[mg]

This is a great example of how useful the File System Access API is.

On http://co-do.xyz/ you can select a directory and let AI get to work inside of it without having to worry about side effects.

The Fily System Access API is the best thing that happened to the web in years. It makes web apps first class productivity applications.

[storystarling]

It definitely makes deployment cheaper, but I'm skeptical about relying on the browser for state management in longer chains. I tried this for a publishing tool and ended up migrating back to LangGraph and Celery just to ensure reliability. The infrastructure savings weren't worth the headache of handling edge cases on the client.

[layer8]

> It makes web apps first class productivity applications.

They won’t be first-class as long as native UI still has the upper hand.

[whywhywhywhy]

This battle is long won in favor of webtech in every realm but 3d/video editing/audio work/things that do gpu heavy lifting like game engines.

Outside those sort of spaces it’s hard to name a popular piece of software still on native that isn’t a wrapped webapp.

[layer8]

Microsoft Office apps, Java IDEs, text editors in general. Most of the time I spend with software is with non-web-UI applications.

To the extent that the battle has been won, the apps it has been won with are nevertheless second-class compared to native-level usability.

[whywhywhywhy]

> Microsoft Office apps

Thought these were web wrappers now if you use the latest

> text editors in general

Definitely not in general, VSCode and Cursor are both webtech and are extremely popular. Only terminal editors are native and then beyond that you have things like SublimeText, Textmate which are extremely niche now.

> Java IDEs

Yeah those and XCode I guess, Java IDE is extremely niche compared to webdev.

[Sammi]

> video editing

cough CapCut cough

[cobolexpert]

In which way does native UI have the upper hand, do you think? To me it seems like a lot of users are largely indifferent to this aspect (e.g. so many applications nowadays being Electron/browser based). If browsers keep gaining capabilities then it seems like this gap will get even smaller.

[TingPing]

I’ve never used a webapp that felt nicer than native software, it’s always very clearly a compromise.

[simonw]

I can't tell what's a web app and what's native these days. Are you sure you can?

[TingPing]

I'd have a very good hit rate, it mostly comes down to knowledge of toolkits. There are native apps that use their own toolkit, mostly written in Rust these days, and they always are worse than traditional toolkits (accessibility, respecting platform settings, visually fitting in, etc). That same issue applies to webapps typically.

[layer8]

The way keyboard-only usage works, if it is workable at all, is usually a dead giveaway. As is the lack of dialog windows and traditional menus, and often latency.

[auggierose]

Unfortunately, this feature of the API is not supported (yet?) either by Safari or Firefox.

[mmis1000]

The guarantee of web page never edit file on your disk(only create new ones) does not hold on this api though. I know it's what makes this api useful. But at the same time, there is big risk that user never expected this and results into giant security issue.

Firefox and safari are generally very conservative about new api that can enable new type of exploits.

At least firefox and safari does implement origin private file system. So, while you can't edit file on user disk directly. You can import the whole project into browser. Finish the edit and export it.

[cxr]

Browsers have had widespread support for processing files via drag-and-drop and the <input> element since HTML5 (< 2015). The last holdout on allowing the filepicker to accept a full directory (and its subdirectories, recursively—rather than 1 or N individual files) was Safari sometime around (before) 2020.

The Chrome team's new, experimental APIs are a separate matter. They provide additional capabilities, but many programs can get along just fine without since they don't don't strictly need them in order to work—if they would ever even have end up using them at all. A bunch of the applications in the original post fall into this category. You don't need new or novel APIs to be able to hash a file, for example. It's a developer education problem (see also: hubris).

[jefftk]

Providing a web app with edit access to a local directory is really needed for this to be usable. Without that you're constantly managing downloaded files and manually replacing things. I do think this is a case where the File System Access API shines.

[cxr]

> Providing a web app with edit access to a local directory is really needed for this to be usable.

"This" what? sha256sum doesn't need read-write access for even one file to be able to compute a hash, let alone a whole directory. You're ignoring most of my comment, focusing on like 20%, and in so doing, missing (and/or deliberately misframing) 100% of the point.

[jefftk]

We're talking about Simon's boosting of https://aifoc.us/the-browser-is-the-sandbox/ which is a prototype of Claude Cowork in the browser. That's what I'm saying needs read-write access.

[cxr]

> https://aifoc.us/the-browser-is-the-sandbox/ which is a prototype of Claude Cowork in the browser

Yup. That's the link, all right—the one we all read and that I'm citing examples from. Thanks for the reminder, I guess: it has been a whole 8 hours since I first looked at it.

What "we" are talking about here, in this subthread, is the fact that "Browsers have had widespread support for processing files" for a long, long time, and that although "Chrome team's new, experimental APIs [...] provide additional capabilities" which are undoubtedly useful for certain programs, they're overkill and don't offer anything new and/or strictly necessary for many, many programs that don't actually need that sort of access—including "A bunch of the applications in the original post [that] fall into this category. You don't need new or novel APIs to be able to hash a file, for example."

Which is to say, we're talking about POLP/POLA. And the point of my comment was to address the very worthwhile matter of POLA violations. But you seem insistent on shutting that discussion down with chatter that looks like it's an on-topic reply or refutation to something, but in reality doesn't actually meaningfully engage with what you're purporting to respond to, or at best comes come across as confused and not particularly attentive.

There are already and will continue to be plenty of opportunities to discuss the acknowledged upsides of the new APIs for the class of programs for which they are strictly necessary. There's a lot of them in this very comment section. It doesn't have to come at the expense of changing the subject in the middle of a different conversation—accompanied by undertones that you're putting some matter to rest.

[pjmlp]

Because it is a ChromeOS Platform API, not that it matters with the Chrome market share.