Claude Whiteboards, Claude Sheets, and Claude Data
claude.com
claude.com
Build vs buy keeps getting pushed further to build, which is great for people in general but it feels like Anthropic is digging its own grave proving the only value it can provide is the model itself, and from what it looks like that's still scaling linearly and unprofitably.
History is repeating itself AI natively.
I don't see people switching to Claude sheets overnight & the messaging from anthropic will likely become less effusive over time (short-/medium-term). But I'd be surprised if this doesn't make significant inroads in the long-term.
I don’t see myself ever using an off-the-shelf analytics package again.
Figma is the default for teams but is losing a lot of stickiness.
I don't quite follow you. Are you talking about Mermaid's lack of decision when laying out diagrams?
I'm working on a tool that makes it easier for me to work with mermaid diagrams (mermaid is a great intermediary format for representing certain types of information, e.g. software architecure, but its syntax does not work well with my brain), so I'm interested in alternative concepts that make it easier to work with these types of things.
What these tools are generally missing is a way for the user to "fix stuff" that isnt just prompting. Once they add things like that, added to model improvements... it's just going to be a cost story: which one is cheaper.
OpenAI did something similar with Sales Assistant last year. Haven't heard of anyone using it.
Claude Design almost completely solves that. One dedicated designer can make a Claude Design System and then everything it makes from there stays on brand with essentially zero effort. And it is super intuitive to get to a usable Design System. Every time I use it I think "This is how AI is supposed to feel."
I don't think there is a reasonable expectation that these new tools will simply end Google Workspace. Google are going to be just fine to do whatever they want with their stuff, and they have the customers.
Anthropic can just do this for a new generation of people and businesses, and establish themselves here over time.
For me, this is simply a first-party UI for Claude. So they have better control over the presentation layer + some extra features on top.
It's not meant to be equivalent to Google Sheets or Excel.
Initially they were adopted by customers with the simplest use cases and the products grew over time to what they are.
The next set of startups might just have a Claude sub when they are small and grow with the product
So yes, not dead and clearly good for people who want that additional layer of control, but also nowhere where it used to be.
is this about growth, the userbase or the product? userbase is growing and the product seems to work fine; as for growth I'm not sure that was what people talked about when they said Figma would *die* but even then 43% seems fine?
I expect that Anthropic will have success in some of their side hustles if they're reinforced by the model ecosystem and double down on them, but you should also expect that at some point in the future we will talk about the "Anthropic Graveyard".
Whiteboarding is the main thing I do in my AI build processes outside of Claude - if they can do it native + multiplayer, it's going to be a huge UX improvement for a lot of users.
Why does it show that? Maybe they have tens of billions of dollars and the bandwidth to do more.
I read this as: We'll build anything and everything into Claude eventually.
How does this work long-term? Is the new reality going to be picking an AI provider for all of your SaSS tooling, similar to Cloud providers for infra? Hopefully competition thrives.
Dots in particular display some insidious methodology, they've hidden the "programmable" portions of an agent. They want to make it 'magic' (all your work goes through a single tool) but... that's pretty much exactly what businesses DON'T want. We need to see what it's doing, accessing, working with, underlying data, etc... It's an explicit business requirement for us.
The more they can 'hide' what the backend does, the harder it is to get away from the product.
Buuuuut, I don't think it's going to work. I can still export it somehow if I want to, and it would probably take 10-15 minutes... The whole point of these agents is that they're 'plug and play' with whatever technology you want.
And if they don't build them this way, someone will whip up an open source version that does it cheaper and more transparently in a couple of weeks.
Software being free now hits both of these companies exactly where they want protection.
this is my biggest gripe with dot. it doesn't integrate very well with my "agent ecosystem". It doesn't see my boxen. It doesn't load any agents.md. It like... is on an island and the only way it talks to the world is.... opaque magic.
Well, isn't that the natural end result of an AGI?
No more software, no more deciding what apps you need. It's all just Claude. Except now you're pulling 10kW to add up a column of numbers.
