An open-source, clean-room reimplementation of Microsoft Word in pure Rust | Hacker News Reader

An open-source, clean-room reimplementation of Microsoft Word in pure Rust

github.com

101 pointsespeed119 comments

Has anyone of the people spamming these apps by storytold on HN actually tried any of them? I know that any new software can have bugs, but the ones here are really not acceptable for a release, let alone labeling it a 'reimplementation'. The quality of the software is just very embarrassing to anyone who might have used any of the originals, or at least can assess basic software qualities. I have tried some of them and really everything is buggy, a lot of options and buttons make the whole thing crash, the UX is very poor compared to the originals. I tried some of the clones of Adobe's apps, and they just don't work. I tried wordcraft as of writing this, it does not work and it is even more buggy than my experience with photocraft, changing a text to a headline made it just crash on my M4 MacBook, after making it VERY slow and unresponsive for 3 minutes.

It is extremely far from a working solution for someone who uses these apps for just casual non-professional use. For professional use, I borrow Wolfgang Pauli's "not even wrong" to say it is not even something to think of. People who think this is a possible replacement just don't have a clue about how big a software like Photoshop and MS Word are, and how much they actually provide their users with.

I know that "Hey I open-source'd Adobe and Microsoft" can get people some hype on TikTok and Twitter, and maybe have some media reporting about them for fun, but I am really surprised that people here are passing on these like they are really what they claim to be. Genuinely, if anything can be used as a counterargument for the promised LLMs revolution, it would be software like this.

I see what you're saying, but we're just getting to the start of it being worthwhile to do something like this. Once these projects have people testing them and proposing changes/improvements (AI driven), I don't think it would be long until they are on par with, if not better than the originals.

The only catch will be if attention coalesces around an individual project instead of hundreds of people vibecoding their own versions.

That's fine, but why not make it clearer that they're extremely early alpha?

Their websites look more polished than the apps themselves, have download options for non-technical users, and in no place that I can see (as of writing) do they mention that they are early development: https://getartcraft.com/apps/wordcraft

In "the before times" it was at least the norm to place at the top of a github repo and download section "this is early alpha" or similar to not let down potential users. This project currently seems like a great way to burn user willingness to try new apps.

“I don’t think it would be long” is an age-old planning fallacy in this field, AI hasn’t changed that. It is certainly easy to clone a UI today because an agent has a target to iterate against, but cloning all the behaviours of the software in question is going to prove to be much more difficult.

These are decades-old software where even the inconsistencies, quirks and bugs are now considered features (because users have come to rely upon them) - a full detailed specification itself would be very complicated. My bet is this OSS-clone hype will last all of 6 months.

Codebase of the behemoths should be around in the GB-range, Millions of engineering hours and QA etc.; I do not see how one person even with 100 agents will be able to coordinate/navigate such a large project.

My bet is: This guy is pumping out that stuff to land a good job somewhere :-) (I doubt that by now someone is paying for the continuously crashing clones?)

Like everything in this stupid, hype-driven AI industry, it's always the "easy" 80% that is done; the actually-difficult final 20% is always waved off with vague future promises.

So AI has made everyone managers. Got it.

My personal hell at work is my PM vibe coding some prototype, showing it to my director or vp to get them excited, and then coming to me to say that it is basically done so I should be able to deliver in a couple weeks.

This problem predates AI by decades.

yeah, but before they had to make their hands dirty to deliver at least something >made by them<, even if its some basic mockups on some slides to show an idea

now, they open discussions with "you cant imagine how many tokens I burnt yesterday to create X" and then showing a widget with 5 buttons that runs within Claude/Browser without any real data or connection to the rest of any parts of the company :-D

Never before has my PM been able to make a prototype in a day or two.

Yes, it was always possible for a PM to get a VP excited about a project and then lead to some mandate. But at least then I got to do a time estimate and they believed me. Today they can create a prototype to show to the VP so that both of them conclude that the project isn't actually very hard and my estimate of two engineers and four months gets rejected because "the prototype is 80% of the way there and it only took two days" even though the prototype is running locally on my PM's machine and is using fake data when we don't have easy access to the real data.

But this is not even the easy 80%. More like a skin-deep “reimplementation”.

People who don't put in effort to build a thing also don't put in effort to test a thing. The effort creates investment.

