7 New Claude Features To Get You Ahead of 99% of People Using AI
A tour of seven recent Claude features, each with a setup path and one real use case, framed as a shift from talking to Claude to managing it.
Posted
3 days ago
Duration
Format
Listicle
educational
Views
41K
578 likes
57 · 43
Big Idea
The argument in one line.
Each of these seven Claude features moves the assistant from something you talk to toward something you manage, by fitting your tools, delegating work, and letting you verify results.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You use Claude daily and want a fast checklist of what shipped recently, with the exact setup step for each feature.
You live in Google Sheets, Docs or Slides and want to know whether the official Claude add-on is worth installing.
You build dashboards or tools for a team or customers and are weighing API cost against running on each user's Claude membership.
You run long Claude Code builds and are tired of being stuck at your desk while they finish.
SKIP IF…
You want a deep technical tutorial on any single feature, since each gets only two to four minutes.
You do not use Claude Code, Artifacts or Google Workspace, so most of the seven features will not apply.
TL;DR
The full version, fast.
Marchese covers seven Claude features that push the tool from conversation toward management. The official Google Workspace add-on puts Claude in Sheets, Docs and Slides with native formatting awareness. Mods let you reshape Claude Code by prompting, but only if they save real time. Dashboards let you trace any number to its source. Projects in Claude Code spawn parallel threads from one chat. Haiku 5.5 cuts API cost enough to rethink product economics. Artifact Docs and Slides keep drafting in one place, and Remote Control with keep-awake frees you from your desk.
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.
01 · Feature 1: Claude in Google Docs, Sheets and Slides
The official Workspace add-on, setup steps, and why native product awareness matters.
01:42 – 06:20
02 · Feature 2: Claude Mods
Customizing Claude Code by prompting, the fake-work warning, and three mods worth trying.
06:20 – 08:17
03 · Sponsor: Bird
Bird gives agents email, SMS, WhatsApp, calls and an eSIM.
08:17 – 10:37
04 · Feature 3: Claude Dashboards
Artifact dashboards that show where every number comes from.
10:37 – 12:34
05 · Feature 4: Claude Projects in Claude Code
One chat that spawns and checks parallel threads.
12:34 – 14:44
06 · Feature 5: Claude Haiku 5.5
Cheaper API cost changes what is worth building for customers.
14:44 – 16:38
07 · Feature 6: Docs and Slides in Claude Artifacts
Editable documents and decks, plus three fixes for AI-looking slides.
16:38 – 17:16
08 · Feature 7: Remote Control plus awake
Run local builds and watch them from your phone.
17:16 – 17:30
09 · From talking to managing
Recap and a pointer to picking the right model and effort level.
Atomic Insights
Lines worth screenshotting.
The value of the official Claude Google Workspace add-on is product awareness: native formatting, sheet formulas and themes, not just access to connectors and skills.
A mod passes one test: what did it actually help you get done? Anything else is fake work that feels productive.
Mods live only in the conversation that made them unless you ask Claude to save them as a plugin and install it.
A good mod to start with is a usage meter that shows token consumption and how close you are to your limits.
A correction-capture mod that offers to save every 'no' as a permanent rule turns repeated mistakes into system improvements.
Dashboards add a 'see where the numbers come from' view that shows raw values and the exact formula, so you can trust but verify.
Dashboards can now pull external data into artifacts, which earlier artifacts could not do.
Claude Projects in Claude Code replace juggling chat windows with one conversation that spawns and checks parallel threads.
Claude Code now partitions work into threads from a single chat by default, so you get the projects experience without opening Projects.
Haiku 5.5 at roughly 10 cents per million input tokens and 50 cents per million output tokens moves AI cost from dollars to cents for many tasks.
Running Opus-class models through the API is often too expensive to build a product around, which is where a cheap small model changes the math.
A practical rule: anytime you would handwrite something longer than an email, draft it as a Claude artifact document instead.
Slides that look like AI slop are fixed by a design system, one original idea before you open AI, and asking for five variations instead of one.
Artifact share links require the recipient to have a Claude account, a gotcha many people miss.
Remote Control plus a keep-awake setting lets you run local Claude builds and monitor them from your phone.