I think thats more like: "We'll build anything and everything into Claude eventually with the statistical median level of vibe-coded features."
That might work for some use cases but I doubt serious Excel nerds are going to drop it because Claude has a spreadsheet like tool.
I suspect the goal is partly starting a walled garden.
"Llms are interchangeable? Our harness/toolset isn't".
It's a way to get a moat I guess.
It doesn't sound like a business for the scale they're aiming for though. Taking a bite of the MS office's pie doesn't seem to me like a trillion dollar company move.
The rest of the world moves on using LLMs as tools.
Anthropic is eventually bought and stripped for parts.
“Please remove this page break I can’t figure out how to get rid of and someone else added”
Co-pilot: “Sorry I’m unable to do that but you can follows these same instructions you can find on the Internet to accomplish that”.
Claude: “I will literally rebuild Word from the ground up to solve this problem if that’s what it takes (but here is how to actually solve your problem), oh and do you want this exported in the best format for this software, this other software, or in these 15 file formats? Oh I don’t have native capabilities for that format - ok let’s build a script and convert….”
You get the picture. Competition is here, and it will get even better. Many SaaS companies will die. Good. Many of them only delivered short-term value solving artificial constrains or gaps in software.
I'm building https://smalldocs.org / https://github.com/espressoplease/smalldocs. My take is that the siloed approach of previous document formats is now redundant. The document formats were isolated because we needed heavy tools for humans to do the work. Now we need lightweight javascript predominately for rending as agents do the work, so document formats can be blended.
This means SmallDocs allows an agent to put together a document (the underlying format is markdown with rich custom rendering on top) with text, diagrams, a multi tab spreadsheet, charts, code, slides, etc.
Some example SmallDocs:
* AlphaFold 3: capabilities, evidence and operational limits (diagram, spreadsheets, charts, code, math): https://smalldocs.org/s/k9BtoviHg1eMsWzrSDBVZM#k=v0oIAbORXDj...
* Profitable growth can run out of cash (slides, spreadsheet, charts): https://smalldocs.org/s/nfCr0EEduLQwZAr0_v9ZR2#k=tv_BjigEYwt...
As a traditional company, if I had invested with a large team to build out these three tools. I wouldn't launch it with a short blog post. It shows how serious they are at maintaining them. People forget there are entire teams/companies supporting SaaS products. Not just the engineers that built them.
I've been using Claude Slides since it came out, and haven't touched PowerPoint (or any other presentation tools) since. It doesn't do everything PPT does, but it doesn't matter -- it makes turning an idea into a shareable presentation so fast and easy it's kind of crazy. PPT is effectively dead to me. And it was probably in my top 3 used apps just a couple years ago.
But the trade-off is a move from a bundled cloud service that is currently unlimited use into a new method of token-based content generation, which can be more expensive via API.
It appears to be a business model made directly from the 2000's with expensive mobile phone data charges and overages.
A typical workflow is: connect via MCP to some data source. Do some stats, transformations, etc. Put the data in a format that Claude understands or will understand better later. Make a UI/story out of it. Keep track of your "things" through visualizations.
Been using this worlflow extensively now. This is their easy wall garden.
And the story repeats. Google had a 20-year lead in search and LLM definitely has threatened it.
They absolutely are. Maybe not quickly (let's face it, Excel users are not sprightly), but this is absolutely going to displace Gsuite, and any Excel users who were thinking of moving to it.
I expect Google will respond.
either fully commit to disrupting your customers and keep a walled garden or stop building in the space, this incremental approach will only ship confusion
Yeah, if we're done working on it, why are we continuing to work on it? That makes no sense.
This project comes from a pure love of grids of numbers - and a deep sense that Excel has failed to keep pace with stuff like columnar data (Apache Arrow), fast queries (Rust Polars), proper number types, etc. so tbh don’t feel very threatened.
Feedback - good and bad very welcome.
Wouldn’t it make a lot more sense if they’d just let Claude connect to your Excel/Google Sheets/whatever?