These apps have to be astroturfed, right? Like I get the desire to give Microsoft and Adobe the middle finger and I guess it's somewhat impressive these apps even do anything, but man are they such buggy pieces of trash. Like you said, I downloaded the Word clone and it's broken in so many ways that I can't even describe what the bugs are. In 5 minutes of messing around, I hit so many random and inconsistent failures of basic functionality like styles, image resizing, and table formatting. Plus the UI is atrocious. All the downsides of Windows apps mixed with all the downsides of a weird cross-platform UI framework.

Agreed, every behavior of this particular dev supports the hypothesis that they are producing mass slop to a "good enough" level to receive massive hype, without the intention to put in the thousands of GPU-hours it would require to polish (actually, the estimated roadmap is growing rapidly in expected timeframe as the AI understands the difficulty of the task). I expect there to be more and more of these crappy surface-level implementations before people lose interest and the dev drops all of it.

Yes. The word one is excellent for doing what I keep using Google Docs for. Just get a nice document going -> print -> pdf. Very easy. Works very well for wordcraft and deckcraft.

And the pdfCraft is so much better than every pdf site on the internet.

Also Gridcraft at least equals Google Sheets for me in functionality. Too bad copy paste from Excel or Google Sheets doesn't work correctly. It does work, which is already impressive.

This is probably not heavy use, I get it, but definitely improves things for me. Also rapid improvements are incoming.

What's the advantage over mature open source products like LibreOffice?

This works at least as well on the web as Google docs. Obviously that's without the collaboration features, as well as using your local filesystem instead of Google Drive/Azure Cloud storage.

But Google Drive was always the worst part of Google Docs by far (imho).

And collaboration ... for my personal office needs ... I sometimes try to do this with my wife and kids, but almost never.

Pdfcraft is far superior to all those stupid pdf sites, and Wordcraft easily does what I use Google Docs for.

Call out: Is pdfCraft better than this site:

https://www.ilovepdf.com/

Esp. in terms of quality regarding document output

For Word-to-PDF specifically, I'd compare them with an Indian-language document, merged table cells and pictures. Those were the cases that broke in my own converter when I tried drawing the pages with pdf-lib.

Rendering the .docx with docx-preview and using the browser's print engine preserved those cases in my tests. The cost was an extra click in the print dialog. That doesn't establish how pdfCraft compares with iLovePDF, but I'd want those documents in any output-quality comparison.

Photocraft is near parity for day to day use. I stopped using gimp and photopea and completely migrate to photocraft. Granted I am not poweruser with tons of plugins. Simple editing and layering, masks, transparency, gif-jpg-png concersions are very doable with this. So yes, it is good enough for production use for majority of not-so-poweruser.

Storytold apps are a gift, and are setting the precedent for AI apps. Everyone was wondering about the relevance of open source in post AI era, and here we have it: a magnitude more ambitious ecosystem.

While Bun proved at-scale rewrite, these are proving huge at-scale re-dos of the entire software ecosystem.

I suspect 24 months from now the landscape of daily software is going to be vastly changed.

And they are already really useful, even in an extremely buggy alpha state. LightCraft is already working pretty well for a photography hobbyist like me. Can't wait to see the state of this software suite in 6 months.

> extremely buggy alpha state

I'm curious about this part. As a user you're finding them extremely buggy - what kind of bugs are you seeing?

It sounds like you're expecting the bugs to be ironed out in six months. I guess I'm wondering if the same process that led to buggy software in the first place can actually be used to get rid of the bugs without creating more.

In the issues page of the repos you can see avg type of open source bugs. For the Word clone there is: spellcheck doesn't work, image embedding is broken, etc.

Well, in the case of LightCraft, some basic things like processing raw photos. It seems to be struggling with processing them compared to older software, it's kinda slow and glitchy. But it's in a really early stage, support for the CR3 format (the one I need) was just added like 2 days ago.

I don’t think it’ll be “the same process”. Someone has to identify the bugs and instruct the LLM to fix them. I doubt that can be automated with GUI software as it stands today - there’ll be a lot of human involvement even if the LLM implements the fixes.

With a human in the loop, iterating on bugfixes can work. It’s just more human-labor-intensive than an initial implementation.

I suppose you could invite users to submit bug reports, give them some guidelines on how to write them, and then have an LLM monitor the bug reports and fix them. But then I think you’d be more likely to see the creation of more bugs, as you say.

Why not use Darktable?

Performance is bad and UI/UX is laughable.

Why not use ART (RawTherapee fork)?

See my previous comment.

As a Lightroom user, I've tried _many_ times to use Darktable, and I've given up every time.