Takeaway
Manage Claude instead of just chatting with it
WHAT TO LEARN
Each recent Claude feature reduces context switching, makes results verifiable, or lowers cost, so you can delegate more and supervise instead of operate.
Install the official Google Workspace add-on to use your Claude skills and connectors inside Sheets, Docs and Slides with native formatting and formula awareness.
Judge every Claude Code mod by what it helped you finish. Start with a usage meter and a correction-capture mod, and save keepers as plugins.
Use the dashboard template and the 'see where the numbers come from' view to verify values and formulas before trusting a business number.
Let Projects in Claude Code coordinate parallel threads from one chat so you stop choosing which window a task belongs in.
Test Haiku 5.5 on your tasks. At about 10 cents per million input tokens it can make API-based products affordable that were not before.
Draft anything longer than an email as an artifact document, and fix generic slides with a design system, an original idea, and five variations.
Pair Remote Control with a keep-awake setting so long local builds keep running while you monitor from your phone.
Glossary
Terms worth knowing.
Mod
A small piece of code that hooks into Claude Code to change how it looks or behaves. You describe what you want and Claude writes the mod.
Artifact
A webpage, document or deck that Claude generates in a chat and that you can edit and share.
Dashboard (Claude)
An artifact template that visualizes data and shows the raw values and formulas behind each number.
Claude Project (Claude Code)
A single ongoing conversation in Claude Code where Claude coordinates parallel threads, distinct from Projects in regular Claude chat.
Remote Control
A slash command that lets you use chats running on your local machine from your phone.
Haiku 5.5
Anthropic's small, fast, low-cost model, described in the video as costing about 75 percent less to run than Haiku 4.5.
“The most dangerous way to lose time is not to spend it having fun, but to spend it doing fake work.”
A quoted line that reframes productivity tooling in one sentence.→ newsletter pull-quote↗ Tweet quote
16:40
“Until now, Claude has mostly been something you talk to, but every one of these features pushes it towards something you manage.”
States the thesis of the whole video in one line.→ TikTok hook↗ 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
Feature number one is Claude's official Google Workspace integration. Before now, the only way to use Claude in Google Sheets, Slides, or Docs was without any first -party plugin. But in October, Anthropic released an official native Google Workspace integration.
And we'll dive into what this means in a second. But first, here's how to actually set it up. Go to Google Workspace Marketplace, search for Claude, install it, and you can install it at the admin or just for yourself.
Then in Docs, Sheets, or Slides, go to Extension, select Add -ons, get add -ons, and install Claude again. Now Claude will open up in the sidebar next to your file and you can change the model, use your connectors and your Claude skills. And it truly feels like Claude directly inside Google Sheets and these Google workspaces.
And what really made this click for me is the access to your skills and connectors. So for example, if I have a skill called review writing, I can use this directly in docs. And if I've connected Slack and want to summarize my week for an end of week update, I can do that right there too, because it has access to that information.
And if you'd rather not, work in Docs or Sheets and instead directly in Claude, you can turn on Google Connectors and work that way instead. So it goes directly to these without ever leaving your Claude terminal.
But honestly, for all of this, it's less about the connectors and the skills, but more about the product awareness. Before, Claude would try and piece together information about your file. But now, because it's an official integration, Claude has access to information like any native formatting, sheet formulas, or themes.
And a few of my friends that live in spreadsheets, they were texting back and forth. They said, literally, it's a joke. It's unbelievable, probably the best use case.
Now that is just the first of seven new Claude features that will get you ahead of 99 % of people. And the features that I go through point to a bigger trend. Claude is starting to fit in the way you already work instead of making you adapt the way you work, which brings us to feature number two, Claude Mods.
So this feature recently went viral, but in my eyes, nobody is talking about what's actually interesting about it. Now, if you're not familiar, Mods lets you customize what the input bar looks like when you ask Claude a question. How it works is there's essentially small pieces of code that hook into what's happening inside Claude code.
And the key part of this is you don't write anything. You describe what you want and then Claude builds the mods itself. The creator of Claude code, Boris Cherny, put it this way.
Mods are absolutely insane. You can now customize Claude to work and look the way you want by just prompting it. And he is right.
