and cheap ones. You got one cloud that figures out what you need and five new tools for docs, slides, designs, dashboards, and videos all built from a chat.
Claude also now works inside Google Docs, Sheets, and Slides and on CarPlay. And Claude Code got three major updates, projects, the wrap -up allowance, and mods. If you only try one thing this week, ask Claude to build you a deck from your notes.
It takes a minute and it shows you how much has changed. If you want help setting any of this up, I teach it inside the Claude Code Club. The link is in the description.
The Hook
The bait, then the rug-pull.
Ten launches in one month is too many to track. Duncan Rogoff boils them down to what each one is, why it helps, and whether it deserves your time.
Frameworks
Named ideas worth stealing.
01:36list
Which Claude 5.5 model to use
Sonnet: everyday work
Opus: long, hard jobs
Haiku: small repeated tasks
Start with Sonnet, reach for Opus when a job is long and hard, and save Haiku for small tasks run all day.
Steal forchoosing a model for any workflow
CTA Breakdown
How they asked for the click.
VERBAL ASK
13:10product
“If you want help setting any of this up, I teach it inside the Claude Code Club. The link is in the description.”
Soft pitch for a $9 Skool community, placed after the recap and also mentioned once near the two-minute mark.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A 12-minute tutorial that reverse-engineers a faceless YouTube channel earning $12K/month and rebuilds its entire production pipeline inside Claude Code.
A live demo of seven chained Claude Code skills that handle gap analysis, ideation, hooks, titles, thumbnails, repurposing, and performance tracking — for under $1 per cycle.
Five prompting habits that turn a default, generic AI promo video into one that actually matches a brand's own look, from finding reference videos to reviewing drafts like a designer.
00:13
00:25
00:32
00:49
00:55
01:15
01:23
01:32
01:45
01:55
02:05
02:15
02:25
02:46
02:57
03:06
03:21
03:25
03:36
03:56
04:06
04:21
04:26
04:33
04:46
04:53
05:03
05:20
05:27
05:51
06:07
06:13
06:27
06:37
06:43
06:57
07:14
07:22
07:37
07:47
07:57
08:08
08:28
08:38
08:48
08:57
09:18
09:28
09:39
09:48
09:58
10:04
10:30
10:38
10:45
10:56
11:21
11:29
11:36
11:49
11:59
12:09
12:19
12:34
12:40
12:59
13:09
13:19
12:41
12:27
11:24
12:27
05:39
13:05