Like so many open source GUI projects, it's clear they have solid technical developers who are building good underlying tech, but they have no one on the project doing any kind of user research or spending time thinking about what good UI/UX looks like.

I do use it sometimes! But I dislike the UX. RawTherapee feels better, but it still feels weird compared to Lightroom. LightCraft is basically a Lightroom clone, so everything feels familiar and standard. The problem is that a bunch of stuff is still missing and/or buggy.

Let's add it to the list of predictions then. Wonder what people were predicting 2 years ago

Businesses that previously relied upon vendors are going to start hiring engineers to build their bespoke stack in-house.

This is software's radiology moment.

What happened in radiology?

No they won't

Why would they hire engineers for this?

Note that they're currently a buggy mess in my experience. That's not unusual for an alpha, but I'll personally wait to call them a gift until I see if the team is capable of polish+reliability. I do think their marketing is ahead of their product at the moment.

> bun proved at-scale rewrite

Not so sure about that one

We already have one entire software ecosystem, why do we need another?

Is there any evidence that LLM development practices result in code that is more reliable or easier to maintain?

And is software truly open source if the preferred way to maintain it is via proprietary commercial services whose behavior is subject to frequent change without notice?

>daily software is going to be vastly changed

Is this true for large & mega corps as well with deep enterprise integration? Thats where the money is for most of them.

These apps are awesome, but they make me a little nervous. I can just imagine when a "clean room Rust" Oracle DB (re)implementation hits GitHub that Oracle would not just sit there and take it but would instead rain lawyers down from the sky... Ditto for other companies that have hornet's nest legal departments.

That doesn't make me nervous at all. I've been waiting for the legal challenges from deep pockets to roll in.

I'm not entirely sure they will, though.

That famous Oracle DB? https://news.ycombinator.com/item?id=18442941

Good luck.

On the other hand, it might be doable for Oracle themselves, what with all the tests they have to test the AI implementation against?

In the context of Orcale and the like, I have a feeling sales would present bigger obstacles than legal.

Jokes about training an LLM to golf with executives aside, how do you establish credibility for a vibe-coded reimplementation of Oracle sufficient to convince even one major Oracle customer to bet their business on it, at any price?

I don't get why this is so impressive? You have libreoffice as well, but this project, like that one, will find that maintenance and resolving user issues and providing cloud features are the hardest and most valuable parts of document editing software. Not just having something that works like other open source document editors that already exist.

Also look through the issues page and you see issues like spell check doesn't work and image embedding doesn't work.

> and providing cloud features

I prefer to not have those in my text editor

Maybe I'm a cynical old man, but all these ArtCraft apps feel a bit too good to be true?

They come out of nowhere (sort of), and they're just a bit long-con trojan horse for some point in the future where the userbase will have to pay the piper.

I hope I'm wrong!

Fork and roll your own.

That's not my point at all.

If these apps eventually have millions of users, most of whom will just be downloading the binary releases and using auto-update to get the latest version through some opaque updater, what's to stop a future update from being nefarious?

Adobe customers are asking that exact same question

This is how software outside gatekept marketplaces always worked.

I would argue it happens even inside marketplaces. We've seen the authors of popular Chrome extensions take payouts and sell their extensions to new owners who have bad intentions for the userbase.

Isn't that the risk of literally any software?

Yes, but there used to be a gatekeeping effect because the development of huge software packages like LibreOffice was too capital-intensive to be enticing for a criminal enterprise. If you have malicious intent and know your way around software, finding a security defect and exploiting it for the umpteenth ransomware campaign is much easier than building a full office suite from scratch and getting a userbase going in order to sneak malware in years down the line. The implied fear is that coding chatbots may make this attack economically viable.

Certainly but often there's a company or well-known entity behind popular software, so that brings some accountability to the table.

In this case, these ArtCraft apps are one person I think?

Why would I want to maintain my own fork of a productivity application for personal use?

Agreed. Especially given that I downloaded one app and got a generic launcher instead that then allowed me to download a specific app.

While they may be a buggy mess, I think it's really interesting that we're starting to see this unfold in real time. People have been forecasting these kind of changes for the last few years, and it seems like it's actually happening now. Really curious to see where this goes, if everyone has their own custom fork of some common codebase for specific features/needs they have across these base applications.

All the apps claim to run in the browser via WASM but no links are provided. Are there any live demos of these hosted somewhere?

So were just done with the original definition of "clean-room implementation" right?

Cause you cant use AI that you cant guarantee wasnt trained on this implementation, and call it a clean-room implementation. At least not if you are going to be honest, but I guess that makes sense given how much these AI companies and their leaders lie.