It is cool. But I can't help but think about the Obama meme where he's just like giving himself the metal like, yo, this feature is sick, but it's him who made it. But it is genuinely cool.
And people have built everything from a mod that checks the blast radius of any change they you're making to a mod that lets you play an online Doom game with other Claude users while you wait for Claude to respond between prompts. But a massive word of caution before we get into what mods you should build.
When I think about mods, I think about this line from Paul Graham. The most dangerous way to lose time is not to spend it having fun, but to spend it doing fake work. And mods are a perfect example of this, where it feels productive, you're making this cool custom thing, but is it actually helping?
So the test for any mod is simple. what did it actually help you get done and before you go crazy with this feature i think a lot of this is marketing and clickbait but there are a couple of ones that are genuinely helpful so start with a usage mod this is where you can see where your token consumption has gone and how close you are to your limits there's an open source one called burn meter that i'll link below but the whole point of mods is that you can actually make these yourself so instead of doing that just use this prompt and make your own version of it so you can go through the motion and see what's possible that will get you started with a valuable use case and then you can just build on from there.
Another one that I like is called reflect and how this works is when you're going back and forth with Claude, when you tell it no or you correct it, a band will pop up offering to save that correction as a permanent rule. This is a way to remind you to stop repeating yourself and improve your system continuously. And if neither of those sound exciting, you can lean on Claude to identify ways to create your own mods by using this prompt to suggest mods based on your session history.
Now when you do create a mod that you like, ask Claude to save it as a plugin and install it. because by default, these mods are just in that specific conversation unless you save it long -term. Now, those specific mods are great, but this is what I'm most interested about.
It shows a trend of where the world is going, and it's more than any one mod. It's about how everyone can now customize their own software. Let's just rewind for a second to when the internet first started taking off.
So people watching this channel, who remembers using MySpace? Let me know in the comments about that. But on MySpace, you could go in and customize your homepage so it felt like yours.
You could change. the color, the theme, the song playing. They also had a feature that were identify your top friends, which quite literally ended friendships.
You had the power to ruin a friendship and personalize the experience. It was truly great. And people love this.
And that's a form of personalized software. And we're going to see more and more of it because AI has made it so easy to write code. So hyper -personalization is a part of the next frontier.
And these smartest businesses are taking this very seriously. But the more that you customize and automate, the harder it can be to keep track of what's actually happening. And that's why this next feature matters that nobody is talking about for whatever reason, but it is the exact way that you can build and scale automations while still trusting Claude.
But before we get to feature number three, there are just some things that Claude can't do, which brings us to today's video. Bird just raised $450 million, led by JP Morgan, my former employer, to solve a simple but glaring problem that exists in the AI world. Everyone is building agents right now that can research, plan, and write.
But the second the job is, now go text the customer or call them back, the agent stops and hands that work to you. Bird's CEO, Robert Viz, put it perfectly. An agent that can't send the email, fire the text, or pick up the phone never actually achieves anything.
So Bird solves this with their agentic harness, which enables agents to email use WhatsApp and make calls and even get their own eSIM. They're creating the bridge that's connecting agents to humans.
So for phone calls, like I said, Bird lets your agent set up its own eSIM and it has an actual number to call from. For texts, Bird enables you to send SMS, WhatsApp and iMessage. And for email, it's as easy as going into their portal, hitting copy setup prompt and sending it to Claude.
In today's world, it's truthfully easy to rebuild a lot of software. But when integrating with global laws and regulations, I personally don't want to touch that within a thousand foot pole. And with Bird, already sending trillions of messages a year across 150 plus countries.
I'll personally pass on trying to recreate this software. And because of its AI native approach, integrating Bird is just one prompt away. So if you're looking to create a communication layer in any product you're building, go click the link in the description to get started today.
All right, feature number three is Claude dashboards. I've made entire videos on my channel about Claude artifacts and dashboards is a key enhancement to them. In short, an artifact is a way to use Claude to launch a simple webpage that you can share.
But dashboards, takes the functionality a bit further. So what dashboards are is they're able to visualize data, but on the back end, show you exactly how it got that information.