Doesn’t even need an MCP or official integration at this point, if there’s computer access there’s access to everything.
And why share data when you can own it?
These seem like natural extensions that I'm sure will be very useful.
All Anthropic needs is to host an S3 clone and they become the default place to put all your company's data because they are the easiest place to analyze the data.
No company that need that level of data ingest and analysis is using Anthropic's tooling over something that is used industry wide.
And I'd argue they sort of abandoned MCP. Claude's support for it is poor.
Let's get Claude office, Claude auth, Claude notebooks.... maybe call it C Suite.
The better solution would be a uniform 'App Suite of the commons' (open-source) with a plugable/switchable data and inference providers.
Reality is, the crazy valuation given to frontier labs will push them to capture more of the value chain, to make it harder for customers to switch models. Having a best-in-class model is not defensible, long term.
ClaudeCloudServices
ClaudeERP
ClaudeCRM
ClaudeNeuralLink
ClaudeConsciousnessUpload
There's a reason my co still has Slack and it ain't because of its great features and bug-free app. It's because every written conversation we've ever had is in there.
Now it'll be harder to do the same for word and harder still for excel, but word is doable, and claude can build one off tools better than excel can, so I definitely think that MS should be wary.
The "I need some place to organize tabular data" use case is so different that I'm not considering it for the rest of this.
But for actual computation, if you throw it in Claude, it'll figure it out. It's much easier to do than to use Excel. Claude itself isn't limited to Excel; it can use pandas and the much, much improved plotting tools compared to Excel. I'm an OpenAI guy myself and used codecs to have it make me ggplot2 graphs, but then I forgot to specify it and it just used straight-up Python — I think it actually generated them as SVG files. As sad as I was about it, they looked better than the ones I had it make using ggplot2.
Regular Excel doesn't even compare there; not only can ggplot2 make many additional types of graphs, but SVG plotting, Python, and pandas can make literally any chart you want, including things like polar-coordinate charts.
Nobody is replacing excel anywhere.
You now have that app forever and nothing connects it to Claude other than the comment saying "this isn't just rounding — it's the single source of truth for the shared ledger boundary, not a formatting concern."
While I agree that spreadsheets are likely to be used less by regular folk, the idea that we'll become reliant on claude feels unlikely.
For good or bad, big companies run on Excel.
At my last job - a large tech company - if it wasn't in Excel, then engineers and HR wouldn't use it. The team had a tool that collected data from various databases for the end user and did some analytics. The code was in Python. The output was an Excel file.
I asked why they didn't use pandas.
"Because the users are just going to take the output and put it in Excel."
It was a very successful tool. No one needed to use it so we had to ensure users liked it. All due to the team lead's wisdom on how big companies operate.
They also wanted us to autogenerate PowerPoint files. We gave up on that. The spec is a mess.
I used to work as a trader for a major investment bank. Everything was excel. I can’t see it staying that way.
Normies know enough Excel to debug their "code". Will they be able to debug code AI produces?
> I used to work as a trader for a major investment bank. Everything was excel. I can’t see it staying that way.
How much of that Excel usage is part of the regular flow? As in several different teams using and sharing Excel files with a common template, so they can have macros that depend on the shape of data in the spreadsheet?
If the traders are mostly just using it for their own individual analyses, I can believe you. If they're sharing the Excel files across disparate departments ... it will be tougher.
A much simpler solution that a full blown replacement is simply having AI read, understand and edit Excel files. Why go with a completely different solution if the AI can solve all your problems within Excel?
I wonder how long ago this was, as now it's likely that Excel file will now be opened and interpreted by an LLM rather than Excel. The corporate world is changing fast.
They are shifting to building stuff for agents, not for humans.
Omarchy? yep.
For one, maintaining a lot of apps and connectors costs time and focus for a marginal return. Also it burns relationships with everyone building on top of your API if they can be disrupted anytime.
Their competition is offering good enough models for vastly better prices and they are making products that their customers don't need.
I feel like there will be a site similar to google graveyard for all these AI+X products that OAI and ANT are throwing out.