Is Microsoft selling the code of its proprietary products for AI trainings? Definitely news to me

Can you tell me what the definition of "clean-room implementation" is?

And after that, can you tell me if you can call something a "clean-room implantation" without disclosing 100% of the training data used for the AI model making this supposedly "clean-room implementation"?

Unless you have a reasonable argument why a random AI would be trained on Microsoft's proprietary code, I don't see what you're basing any of this on. If you need the definition of clean room, go to Wikipedia.

I don’t know about office, but there have been several major leaks of the Windows source code. It seems pretty reasonable to think that that leaked code has made it into Claude or Codex in some form.

How would the source for Microsoft Word be in the training data for (for example) Claude?

How did those mathematician's data probably wind up in OpenAI's training data? They are willing to rip off Disney (one of the most litigious IP companies on the planet). There's no reason they wouldn't be willing to rip off the MS Word files uploaded by MS devs (and even less reason they won't steal your work too).

I really think the AI companies think there's no turning back on the infringement and they have to infringe in a "too big to fail" manner to the point where they can force an exception for AI not caring about the fallout for IP rights.

I think the idea here is that all these reverse-engineering projects are _saying_ they're doing clean room implementations for quasi-legal protection.

So they are claiming to follow the legally protected "decompile -> write spec -> new code from spec only" process but they're likely skipping the "write spec" step, and hoping/assuming that it's basically impossible to prove that they're doing so.

So like I said, they are lying.

> hoping/assuming that it's basically impossible to prove that they're doing so.

That would quickly fall apart in a courtroom, when they can’t produce the spec or show any evidence of development being done from the spec.

In any case, with LLMs I don’t see why that step would need to be skipped.

Sure but the idea here is that so many people are doing it, it will be a game of whack-a-mole for rightsholders to go after every single person who is trying this, or forking repos, etc..

And if you get a letter from lawyers, then you tell your agents to write the spec at that point, and you produce the spec..

I'm not saying this is correct or legal, but I've noticed that everyone doing LLM-powered RE these days is stating "clean-room implementation" and my gut tells me that most of them are not actually doing it.

Laws are unlikely to stop indivduals from building vibe-coded clones of commercial software, but they may stop above-board communities from forming around them, and will almost certainly prevent commercial sponsorship of their development.

They may also lead commercial AI companies to severely restrict the capabilities of publically available models, especially in terms of reverse-engineering, which is a cost that could be far greater than the value provided by vibe-cloned software.

Given that nearly every commercial software license agreement prohibits decompilation, can you even legally carry out the first step without residing in a jusrisdiction that doesn't respect international copyright conventions?

And isn't importing material produced in violation of copyright into countries that do respect these conventions prohibited by law?

I have no doubt that every commpany that has a "prevent decompilation" term in their license is terrified of that ever being tested in court.

I do wonder if some company with money to burn and their backs against the wall will find a sympathetic judge in the argument that precedent be damned, clean-room implementations made with AI cannot be considered clean-room. It's too little too late to stop the rapid advancement of open tech, but that hasn't stopped courts from trying to send a message.

If these models will happily recite AGPL 3.0 code as “clean room implementations”, I’m glad that they will also help hobbyists experiment with software. Seems like a fair trade

Trained on the source code?

It's neural piracy, and I am down for it.

Why make a clone when you can redo the whole dang UI? Sure we'd like WYSIWYG, but I am absolutely convinced now in my 34th year of computer use that I don't want a GUI with a bunch of toolbars and dropdown menus that is laid out by decades of cruft accumulation.

LyX or some sort of Markdown editor with robust templates/stylesheets strikes me as more interesting.

A Word clone would have to offer something which Google Docs doesn't which is also what users want.

A distributed word processor that is not tied to the behemoth's cloud infrastructure and authentication walled garden would be very nice.

Like Collabora (https://www.collaboraonline.com/)? Which is distributed libreoffice you can self-host and also integrates with nextcloud.

I see that in addition to the Adobe suite apps last week, they've also now done implementations of excel/spreadsheet (https://github.com/storytold/gridcraft), powerpoint/presentation (https://github.com/storytold/deckcraft), and also AutoCAD/3D modeling (https://github.com/storytold/cadcraft) as well.

What a time to be alive.

"begun", none of those are done. Most are pre alpha state of you try them out.

Because nobody did before they published it. And they don't work.

Some work better then others though, eg my super basic usecases in photocraft mostly worked.