Now, what does this actually mean? Let's just show you in practice what it looks like because that's when it clicked for me. So to use it, go to artifacts, pick the dashboard template, and then say something like, create a dashboard that would be the most useful to me based on this project.
It's a perfect starting part. I don't want you to think about how you can use it. Just go use it and you can judge for yourself.
Now that creates a visualized dashboard. And you might be thinking, Austin, that's cool. You could have just done that since the beginning of AI.
Who cares? Here's the most interesting thing about this. On the right, next to the number, there's an option that says, see where the numbers come from.
When I click that, it shows me the raw values and the exact formula for how we got to that information. Now, I've spoken a lot about AI anxiety on this channel. And a key reason for this happening is because you don't really know how your AI system works and if the information is truly correct.
And as you build more complex systems, how well do you actually... actually understand them? And can you have confidence in the actual numbers?
So there's a saying, trust but verify. Before this feature, with these Claude overview pages that I would make myself, you had to kind of just trust. But now, Claude has built a first -party way for you to then verify.
This two -part approach to creating dashboards and analytic tools that update periodically based on your information is a major upgrade to the Artifact Suite. And previously, you weren't able to pull external data into artifacts, and now you can. So this combination of external data and being able to verify how the data is actually being visualized, it is an absolute game changer for any real dashboard that impacts a business.
Now, the natural next question is, as you build these dashboards and these analytics tools, how do you actually get more work done without managing every conversation yourself? And that's where feature number four comes in, clawed projects. And when I say this feature, I do have to be specific about which projects because Anthropic used the same names twice.
This is not a project that's in regular Clawed, Chad. This is a new Clawed project that sits inside Clawed. code and i touched on this feature a bit in a recent video but in short here is the problem it solves let's say you're working on a bunch of different things in cloud and you have multiple chat windows running you're going to often say to yourself all right which chat should i put this in or you have to constantly switch between them and this just creates tension and it's a bottleneck for you actually building a project is one ongoing conversation and claude acts as the coordinator so it'll split the work into threads and run at the same time check them and tells you which need your input simply put it you to have one single chat and it'll spawn off new threads based on the conversation that you're having.
it'll abstract this coordination layer that you're so used to doing. And this will take you a layer above. So you're no longer in that coordination layer.
And this is a learning curve. And likely over the past two, three years, you've just been changing chat windows. And that's just a pattern that you're so used to.
This goes against that, but it brings it all into a single window, reducing your context switching. At first people think of this as a technical feature because it's in Claude code, but you know, on this channel, we don't call it Claude code. We call it Claude power user.
And this is a perfect power user feature. And this feature has been so powerful that you actually don't even need to go. to projects to have the same experience.
Now, when you're using Clawed Code, it'll start automatically partitioning chats from a single thread by default. So it's essentially forcing you to use this project's feature whenever you're using Clawed Code, and I absolutely love it. One thing I'm excited to continue to see is how these tools empower multi -threading as we start managing more and more agents.
Now, before we get to the last three features, if this is your first video of mine, welcome to the channel. But if this is your second or more, you know the drill. Here is our anti -slop agreement.
The visuals, the test. the hours of research that went in this video. This is entirely built for humans, not for AI robots.
So all that I ask as part of this agreement is that you can subscribe to this channel so it can help reach more people. And if you already subscribe, you're a legend, but make sure to sign up for post notifications and like the video. And as a thank you for all the support you guys have given me, we're almost at 100K subs, which is wild.
I give away a Claude Max subscription every single video. And so this video's winner is Benjamin Wager, MD, for building an education platform. And to enter the next video's giveaway, comment below.
with what you're building or any problems that you're running into, I would love to help as many of you as possible. Okay, enough of that. Let's get to feature number five, which is Claude Haiku 5 .5.
In this video, we've talked about giving Claude more work to do, but there's another factor that determines how much you can actually delegate, and that's how much it costs you to run. And that's why Haiku 5 .5 is worth paying attention to. Now, I'm not going to turn this video into a models video, but here's the one thing to know.
Anthropic calls this the cheapest, fastest, and most capable small model that they've ever released. And it says it costs around $75. percent less to run than haiku 4 .5 and this trend of decreasing cost while increasing the quality is something that's really changing what's possible to build one thing i've talked about a lot on this channel is building local dashboards that leverage your cloud membership so you don't have to pay api fees now that's fine for personal use but if you're building for customers or a team you have to use the api and the reality is that running opus or fable through the api can just be too expensive to build a product around but now with haiku 5 .5 this starts to change the math for a lot of people that are watching this video.
At roughly 10 cents per million input tokens and 50 cents per million output tokens, it's starting to get into a cost of cents, not dollars. So for me, when we're scaling buildpartner .ai, I'm starting to ask a different question. I'm wondering, do we still need AI to run on each user's machine through their own cloud membership?
Is that still needed? Or could we build it directly into the dashboard now? The models are finally getting smart enough, fast enough, and cheap enough to start making the economic actually work.
So here's what to do. Experiment with Haiku and see how it feels on your task in speed and response time. And then ask yourself, if you're building a product or have an existing tool, can you swap whatever model you're using right now for Haiku?
And then did the lower API cost open up the use case that wasn't worth building before? And so on -screen is a prompt to help you optimize your setup with Haiku. Cheaper models make it easier to justify using AI for more tasks.
But there is another part of the workflow that matters. You don't want to generate something useful and then spend another 10 minutes, moving it into another app that you actually collaborate on.
That's where feature number six comes in, Docs and Slides in Cloud Artifacts. One of my biggest workflow improvements with Cloud has been being able to go from a prompt to an editable document or presentation in the same place. With Artifact Docs, you can ask for a document in any chat and it'll just open it right away.
Then you can edit it inline like any document editor, and then you can highlight text and then have Cloud answer a question or make a change for you directly in the document itself. It's a Cloud -native approach. to editing a large document, which has really changed my workflow.
And this is something that I just now use every single week as part of my YouTube creation process. And my general rule of thumb now is anytime you handwrite something longer than an email, create a cloud artifact document, marketing material, a LinkedIn post, a proposal, whatever you're working on. If it is a longer document, bring it to cloud artifact documents.
Now, before we get to artifact slides, I did go deeper on this in a video on my channel, which I will link in the description. With slides, it's the same idea, but you're creating a slide deck. So it's an artifact that's designed to look like a PowerPoint, but if you want to export it to PowerPoint, PDF, Google Slides, that's totally an option.
Now, the biggest complaint that most people have when using this is that they all look AI generated or AI slop. So there are three good ways to fix this. The first is you need to create a design system.
Bring in any reference assets, your brand files, any design systems you currently have, and bring that directly into Cloud. This will create a design system that will inform exactly how to create the slides that you create whenever you build them. And if you don't have a brand look yet, design DesignMD .app is a great place to pull design inspiration from.
This is just a general thing where once you create a design system, I would spend a couple of hours on this. It'll start making every asset that Claude creates in your brand. The second is for the content it creates, use the one thought rule.
Craig Brax recommends you come up with at least one original idea before you open AI. Then use it to refine, challenge, and expand on that idea. So create a deck that starts from your idea, helps it lean away from the default deck look.
And then the third is don't ask. Ask for one version, ask for five. This will let AI iterate on five different variations, and then you choose the one that you like the most, and then it'll start iterating from there.
And when you do land on a deck that you like, one gotcha that a lot of people miss is the sharing. If you send someone a link, they need a cloud account to open it. Just keep that in mind.
Now getting to feature number seven, remote control plus awake. This is one that I've personally fallen in love with over the past couple of weeks, and I think most people don't use it enough. I am a massive believer in running things locally on your community.
instead of on the cloud. Primary reason is that Claude can then edit your files. It just makes it a simpler configuration.
It feels more like having AI working with you on your computer than a robot working remotely. And it also keeps you in the loop and you stay more in tune with the actual actions. But the problem with running everything local is you need your computer on.
And if it's not up and running, it's going to be hard to keep working. And then also it can be hard to work from your phone. So here's a combination I use to maximize using my local machine while not being a prisoner.
my laptop. The first is remote control. To use this just type slash remote control and this will let you use whatever chats are going on in your local machine directly on your mobile phone.