This reminds me of the shareware clones of popular PC software sold for $5/floppy through mail order and mall kiosks that failed to dominate the commercial software industry in the '80s.

The entire Excel implementation is under 50 MiB. The Word one is under 90MiB. Insane. I'm sure the features are incomplete, but both can open the normal Word and Excel files I have laying around. I'm extremely impressed.

Has anyone compiled a list of best practices for using AI agents to develop large software projects in Rust?

Types help: https://news.ycombinator.com/item?id=49806799

Agent memory helps: https://news.ycombinator.com/item?id=49948328

Storytold's projects include some: https://github.com/storytold/photocraft/blob/main/AGENTS.md

More?

Just create a blueprint with specs and architecture decisions, software cycle, with product and research docs, plus other needed docs like legal or such, and always use it to make the code. Don’t use memory because the way it gets called by the model isn’t good, and for any updates make the model update the specs too. That’s what I do and the output not just consistent but also future proof for updates or revisions or audits.

same. got hundreds of specs started from architecture decisions

most system design pattern for large projects applicable to rust 1:1

as for skills, did similar research 3 months ago, and couldn’t find anything comprehensive …

except this link https://github.com/leonardomso/rust-skills/tree/master/rules

ended up asking claude for code review 3 time and logging best practices in markdown gives compound effect

as a heavy temporal.io user i was never really bothered with async tokio rust gotchas

https://github.com/temporalio/sdk-rust

python libs written in rust under hood is a good source of rustish domain specific best practices too … if you are ok to see lot of “unsafe” code and clever tricks

Why not use LibreOffice which is already open source? I'm just trying to understand what we are solving here and why do we need a 1:1 of some other tool while existing tools which are much refined already exist?

I’d assume to try and be a simpler replacement for those already used to Word, like my father. He absolutely hates Word and Microsoft but he does not want to switch because he has been using Word for the past 20 years.

Of course, the quality of this is nowhere near good enough, but in theory :)

I have seen many claims about clean room implementations. I've seen the numerous refutations of reverse engineering binaries. I want to believe. But the pace is too coincidental imo. I know someone from Storycraft has been defending this in each thread and has acknowledged the lawsuit issue on other projects, but still, I'm hesitantly curious about patent violation considerations. Feel like I'm missing something here.

The part that interests me most about WordCraft is its use of imgui for the user interface. I hadn't considered imgui for a desktop UI so far but this project may change my mind.

To me, the key point is that this is the type of mass market disruption that AI can create. This is just the first of many that we'll see where the corporations are not the beneficiaries of AI.

How does one decide on the layer that each thing lives at architecture wise? Is it just that everything is one layer above whatever subsystems it depends on?

Pointless. LibreOffice exists. Trying to reverse engineer the closed source version is an exercise in futility.

Is it easier to make all these apps from scratch or use AI to supercharge an existing open source alternative?

Somewhere there is a team of lawyers high fiving each other. "Time to shine boys!"

Damn! I was in the process of doing this after I heard about Artcraft. They beat me to it!

Visio and Project FTW.

You'd rewrite the Bible in Rust at this point

Rich text editing and egui. Interesting choice

Why would people use this rather than existing alternatives with a more established community and maintainers?

Free.

And they work on Linux. A genius got MS Office working with Wine on Linux in a Nix Flake [1], but that's via a ton of Wine-glue.

These Artcraft apps seem to work on Linux completely fine, at least with the bit I've played with.

[1] https://news.ycombinator.com/item?id=49746401

By alternatives, I mean stuff like LibreOffice and its counterparts.

I can't speak for anyone else, but the reason I did the MS Office on Linux thing was because there is a subset of people who want the proper noun Microsoft Office and will use that as an excuse to not move to Linux. My parents, for example.

I don't really use MS Office for anything but Excel since I don't really do WYSIWYG document editing and even if I did I probably would use LibreOffice or OnlyOffice or something. LibreOffice Calc just kind of sucks as of right now.

> A genius got MS Office working with Wine on Linux in a Nix Flake

“Your scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should."

What is Word used for (or who uses it) nowadays? I don't remember the last time I typed anything on it.

lawyers at a bare minimum

Pretty much everyone who writes for a living and isn't doing "literature".

I.e. lawyers, accountants, clerks generally

AutoDesk is next ?

They already have a CAD thing, so I think that’s a safe bet.

https://github.com/storytold/cadcraft they already made one

This makes me so happy. Thank you!

Wow. This is insane.

Very nice!

Reply on news.ycombinator.com