But once you do that you need to make sure that your laptop or your computer doesn't fall asleep. So the second step is you need your computer to stay awake. So if you're using Claude desktop app this is actually built in.
Go to settings then Claude code and then turn on keep computer awake while Claude works. And if you're on a laptop there's a separate switch called keep awake on battery power. So you may want to turn that on as well.
And if you'd rather just manually remind your computer to stay awake and this is something that i do i use a command called awake in my terminal which you can have claude set up by using this prompt essentially now whenever i type awake it'll keep my computer up for however long you set it i usually set it to three hours so by doing these two things remote control and keeping my computer awake it's a massive quality of life improvement i no longer feel tied to my computer so what i'll usually do is after a long planning session i'll get to a point where it's time to build and those builds for claude can take up to 30 or 40 minutes so what i'll do is I'll head to the gym or I'll go into my sauna, which fun fact.
This is actually a sauna in my office, and I'll keep an eye on my phone as it's iterating and building. Now, all of a sudden, I don't need to be in front of my laptop to keep being productive. Zooming out for a second, here's how I think about all seven features that I covered in this video.
Until now, Claude has mostly been something you talk to, but every one of these features pushes it towards something you manage. You can now work where your team works, right? Use Google integrations, docs, and slides to work in the tools you already like.
You can design your work environment. You can use mods to adjust Claude to your work. You can delegate.
You can use projects, parallel threads, and haiku to hand off tasks and split up the workload. You can check the work. You can use dashboards to trace any number back to its source.
And five, you can stay connected. You can use remote control to keep things moving without being chained to your desk. But like any manager using these tools, you will have a budget.
And the biggest way to save on budget is picking the right model and effort level for the job. So to make sure you know how to properly do that, something that most people don't actually understand, go check out out this video if you like this i guarantee you'll love that i'll see you over there peace
The Hook
The bait, then the rug-pull.
Seven Claude features in under eighteen minutes, each with a setup step and a use case. The thread running through them is a shift from chatting with Claude to managing it like a team.
Frameworks
Named ideas worth stealing.
16:38model
Talk to, then manage
Work where your team works
Design your environment
Delegate
Check the work
Stay connected
Five verbs that recap how the seven features shift Claude from a chat partner to something you manage, within a budget set by model and effort choices.
15:34list
Three fixes for AI-looking slides
Create a design system from your brand assets
Use the one-thought rule: start from an original idea
Ask for five versions, not one
A short checklist for getting artifact slide decks that do not look generic.
CTA Breakdown
How they asked for the click.
VERBAL ASK
17:16next-video
“Go check out this video, if you like this I guarantee you'll love that.”
Ends on a pointer to a related video about picking the right model and effort level, tied to the budget point in the recap.
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.
Eight free plugins and habits that cut Claude Code's token bill, kill chat-planning chaos, and replace typing with talking, none of which depend on picking a smarter model.
A rapid-fire tour of seventeen ways to use Claude's design tool for real work, from rebuilding old branding to wireframing a product before writing any code.
Four ways to use Claude artifacts beyond a pretty preview: plan visually, collaborate in-line, ship micro tools, and redesign how work gets handed off.
A walkthrough of the 13 skills Anthropic added to the Claude for Small Business plugin, and the orchestration-vs-utility framework for deciding which ones actually matter.
A JP Morgan engineer turned $25M startup COO lays out the seven pieces of Claude he'd learn again, in the exact order that compounds instead of the order most people stumble into.
00:15
00:32
00:39
00:59
01:11
01:28
01:34
02:05
02:21
02:31
02:38
02:57
03:07
03:23
03:36
03:47
04:03
04:13
04:24
04:39
04:59
05:03
05:23
05:30
05:47
05:54
06:12
06:46
06:58
07:06
07:18
07:32
07:45
07:59
08:20
08:51
09:07
09:18
09:30
09:44
09:57
10:10
10:17
10:54
11:03
11:15
11:29
11:42
11:55
12:08
12:16
12:44
13:01
13:14
13:23
13:34
13:54
14:07
14:19
14:32
14:59
15:12
15:25
15:52
15:58
16:12
16:42
17:16
17:19
14:50
17:42
14:05
15:46
Video of the Day15:55
21:11